diff --git a/.gitattributes b/.gitattributes
index a6344aac8c09253b3b630fb776ae94478aa0275b..aa7aacd0134a92c3c1943fdecc75cd8b7420cce6 100644
--- a/.gitattributes
+++ b/.gitattributes
@@ -33,3 +33,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+model.safetensors.index.json filter=lfs diff=lfs merge=lfs -text
+tokenizer.json filter=lfs diff=lfs merge=lfs -text
diff --git a/README.md b/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/chat_template.jinja b/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0093efaa15b9ee3b0d8799ec64933fe0897b6687
--- /dev/null
+++ b/chat_template.jinja
@@ -0,0 +1,117 @@
+[gMASK]<sop>
+{%- if tools -%}
+{%- macro tool_to_json(tool) -%}
+    {%- set ns_tool = namespace(first=true) -%}
+    {{ '{' -}}
+    {%- for k, v in tool.items() -%}
+        {%- if k != 'defer_loading' and k != 'strict' -%}
+            {%- if not ns_tool.first -%}{{- ', ' -}}{%- endif -%}
+            {%- set ns_tool.first = false -%}
+            "{{ k }}": {{ v | tojson(ensure_ascii=False) }}
+        {%- endif -%}
+    {%- endfor -%}
+    {{- '}' -}}
+{%- endmacro -%}
+<|system|>
+# Tools
+
+You may call one or more functions to assist with the user query.
+
+You are provided with function signatures within <tools></tools> XML tags:
+<tools>
+{% for tool in tools %}
+{%- if 'function' in tool -%}
+    {%- set tool = tool['function'] -%}
+{%- endif -%}
+{% if tool.defer_loading is not defined or not tool.defer_loading %}
+{{ tool_to_json(tool) }}
+{% endif %}
+{% endfor %}
+</tools>
+
+For each function call, output the function name and arguments within the following XML format:
+<tool_call>{function-name}<arg_key>{arg-key-1}</arg_key><arg_value>{arg-value-1}</arg_value><arg_key>{arg-key-2}</arg_key><arg_value>{arg-value-2}</arg_value>...</tool_call>{%- endif -%}
+{%- macro visible_text(content) -%}
+    {%- if content is string -%}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping -%}
+        {%- for item in content -%}
+            {%- if item is mapping and item.type == 'text' -%}
+                {{- item.text }}
+            {%- elif item is string -%}
+                {{- item }}
+            {%- endif -%}
+        {%- endfor -%}
+    {%- else -%}
+        {{- content }}
+    {%- endif -%}
+{%- endmacro -%}
+{%- set ns = namespace(last_user_index=-1, thinking_indices='') -%}
+{%- for m in messages %}
+    {%- if m.role == 'user' %}
+        {%- set ns.last_user_index = loop.index0 -%}
+    {%- elif m.role == 'assistant' %}
+        {%- if m.reasoning_content is string %}
+            {%- set ns.thinking_indices = ns.thinking_indices ~ ',' ~ ns.last_user_index ~ ',' -%}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- set ns.has_thinking = false -%}
+{%- for m in messages -%}
+{%- if m.role == 'user' -%}<|user|>{{ visible_text(m.content) }}{% set ns.has_thinking = (',' ~ loop.index0 ~ ',') in ns.thinking_indices -%}
+{%- elif m.role == 'assistant' -%}
+<|assistant|>
+{%- set content = visible_text(m.content) %}
+{%- if m.reasoning_content is string %}
+    {%- set reasoning_content = m.reasoning_content %}
+{%- elif '</think>' in content %}
+    {%- set reasoning_content = content.split('</think>')[0].split('<think>')[-1] %}
+    {%- set content = content.split('</think>')[-1] %}
+{%- elif loop.index0 > ns.last_user_index and not (enable_thinking is defined and not enable_thinking) %}
+    {%- set reasoning_content = '' %}
+{%- elif loop.index0 < ns.last_user_index and ns.has_thinking %}
+    {%- set reasoning_content = '' %}
+{%- endif %}
+{%- if ((clear_thinking is defined and not clear_thinking) or loop.index0 > ns.last_user_index) and reasoning_content is defined -%}
+{{ '<think>' + reasoning_content +  '</think>'}}
+{%- else -%}
+{{ '</think>' }}
+{%- endif -%}
+{%- if content.strip() -%}
+{{ content.strip() }}
+{%- endif -%}
+{% if m.tool_calls %}
+{% for tc in m.tool_calls %}
+{%- if tc.function %}
+    {%- set tc = tc.function %}
+{%- endif %}
+{{- '<tool_call>' + tc.name -}}
+{% set _args = tc.arguments %}{% for k, v in _args.items() %}<arg_key>{{ k }}</arg_key><arg_value>{{ v | tojson(ensure_ascii=False) if v is not string else v }}</arg_value>{% endfor %}</tool_call>{% endfor %}
+{% endif %}
+{%- elif m.role == 'tool' -%}
+{%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+    {{- '<|observation|>' -}}
+{%- endif %}
+{%- if m.content is string -%}
+    {{- '<tool_response>' + m.content + '</tool_response>' -}}
+{%- else -%}
+    {{- '<tool_response><tools>\n' -}}
+    {% for tr in m.content %}
+        {%- for tool in tools -%}
+            {%- if 'function' in tool -%}
+                {%- set tool = tool['function'] -%}
+            {%- endif -%}
+            {%- if tool.name == tr.name -%}
+                {{- tool_to_json(tool) + '\n' -}}
+            {%- endif -%}
+        {%- endfor -%}
+    {%- endfor -%}
+    {{- '</tools></tool_response>' -}}
+{% endif -%}
+{%- elif m.role == 'system' -%}
+<|system|>{{ visible_text(m.content) }}
+{%- endif -%}
+{%- endfor -%}
+{%- if add_generation_prompt -%}
+    <|assistant|>{{- '</think>' if (enable_thinking is defined and not enable_thinking) else '<think>' -}}
+{%- endif -%}
\ No newline at end of file
diff --git a/config.json b/config.json
new file mode 100644
index 0000000000000000000000000000000000000000..2bfa53b86e60ef08e01fb978088709035fd8a536
--- /dev/null
+++ b/config.json
@@ -0,0 +1,2839 @@
+{
+  "architectures": [
+    "GlmMoeDsaForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 0,
+  "dtype": "bfloat16",
+  "eos_token_id": [
+    154820,
+    154827,
+    154829
+  ],
+  "ep_size": 1,
+  "first_k_dense_replace": 3,
+  "hidden_act": "silu",
+  "hidden_size": 6144,
+  "index_head_dim": 128,
+  "index_n_heads": 32,
+  "index_topk": 2048,
+  "indexer_rope_interleave": true,
+  "initializer_range": 0.02,
+  "intermediate_size": 12288,
+  "kv_lora_rank": 512,
+  "max_position_embeddings": 202752,
+  "mlp_layer_types": [
+    "dense",
+    "dense",
+    "dense",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse",
+    "sparse"
+  ],
+  "model_type": "glm_moe_dsa",
+  "moe_intermediate_size": 2048,
+  "moe_layer_freq": 1,
+  "n_group": 1,
+  "n_routed_experts": 256,
+  "n_shared_experts": 1,
+  "norm_topk_prob": true,
+  "num_attention_heads": 64,
+  "num_experts_per_tok": 8,
+  "num_hidden_layers": 78,
+  "num_key_value_heads": 64,
+  "num_nextn_predict_layers": 1,
+  "pad_token_id": 154820,
+  "pretraining_tp": 1,
+  "q_lora_rank": 2048,
+  "qk_head_dim": 256,
+  "qk_nope_head_dim": 192,
+  "qk_rope_head_dim": 64,
+  "quantization_config": {
+    "autoround_version": "0.13.0",
+    "bits": 4,
+    "data_type": "int",
+    "extra_config": {
+      ".*layers\\.0\\.mlp.*": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      ".*layers\\.1\\.mlp.*": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      ".*layers\\.2\\.mlp.*": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      ".*weights_proj.*": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.0.mlp.down_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.0.mlp.gate_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.0.mlp.up_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.0.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.0.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.0.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.0.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.0.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.0.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.0.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.0.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.1.mlp.down_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.1.mlp.gate_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.1.mlp.up_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.1.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.1.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.1.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.1.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.1.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.1.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.1.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.1.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.10.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.10.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.10.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.10.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.10.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.10.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.10.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.10.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.10.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.10.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.10.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.11.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.11.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.11.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.11.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.11.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.11.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.11.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.11.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.11.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.11.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.11.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.12.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.12.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.12.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.12.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.12.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.12.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.12.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.12.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.12.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.12.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.12.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.13.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.13.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.13.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.13.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.13.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.13.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.13.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.13.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.13.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.13.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.13.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.14.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.14.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.14.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.14.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.14.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.14.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.14.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.14.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.14.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.14.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.14.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.15.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.15.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.15.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.15.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.15.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.15.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.15.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.15.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.15.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.15.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.15.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.16.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.16.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.16.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.16.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.16.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.16.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.16.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.16.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.16.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.16.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.16.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.17.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.17.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.17.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.17.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.17.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.17.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.17.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.17.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.17.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.17.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.17.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.18.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.18.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.18.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.18.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.18.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.18.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.18.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.18.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.18.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.18.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.18.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.19.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.19.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.19.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.19.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.19.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.19.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.19.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.19.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.19.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.19.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.19.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.2.mlp.down_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.2.mlp.gate_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.2.mlp.up_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.2.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.2.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.2.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.2.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.2.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.2.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.2.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.2.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.20.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.20.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.20.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.20.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.20.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.20.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.20.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.20.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.20.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.20.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.20.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.21.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.21.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.21.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.21.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.21.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.21.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.21.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.21.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.21.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.21.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.21.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.22.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.22.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.22.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.22.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.22.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.22.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.22.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.22.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.22.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.22.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.22.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.23.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.23.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.23.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.23.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.23.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.23.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.23.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.23.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.23.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.23.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.23.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.24.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.24.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.24.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.24.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.24.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.24.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.24.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.24.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.24.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.24.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.24.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.25.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.25.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.25.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.25.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.25.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.25.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.25.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.25.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.25.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.25.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.25.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.26.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.26.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.26.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.26.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.26.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.26.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.26.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.26.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.26.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.26.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.26.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.27.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.27.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.27.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.27.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.27.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.27.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.27.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.27.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.27.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.27.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.27.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.28.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.28.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.28.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.28.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.28.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.28.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.28.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.28.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.28.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.28.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.28.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.29.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.29.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.29.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.29.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.29.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.29.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.29.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.29.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.29.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.29.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.29.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.3.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.3.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.3.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.3.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.3.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.3.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.3.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.3.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.3.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.3.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.3.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.30.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.30.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.30.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.30.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.30.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.30.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.30.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.30.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.30.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.30.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.30.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.31.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.31.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.31.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.31.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.31.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.31.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.31.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.31.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.31.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.31.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.31.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.32.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.32.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.32.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.32.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.32.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.32.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.32.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.32.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.32.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.32.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.32.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.33.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.33.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.33.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.33.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.33.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.33.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.33.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.33.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.33.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.33.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.33.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.34.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.34.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.34.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.34.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.34.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.34.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.34.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.34.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.34.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.34.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.34.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.35.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.35.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.35.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.35.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.35.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.35.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.35.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.35.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.35.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.35.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.35.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.36.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.36.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.36.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.36.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.36.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.36.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.36.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.36.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.36.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.36.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.36.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.37.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.37.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.37.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.37.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.37.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.37.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.37.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.37.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.37.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.37.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.37.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.38.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.38.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.38.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.38.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.38.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.38.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.38.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.38.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.38.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.38.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.38.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.39.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.39.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.39.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.39.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.39.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.39.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.39.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.39.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.39.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.39.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.39.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.4.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.4.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.4.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.4.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.4.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.4.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.4.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.4.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.4.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.4.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.4.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.40.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.40.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.40.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.40.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.40.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.40.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.40.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.40.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.40.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.40.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.40.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.41.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.41.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.41.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.41.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.41.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.41.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.41.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.41.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.41.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.41.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.41.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.42.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.42.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.42.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.42.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.42.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.42.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.42.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.42.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.42.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.42.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.42.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.43.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.43.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.43.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.43.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.43.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.43.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.43.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.43.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.43.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.43.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.43.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.44.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.44.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.44.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.44.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.44.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.44.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.44.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.44.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.44.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.44.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.44.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.45.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.45.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.45.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.45.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.45.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.45.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.45.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.45.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.45.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.45.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.45.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.46.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.46.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.46.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.46.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.46.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.46.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.46.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.46.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.46.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.46.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.46.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.47.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.47.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.47.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.47.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.47.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.47.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.47.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.47.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.47.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.47.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.47.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.48.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.48.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.48.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.48.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.48.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.48.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.48.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.48.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.48.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.48.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.48.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.49.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.49.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.49.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.49.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.49.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.49.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.49.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.49.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.49.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.49.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.49.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.5.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.5.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.5.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.5.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.5.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.5.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.5.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.5.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.5.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.5.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.5.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.50.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.50.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.50.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.50.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.50.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.50.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.50.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.50.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.50.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.50.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.50.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.51.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.51.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.51.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.51.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.51.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.51.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.51.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.51.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.51.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.51.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.51.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.52.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.52.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.52.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.52.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.52.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.52.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.52.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.52.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.52.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.52.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.52.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.53.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.53.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.53.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.53.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.53.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.53.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.53.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.53.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.53.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.53.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.53.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.54.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.54.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.54.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.54.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.54.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.54.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.54.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.54.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.54.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.54.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.54.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.55.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.55.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.55.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.55.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.55.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.55.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.55.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.55.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.55.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.55.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.55.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.56.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.56.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.56.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.56.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.56.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.56.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.56.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.56.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.56.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.56.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.56.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.57.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.57.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.57.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.57.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.57.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.57.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.57.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.57.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.57.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.57.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.57.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.58.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.58.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.58.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.58.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.58.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.58.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.58.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.58.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.58.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.58.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.58.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.59.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.59.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.59.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.59.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.59.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.59.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.59.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.59.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.59.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.59.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.59.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.6.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.6.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.6.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.6.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.6.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.6.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.6.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.6.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.6.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.6.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.6.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.60.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.60.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.60.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.60.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.60.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.60.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.60.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.60.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.60.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.60.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.60.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.61.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.61.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.61.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.61.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.61.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.61.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.61.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.61.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.61.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.61.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.61.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.62.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.62.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.62.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.62.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.62.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.62.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.62.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.62.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.62.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.62.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.62.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.63.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.63.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.63.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.63.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.63.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.63.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.63.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.63.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.63.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.63.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.63.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.64.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.64.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.64.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.64.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.64.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.64.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.64.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.64.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.64.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.64.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.64.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.65.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.65.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.65.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.65.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.65.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.65.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.65.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.65.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.65.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.65.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.65.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.66.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.66.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.66.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.66.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.66.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.66.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.66.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.66.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.66.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.66.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.66.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.67.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.67.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.67.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.67.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.67.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.67.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.67.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.67.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.67.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.67.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.67.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.68.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.68.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.68.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.68.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.68.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.68.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.68.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.68.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.68.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.68.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.68.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.69.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.69.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.69.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.69.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.69.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.69.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.69.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.69.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.69.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.69.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.69.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.7.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.7.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.7.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.7.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.7.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.7.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.7.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.7.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.7.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.7.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.7.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.70.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.70.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.70.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.70.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.70.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.70.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.70.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.70.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.70.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.70.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.70.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.71.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.71.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.71.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.71.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.71.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.71.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.71.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.71.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.71.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.71.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.71.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.72.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.72.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.72.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.72.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.72.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.72.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.72.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.72.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.72.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.72.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.72.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.73.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.73.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.73.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.73.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.73.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.73.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.73.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.73.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.73.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.73.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.73.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.74.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.74.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.74.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.74.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.74.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.74.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.74.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.74.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.74.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.74.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.74.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.75.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.75.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.75.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.75.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.75.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.75.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.75.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.75.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.75.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.75.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.75.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.76.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.76.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.76.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.76.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.76.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.76.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.76.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.76.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.76.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.76.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.76.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.77.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.77.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.77.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.77.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.77.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.77.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.77.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.77.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.77.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.77.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.77.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.8.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.8.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.8.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.8.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.8.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.8.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.8.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.8.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.8.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.8.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.8.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.9.mlp.shared_experts.down_proj": {
+        "bits": 8
+      },
+      "model.layers.9.mlp.shared_experts.gate_proj": {
+        "bits": 8
+      },
+      "model.layers.9.mlp.shared_experts.up_proj": {
+        "bits": 8
+      },
+      "model.layers.9.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "float"
+      },
+      "model.layers.9.self_attn.indexer.wk": {
+        "bits": 8
+      },
+      "model.layers.9.self_attn.indexer.wq_b": {
+        "bits": 8
+      },
+      "model.layers.9.self_attn.kv_a_proj_with_mqa": {
+        "bits": 8
+      },
+      "model.layers.9.self_attn.kv_b_proj": {
+        "bits": 8
+      },
+      "model.layers.9.self_attn.o_proj": {
+        "bits": 8
+      },
+      "model.layers.9.self_attn.q_a_proj": {
+        "bits": 8
+      },
+      "model.layers.9.self_attn.q_b_proj": {
+        "bits": 8
+      },
+      "model.layers.78.mlp.gate": {
+        "bits": 16,
+        "data_type": "fp"
+      },
+      "model.layers.78.self_attn.indexer.weights_proj": {
+        "bits": 16,
+        "data_type": "fp"
+      }
+    },
+    "group_size": 128,
+    "iters": 0,
+    "packing_format": "auto_round:auto_gptq",
+    "quant_method": "auto-round",
+    "sym": true,
+    "block_name_to_quantize": [
+      "model.layers"
+    ]
+  },
+  "rms_norm_eps": 1e-05,
+  "rope_interleave": true,
+  "rope_parameters": {
+    "rope_theta": 1000000,
+    "rope_type": "default"
+  },
+  "routed_scaling_factor": 2.5,
+  "scoring_func": "sigmoid",
+  "tie_word_embeddings": false,
+  "topk_group": 1,
+  "topk_method": "noaux_tc",
+  "transformers_version": "5.4.0",
+  "use_cache": true,
+  "v_head_dim": 256,
+  "vocab_size": 154880
+}
\ No newline at end of file
diff --git a/generation_config.json b/generation_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..ebfbae00e215f99095939ebe38c70bc93d1165df
--- /dev/null
+++ b/generation_config.json
@@ -0,0 +1,13 @@
+{
+  "_from_model_config": true,
+  "do_sample": true,
+  "eos_token_id": [
+    154820,
+    154827,
+    154829
+  ],
+  "pad_token_id": 154820,
+  "temperature": 1.0,
+  "top_p": 0.95,
+  "transformers_version": "5.4.0"
+}
diff --git a/model-00001-of-00075.safetensors b/model-00001-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..c0971d312e48cb907222ee1de5c689f1001371e2
--- /dev/null
+++ b/model-00001-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3abae9a7f485f43691fcf7f6d6615340ebd6cae24a1390166f24a33ad31ebe84
+size 5368688176
diff --git a/model-00002-of-00075.safetensors b/model-00002-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..060301e0b4f586d4d11e83fdca8403ef04c33613
--- /dev/null
+++ b/model-00002-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:811fb060d00f21257226be35ca2658d5b9f09f4f2750f4a582acbec30cd429d7
+size 5365562632
diff --git a/model-00003-of-00075.safetensors b/model-00003-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..a1b9593a88abb505dc7309a8ec086b7d98a771b1
--- /dev/null
+++ b/model-00003-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c2211dfb9d98139969d37f2fa43d67a043ec15406038b4a36a62b550ce8abd81
+size 5365562624
diff --git a/model-00004-of-00075.safetensors b/model-00004-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..13c3fbec323a60b59855707463cb9451c4834c6d
--- /dev/null
+++ b/model-00004-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9803461fc68de67edb1b1daec21061984362b4a8d68ea49416170169ea19ac47
+size 5365562632
diff --git a/model-00005-of-00075.safetensors b/model-00005-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..8c5020f8446ac75c34a201b23783d3895104e5f3
--- /dev/null
+++ b/model-00005-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2cbf72e8e4ef69c1f47645ff4362dd6982485710945c5b481398b644d7acfcae
+size 5365562632
diff --git a/model-00006-of-00075.safetensors b/model-00006-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..a125c8d54232ecb73da17f768f390761fda70155
--- /dev/null
+++ b/model-00006-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b9e8088d7e1f4579eabfc8f6251c5cee53c4294b6c889cd7d86f9c5672b6db68
+size 5365562624
diff --git a/model-00007-of-00075.safetensors b/model-00007-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..82fb6085d7883d6cfc038c6c81f2a802413d353b
--- /dev/null
+++ b/model-00007-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ec63625917b57777d6d9c02af3c115d800c2188d1edb13ebb34025174160c96b
+size 5365562632
diff --git a/model-00008-of-00075.safetensors b/model-00008-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..40b1e7e5683c3b08ec3fd06c3f246b9da3f6e738
--- /dev/null
+++ b/model-00008-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7dd89b34bdb47de415d1040f630b5a05e9a9e18d350eda502767785c27024164
+size 5365564568
diff --git a/model-00009-of-00075.safetensors b/model-00009-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..53ba3b97c0cf926f76fbf6e7e206f6de150dfeed
--- /dev/null
+++ b/model-00009-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fc55f84d8e701fe0496a0f23c34d0a9bf8fe342352bd374e7dc1e91e401f167e
+size 5365565040
diff --git a/model-00010-of-00075.safetensors b/model-00010-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..e224bdbd2eeb697052b5104c5d785f82900f5581
--- /dev/null
+++ b/model-00010-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:812295dc6cd2d21b422ba7c15d03e7a398b2c34d50d82404636e002bef8f2a95
+size 5365565048
diff --git a/model-00011-of-00075.safetensors b/model-00011-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..894355ba19a225a8849667f9cd0fd48579e68a2b
--- /dev/null
+++ b/model-00011-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:486213678345d3a27fd0a4e6bac1972bfe8d380b430bfc72f6015319fac4e971
+size 5365565048
diff --git a/model-00012-of-00075.safetensors b/model-00012-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..9b17705e3905ac254c4d341a4e8a5626846f8a2e
--- /dev/null
+++ b/model-00012-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7f68011f12624ec7294c273904ddd4df98a1002c81f7599e7633689beb9c880a
+size 5365565040
diff --git a/model-00013-of-00075.safetensors b/model-00013-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..aaacaea1c29352dd5ae2bb01fd05d404d407c40c
--- /dev/null
+++ b/model-00013-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:48adac517365a2ca6a8c1f5a7524ecec164ad752ce4337f91d1fe19127b18195
+size 5365565048
diff --git a/model-00014-of-00075.safetensors b/model-00014-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..ca026f944da0544f98b36abf851c5e92a7b5fdfc
--- /dev/null
+++ b/model-00014-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:72b448add7877c93da80a139c8043f30942ea72d7c2b8be08ee04af7614c6f6c
+size 5365565048
diff --git a/model-00015-of-00075.safetensors b/model-00015-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..75d3d9971c3a6ffbbc369b5c0337d9a8b966ced2
--- /dev/null
+++ b/model-00015-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ccdabf94269f794c8706a245b77e0e9eff46a0120b9d663cf78d4d78e97b6b56
+size 5368515040
diff --git a/model-00016-of-00075.safetensors b/model-00016-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..e555735889b8d623319a6d65470ba8ff97455aee
--- /dev/null
+++ b/model-00016-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d6cc9c53205684d0312e13273c3ef3cfa83fac45305bf2ce4d702a55c9b712c2
+size 5333003624
diff --git a/model-00017-of-00075.safetensors b/model-00017-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..175d7ed3c28df67d9a94c6ca209f56e47206bcb3
--- /dev/null
+++ b/model-00017-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b8277bf4b5b71152264e8cf3c36937954f983bb97f7c15422dd988971a89e697
+size 5367233880
diff --git a/model-00018-of-00075.safetensors b/model-00018-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..b14dbdde45528ae894e9f02979309d6996a3bdba
--- /dev/null
+++ b/model-00018-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2aa6322300f1ed1c1ad647a49d0e866d8d9149ce2017787d018468aad0fb0924
+size 5365564920
diff --git a/model-00019-of-00075.safetensors b/model-00019-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..4156a0384e910c1222afaba2dce48f0ac690f973
--- /dev/null
+++ b/model-00019-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:844d63a3542fe1d0fc554f34391891e7e8aeeb3b607e0df58970aaf29caa5885
+size 5365564952
diff --git a/model-00020-of-00075.safetensors b/model-00020-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..9e95edcaaa264b127e9f7776fe10966a6afa7f2c
--- /dev/null
+++ b/model-00020-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:28ac57596a415490b6395949b04748a10a7c34f8550a5eca7110cee9837cb8e6
+size 5365564968
diff --git a/model-00021-of-00075.safetensors b/model-00021-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..55612c3d167bfc1bf6ed6cc65dc807e0463cc0e4
--- /dev/null
+++ b/model-00021-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b5e9322c91e504018edc1eaf98bc8f12e89ff65d8185e4e7d8b517e316f276d5
+size 5365564976
diff --git a/model-00022-of-00075.safetensors b/model-00022-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..92303f818fbd75a5543563dbfd08c5b1f675cce0
--- /dev/null
+++ b/model-00022-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ff9abc96d3b07f604dff23668e5e2b25b613c957dc75173a780e61057284d9e9
+size 5365564976
diff --git a/model-00023-of-00075.safetensors b/model-00023-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..b59f71cd395622f8dbb8ab0b1f0cf438287d51ad
--- /dev/null
+++ b/model-00023-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8898c99133a028a488eacfff87f363caaf918576a06d9f29eb95310b3e21384d
+size 5365564968
diff --git a/model-00024-of-00075.safetensors b/model-00024-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..d2a587d199f572e7112904b3b9000b09259d9012
--- /dev/null
+++ b/model-00024-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dd6c3d794cac63b8d20058438428ff10e8c99e91e768c72dbdf9b846c5d3d473
+size 5365564976
diff --git a/model-00025-of-00075.safetensors b/model-00025-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..0a8d48168a402cdfcba777ac63685d5d4afcbabf
--- /dev/null
+++ b/model-00025-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:51e7cfc3c52229c749822117a276ebb2760fb9d83e9e1c6ebce3e262cc9db7ed
+size 5365564976
diff --git a/model-00026-of-00075.safetensors b/model-00026-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..ec3918068f0a567ae123600f4de047434c405b58
--- /dev/null
+++ b/model-00026-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e6ce74b934bce565399fab2918027ddd184cabeab1eff1d663f7e3ead36052c4
+size 5365564968
diff --git a/model-00027-of-00075.safetensors b/model-00027-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..0a287c8c5b8a91945eb9eeb12f7fbe0b62c83714
--- /dev/null
+++ b/model-00027-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b9230049da8accd29f58b7e3bed2abae63f44f0cf02a5bda56588b57485a956
+size 5365564976
diff --git a/model-00028-of-00075.safetensors b/model-00028-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..026e090d64da28fdddbf19793d88b5d1ff617d35
--- /dev/null
+++ b/model-00028-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b1752884eb7c6b1dd6894715ce767fe86c267bc075ce918a349e005c9932841
+size 5365564976
diff --git a/model-00029-of-00075.safetensors b/model-00029-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..b45078d86aff37f1014ed913102171bfea702c04
--- /dev/null
+++ b/model-00029-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6c0d17423b427e4aa8c73c17d97d26d3414858833971dbf0071938efaf794eeb
+size 5365564968
diff --git a/model-00030-of-00075.safetensors b/model-00030-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..94011fac2ab681090cc5acf2918ab86da1ecbdea
--- /dev/null
+++ b/model-00030-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:07211a77e3332004912734b3f2cd6361ebff2dc5c759ea44d3e8e3a23bd1816a
+size 5365564976
diff --git a/model-00031-of-00075.safetensors b/model-00031-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..6ad67a240c308936ba8bdadd25384cc8ac23c35e
--- /dev/null
+++ b/model-00031-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:80deb888b7718e3a2bc8898f35633832084e1ab9fa957dc5b51bb719354c5524
+size 5365564976
diff --git a/model-00032-of-00075.safetensors b/model-00032-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..d10f5b3c7ee297ba6a53da1f91205525a32a722e
--- /dev/null
+++ b/model-00032-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a7114e5eb37ff04de14ecac90117b54d946d6e50e33f7cef359e9e3b49defa39
+size 5365564968
diff --git a/model-00033-of-00075.safetensors b/model-00033-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..fbb835c2e8b4f7c3d702ff563ece55579c0e90dc
--- /dev/null
+++ b/model-00033-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4483ba090cdd3a230c983fc1f377ac7dbaee1d5ed30721bffd3dd9ed0a3fd59e
+size 5365564992
diff --git a/model-00034-of-00075.safetensors b/model-00034-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..bcba2b54d6815f532ee822ff78af3d34768f7fe8
--- /dev/null
+++ b/model-00034-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cf6003585a116927d18522b18f89b417d38e8fe465f60c890fe37753b5d5d846
+size 5365565032
diff --git a/model-00035-of-00075.safetensors b/model-00035-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..2d8e1a2d9b146276565682a1b493260cb0d4b1d1
--- /dev/null
+++ b/model-00035-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:42893ee9b9046f1777fce11674be8ac6b5214e507d7ba9295c8424e3fec53383
+size 5365565024
diff --git a/model-00036-of-00075.safetensors b/model-00036-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..e4ee71c7e6287baa891b28c5727d9a62c8a99cc1
--- /dev/null
+++ b/model-00036-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aa3d30b4f3111e71cba4e9fdeaf21a1b82fb0e0c60694eef40b2cb2da6d787fd
+size 5365565032
diff --git a/model-00037-of-00075.safetensors b/model-00037-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..24340a51f504882c72bb73c40366d3596a057e34
--- /dev/null
+++ b/model-00037-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:079b1d0dc9bb5e09e0d9c9aaa1c851d6e54fa78f726bda586afe0c8e4a2ce86e
+size 5365565032
diff --git a/model-00038-of-00075.safetensors b/model-00038-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..8359ce472e067062304223d81b39b1d5590664a7
--- /dev/null
+++ b/model-00038-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:53aa1a901f7668c805b32290fb6b12b84cb130a9feac4f6dcc44c3de2aecae20
+size 5365565024
diff --git a/model-00039-of-00075.safetensors b/model-00039-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..7ab0d5bb215772f46da24c125bff1f0a2b572a58
--- /dev/null
+++ b/model-00039-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ce6782494a0e95cbf884c3e46ba34628f7e59933597545e1f2ac2fc1c17177d
+size 5365565032
diff --git a/model-00040-of-00075.safetensors b/model-00040-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..6838958db6c9e9f1dc28d81172e1621cf855e552
--- /dev/null
+++ b/model-00040-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:630a428b4ce8714f05bd09bad190f98a983393fbde07410a0580426df6a841d2
+size 5365565032
diff --git a/model-00041-of-00075.safetensors b/model-00041-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..74bfa04da872a2fe52da506aa2be0a67c80bfceb
--- /dev/null
+++ b/model-00041-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9fa25b73a8779ecdfd595f3ff850ac7aa0c6d15477a8772a62b4aa3d65c61e44
+size 5365565024
diff --git a/model-00042-of-00075.safetensors b/model-00042-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..8b0f89f83ce6678d2b40a047b0b98452fcb8a520
--- /dev/null
+++ b/model-00042-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a7c681f82a0f42547d01191ddf7478d305725c3cf9689df8016b50695c67439f
+size 5365565032
diff --git a/model-00043-of-00075.safetensors b/model-00043-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..2e0244dc5f82753808187d75bb0f7ae7e7228423
--- /dev/null
+++ b/model-00043-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0a6c3284e9c31940f4ae1ff4611862a35cb357594c1a34a6ef7fb4105909189e
+size 5365565032
diff --git a/model-00044-of-00075.safetensors b/model-00044-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..d2a72601c7586869d0e47009ab1c1b8902bc261c
--- /dev/null
+++ b/model-00044-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8800dc8738568b5f7eb2b8f56525381cbd0ac7daaaf93374c1d52a2db5ef20e8
+size 5365565024
diff --git a/model-00045-of-00075.safetensors b/model-00045-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..0203f35bd92f1057786a5c6b6f045b42a7df45b9
--- /dev/null
+++ b/model-00045-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc07e23395c3d3d30fe8352b4687676420c778d0e2ab6e6f476657d3293e08a7
+size 5365565032
diff --git a/model-00046-of-00075.safetensors b/model-00046-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..d9520beb531fd57a02f6783a4440cb32be88fb98
--- /dev/null
+++ b/model-00046-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4937ec79c9c2a0820fc47e336ec888140e111f959d4fa98be6435828881439ea
+size 5365565032
diff --git a/model-00047-of-00075.safetensors b/model-00047-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..db9e0b05a9123d256ec218c2b0c02d44a7521cd6
--- /dev/null
+++ b/model-00047-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bb1b0983a20c9471840da1996ac74e7ae320b4a19b1915107975d70a53d7d5ba
+size 5365565024
diff --git a/model-00048-of-00075.safetensors b/model-00048-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..c40bbdcd4ba99dc433cb06d52b00d57ed473d7f0
--- /dev/null
+++ b/model-00048-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1dd3376c3e34a24aab91ba17d45415ce850cd35b57b1832e293003895753a2f1
+size 5365565032
diff --git a/model-00049-of-00075.safetensors b/model-00049-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..c3fff411cefc784172504d12e0fda394b383b6f2
--- /dev/null
+++ b/model-00049-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:24fae85693e88da30a3b96fc730271335e4a9674900f899f1b8d24c139c581b6
+size 5365565032
diff --git a/model-00050-of-00075.safetensors b/model-00050-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..4521c1d1f95f0a5e33de764b8bc55bf564746aa7
--- /dev/null
+++ b/model-00050-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:89287d7d3a39e8260f2394d702983a8be9212c5887aecc93b713d1b3c189e967
+size 5365565024
diff --git a/model-00051-of-00075.safetensors b/model-00051-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..55e873c02e7b51b93f3de052f52f76abfc99cf5f
--- /dev/null
+++ b/model-00051-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8bf7ec6a853000a24483a3195b4ef225b859695fb7d8b5b43cbf4879f55c7da1
+size 5365565048
diff --git a/model-00052-of-00075.safetensors b/model-00052-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..3c4935bc5498daf89cc62426ca9a234d37a0bebe
--- /dev/null
+++ b/model-00052-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:58e5a7930c05cea7fafc60726f3c1bde530c9935670e85b5700dc302e4c4f284
+size 5365565048
diff --git a/model-00053-of-00075.safetensors b/model-00053-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..6b5677ffb448eea072b83f97dee5dc65d38447ec
--- /dev/null
+++ b/model-00053-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fff8f90831b5d1c897fe914fbeeca90846675582e5110a245d21b33379e74902
+size 5365565040
diff --git a/model-00054-of-00075.safetensors b/model-00054-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..7d30677c90e1144d761aadb5d0b770ab5bd8a55e
--- /dev/null
+++ b/model-00054-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:092fc03aa5d1cdbe044b85d990384a01aa05a5ffd1214af4edcbeb5d67b0b211
+size 5365565048
diff --git a/model-00055-of-00075.safetensors b/model-00055-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..775fd6e748a1b98053d43504eccf674d6c41f1eb
--- /dev/null
+++ b/model-00055-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9922645eb05893f3cfa5a68782ff52271a788e26d0cb36b23f9637e7ba3c10c6
+size 5365565048
diff --git a/model-00056-of-00075.safetensors b/model-00056-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..a25229b5945ebc8fa12c16176571c9239186dc6e
--- /dev/null
+++ b/model-00056-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:713632dfff9d034ef20cdeb90c39f0117e671d5a047993418b1db18d1146a77d
+size 5365565040
diff --git a/model-00057-of-00075.safetensors b/model-00057-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..ea3a7ec56e4c3bdfe435a8610d7c10af56b88d84
--- /dev/null
+++ b/model-00057-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7e8f60e65f6dadd9c047126013a8894aa46f9af3585dece6e07c7276710bcd69
+size 5365565048
diff --git a/model-00058-of-00075.safetensors b/model-00058-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..bfb6b5b4a25d264d49061243061e79c285f136cc
--- /dev/null
+++ b/model-00058-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a9f84dfa12ed3de67657a830231a9a1ca6deb1a7302026a2fba4bc35ae569e87
+size 5368819760
diff --git a/model-00059-of-00075.safetensors b/model-00059-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..16e6a8de20b10899f96d39aa98f0508ea74e3252
--- /dev/null
+++ b/model-00059-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5857de49bfa1308018f7f9ffa924f9e9e1d88d8505cf549c85a266ccef3bf24b
+size 5363808384
diff --git a/model-00060-of-00075.safetensors b/model-00060-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..3784a4abd091e9d91423c8b1f80de1dcd53917a4
--- /dev/null
+++ b/model-00060-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9ac99b198742bb8483c104532c0061f00b65d0467c2aae58d1b4ffa465c5f653
+size 5368812232
diff --git a/model-00061-of-00075.safetensors b/model-00061-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..755f2da52cf9feb1645c20626f0d13dae93dc543
--- /dev/null
+++ b/model-00061-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b3a6a981ec27f1632bdaa853fe87717247c0d298505061ef6bfebb9ffa19972a
+size 5365564944
diff --git a/model-00062-of-00075.safetensors b/model-00062-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..d7a689a06f7ea6ee02c5cec2298a4a7ef9ae13fe
--- /dev/null
+++ b/model-00062-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5f6d3f45ad1b43d1f14df58bf538214702171b818a25e2e71d2672b4984554f
+size 5365564968
diff --git a/model-00063-of-00075.safetensors b/model-00063-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..ddc0514e041cba30c3cbc4afe581dfb9d9681937
--- /dev/null
+++ b/model-00063-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a7fb85db3ffadc566b2e0358357fb8d8cdbc006c5a546780f82a1fb8dad15ac9
+size 5365564976
diff --git a/model-00064-of-00075.safetensors b/model-00064-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..b3cdd9b4c4d337742856b0370025e47a7443bd17
--- /dev/null
+++ b/model-00064-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8da920892ddeabb957b6f7b0dcc8cbe2ec1f510065fc6623b606e150f2f6c72b
+size 5365564976
diff --git a/model-00065-of-00075.safetensors b/model-00065-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..06f421b0a06e094d8e73a54f4b44c7b82339bdbf
--- /dev/null
+++ b/model-00065-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:144c03df6390cec24d94d68faee1f61652ea0231d1aed37a0655d6977942203f
+size 5365564968
diff --git a/model-00066-of-00075.safetensors b/model-00066-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..e99c527cdd568221ca484c2a03ad077dca2e43d1
--- /dev/null
+++ b/model-00066-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0fae1ac538bf79cdee46b1edce36080ca9aa64399dad391ab735d9c6b7c79d24
+size 5365564976
diff --git a/model-00067-of-00075.safetensors b/model-00067-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..3034e4fc2801b0c43e89453bb3ba8408dbcf354b
--- /dev/null
+++ b/model-00067-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:91d85ba79f1894d573f1b915cbff14809087ec5212b5ec8147e7bc60a5611cd3
+size 5365564976
diff --git a/model-00068-of-00075.safetensors b/model-00068-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..59202268b2baf35c12d2e008b39c089ebeb5aa4c
--- /dev/null
+++ b/model-00068-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9b2fd1e3d19245bd4507149d8810a4e3ac52a712fcad095e8a526d827b6c4c76
+size 5365564968
diff --git a/model-00069-of-00075.safetensors b/model-00069-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..48fa5ee434dd1be9b370d086d5c41ca466dbc2c4
--- /dev/null
+++ b/model-00069-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3d4743513c449880a02ec7cf0c059e366c1f7294d3851998db7c4d9b7ce7c5a4
+size 5365564976
diff --git a/model-00070-of-00075.safetensors b/model-00070-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..fdbfabab2656912df4c4bdfd7461ae8fb6b0e246
--- /dev/null
+++ b/model-00070-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:77a7ba10bbf15c7cb2e7008b6127363dd7d7f0f447eff07cbf7044ba7c242087
+size 5365564976
diff --git a/model-00071-of-00075.safetensors b/model-00071-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..3761d4f441115dd28f78d5d79b7d61747d99f76a
--- /dev/null
+++ b/model-00071-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e190a8a508c822097075f67d817da630927843629ebae784ef1862a90ed2f385
+size 5365564968
diff --git a/model-00072-of-00075.safetensors b/model-00072-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..9b30ebcabb7fae77c1a8f479033c10715d14c555
--- /dev/null
+++ b/model-00072-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:769500afa23e7069af7d4d391dbb918c55bc02edc526f1fbc4790d68c8db4d1e
+size 5365564976
diff --git a/model-00073-of-00075.safetensors b/model-00073-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..fb1cfb06de3b3cda45da7b33d4e95be4bde25590
--- /dev/null
+++ b/model-00073-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0306f7698a88f376f6581c1bbec824b0a1605da2315d961609f25e54d4e99629
+size 5365564976
diff --git a/model-00074-of-00075.safetensors b/model-00074-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..272c02a0eb11f644f84cb2c264417b2eccc8b633
--- /dev/null
+++ b/model-00074-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0beb8928569d6ce04f169e097e53401cba68d61f03c81dea530f5e136a069b3f
+size 5233389784
diff --git a/model-00075-of-00075.safetensors b/model-00075-of-00075.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..9c65588da5e0526629b1236b935b60ade8929331
--- /dev/null
+++ b/model-00075-of-00075.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:850a1c490fa492daacf425e4a437d7e96801f585ed14691ad0bf5662d965b6b3
+size 1903165536
diff --git a/model.safetensors.index.json b/model.safetensors.index.json
new file mode 100644
index 0000000000000000000000000000000000000000..0375e79dc86ff40e0f9ef1c527a9ae93dda053df
--- /dev/null
+++ b/model.safetensors.index.json
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f7b8e22fe7ca310637387efc6bf1401bfba91098e03e242c71c6c69e0800d8d
+size 16231052
diff --git a/model_extra_tensors.safetensors b/model_extra_tensors.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..5335e352d4baaa20359c02fb3ae4d8793de4592d
--- /dev/null
+++ b/model_extra_tensors.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:673cee42724fe62d7fcd57bf7119bf3f66380f39ffb9ea07ea6f5c8c0d8b6e7e
+size 5173812096
diff --git a/quantization_config.json b/quantization_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..fd0ee90dcdc1a464b5344dff37f19e10da8f01c2
--- /dev/null
+++ b/quantization_config.json
@@ -0,0 +1,2689 @@
+{
+  "bits": 4,
+  "data_type": "int",
+  "group_size": 128,
+  "sym": true,
+  "iters": 0,
+  "autoround_version": "0.13.0",
+  "quant_method": "auto-round",
+  "packing_format": "auto_round:auto_gptq",
+  "extra_config": {
+    "model.layers.0.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.0.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.0.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.0.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.0.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.0.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.0.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.0.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.0.mlp.gate_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.0.mlp.up_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.0.mlp.down_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.1.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.1.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.1.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.1.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.1.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.1.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.1.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.1.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.1.mlp.gate_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.1.mlp.up_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.1.mlp.down_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.2.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.2.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.2.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.2.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.2.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.2.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.2.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.2.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.2.mlp.gate_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.2.mlp.up_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.2.mlp.down_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.3.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.3.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.3.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.3.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.3.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.3.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.3.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.3.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.3.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.3.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.3.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.4.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.4.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.4.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.4.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.4.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.4.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.4.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.4.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.4.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.4.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.4.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.5.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.5.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.5.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.5.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.5.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.5.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.5.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.5.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.5.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.5.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.5.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.6.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.6.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.6.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.6.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.6.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.6.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.6.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.6.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.6.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.6.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.6.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.7.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.7.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.7.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.7.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.7.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.7.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.7.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.7.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.7.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.7.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.7.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.8.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.8.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.8.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.8.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.8.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.8.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.8.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.8.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.8.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.8.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.8.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.9.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.9.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.9.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.9.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.9.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.9.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.9.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.9.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.9.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.9.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.9.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.10.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.10.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.10.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.10.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.10.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.10.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.10.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.10.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.10.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.10.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.10.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.11.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.11.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.11.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.11.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.11.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.11.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.11.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.11.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.11.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.11.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.11.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.12.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.12.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.12.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.12.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.12.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.12.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.12.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.12.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.12.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.12.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.12.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.13.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.13.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.13.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.13.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.13.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.13.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.13.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.13.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.13.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.13.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.13.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.14.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.14.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.14.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.14.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.14.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.14.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.14.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.14.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.14.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.14.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.14.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.15.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.15.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.15.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.15.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.15.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.15.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.15.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.15.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.15.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.15.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.15.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.16.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.16.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.16.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.16.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.16.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.16.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.16.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.16.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.16.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.16.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.16.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.17.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.17.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.17.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.17.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.17.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.17.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.17.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.17.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.17.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.17.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.17.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.18.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.18.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.18.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.18.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.18.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.18.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.18.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.18.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.18.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.18.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.18.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.19.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.19.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.19.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.19.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.19.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.19.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.19.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.19.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.19.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.19.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.19.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.20.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.20.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.20.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.20.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.20.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.20.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.20.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.20.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.20.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.20.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.20.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.21.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.21.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.21.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.21.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.21.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.21.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.21.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.21.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.21.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.21.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.21.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.22.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.22.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.22.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.22.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.22.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.22.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.22.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.22.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.22.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.22.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.22.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.23.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.23.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.23.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.23.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.23.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.23.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.23.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.23.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.23.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.23.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.23.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.24.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.24.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.24.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.24.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.24.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.24.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.24.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.24.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.24.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.24.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.24.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.25.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.25.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.25.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.25.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.25.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.25.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.25.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.25.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.25.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.25.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.25.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.26.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.26.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.26.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.26.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.26.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.26.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.26.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.26.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.26.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.26.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.26.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.27.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.27.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.27.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.27.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.27.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.27.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.27.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.27.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.27.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.27.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.27.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.28.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.28.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.28.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.28.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.28.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.28.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.28.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.28.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.28.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.28.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.28.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.29.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.29.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.29.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.29.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.29.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.29.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.29.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.29.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.29.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.29.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.29.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.30.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.30.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.30.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.30.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.30.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.30.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.30.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.30.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.30.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.30.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.30.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.31.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.31.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.31.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.31.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.31.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.31.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.31.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.31.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.31.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.31.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.31.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.32.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.32.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.32.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.32.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.32.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.32.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.32.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.32.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.32.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.32.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.32.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.33.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.33.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.33.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.33.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.33.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.33.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.33.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.33.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.33.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.33.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.33.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.34.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.34.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.34.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.34.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.34.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.34.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.34.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.34.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.34.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.34.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.34.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.35.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.35.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.35.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.35.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.35.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.35.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.35.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.35.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.35.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.35.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.35.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.36.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.36.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.36.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.36.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.36.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.36.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.36.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.36.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.36.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.36.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.36.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.37.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.37.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.37.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.37.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.37.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.37.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.37.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.37.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.37.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.37.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.37.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.38.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.38.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.38.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.38.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.38.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.38.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.38.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.38.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.38.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.38.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.38.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.39.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.39.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.39.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.39.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.39.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.39.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.39.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.39.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.39.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.39.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.39.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.40.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.40.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.40.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.40.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.40.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.40.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.40.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.40.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.40.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.40.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.40.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.41.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.41.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.41.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.41.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.41.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.41.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.41.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.41.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.41.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.41.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.41.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.42.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.42.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.42.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.42.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.42.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.42.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.42.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.42.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.42.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.42.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.42.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.43.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.43.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.43.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.43.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.43.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.43.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.43.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.43.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.43.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.43.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.43.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.44.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.44.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.44.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.44.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.44.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.44.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.44.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.44.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.44.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.44.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.44.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.45.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.45.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.45.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.45.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.45.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.45.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.45.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.45.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.45.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.45.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.45.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.46.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.46.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.46.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.46.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.46.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.46.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.46.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.46.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.46.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.46.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.46.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.47.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.47.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.47.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.47.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.47.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.47.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.47.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.47.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.47.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.47.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.47.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.48.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.48.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.48.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.48.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.48.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.48.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.48.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.48.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.48.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.48.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.48.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.49.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.49.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.49.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.49.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.49.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.49.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.49.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.49.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.49.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.49.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.49.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.50.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.50.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.50.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.50.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.50.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.50.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.50.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.50.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.50.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.50.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.50.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.51.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.51.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.51.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.51.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.51.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.51.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.51.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.51.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.51.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.51.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.51.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.52.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.52.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.52.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.52.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.52.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.52.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.52.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.52.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.52.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.52.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.52.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.53.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.53.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.53.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.53.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.53.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.53.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.53.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.53.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.53.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.53.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.53.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.54.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.54.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.54.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.54.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.54.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.54.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.54.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.54.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.54.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.54.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.54.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.55.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.55.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.55.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.55.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.55.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.55.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.55.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.55.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.55.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.55.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.55.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.56.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.56.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.56.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.56.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.56.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.56.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.56.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.56.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.56.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.56.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.56.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.57.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.57.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.57.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.57.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.57.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.57.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.57.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.57.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.57.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.57.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.57.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.58.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.58.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.58.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.58.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.58.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.58.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.58.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.58.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.58.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.58.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.58.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.59.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.59.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.59.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.59.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.59.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.59.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.59.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.59.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.59.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.59.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.59.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.60.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.60.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.60.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.60.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.60.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.60.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.60.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.60.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.60.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.60.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.60.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.61.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.61.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.61.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.61.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.61.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.61.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.61.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.61.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.61.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.61.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.61.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.62.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.62.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.62.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.62.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.62.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.62.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.62.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.62.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.62.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.62.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.62.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.63.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.63.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.63.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.63.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.63.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.63.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.63.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.63.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.63.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.63.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.63.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.64.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.64.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.64.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.64.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.64.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.64.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.64.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.64.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.64.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.64.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.64.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.65.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.65.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.65.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.65.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.65.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.65.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.65.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.65.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.65.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.65.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.65.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.66.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.66.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.66.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.66.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.66.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.66.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.66.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.66.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.66.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.66.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.66.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.67.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.67.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.67.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.67.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.67.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.67.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.67.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.67.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.67.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.67.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.67.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.68.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.68.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.68.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.68.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.68.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.68.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.68.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.68.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.68.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.68.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.68.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.69.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.69.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.69.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.69.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.69.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.69.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.69.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.69.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.69.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.69.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.69.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.70.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.70.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.70.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.70.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.70.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.70.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.70.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.70.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.70.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.70.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.70.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.71.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.71.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.71.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.71.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.71.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.71.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.71.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.71.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.71.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.71.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.71.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.72.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.72.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.72.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.72.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.72.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.72.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.72.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.72.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.72.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.72.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.72.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.73.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.73.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.73.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.73.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.73.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.73.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.73.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.73.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.73.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.73.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.73.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.74.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.74.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.74.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.74.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.74.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.74.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.74.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.74.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.74.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.74.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.74.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.75.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.75.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.75.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.75.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.75.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.75.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.75.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.75.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.75.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.75.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.75.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.76.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.76.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.76.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.76.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.76.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.76.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.76.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.76.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.76.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.76.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.76.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    "model.layers.77.self_attn.q_a_proj": {
+      "bits": 8
+    },
+    "model.layers.77.self_attn.q_b_proj": {
+      "bits": 8
+    },
+    "model.layers.77.self_attn.kv_a_proj_with_mqa": {
+      "bits": 8
+    },
+    "model.layers.77.self_attn.kv_b_proj": {
+      "bits": 8
+    },
+    "model.layers.77.self_attn.o_proj": {
+      "bits": 8
+    },
+    "model.layers.77.self_attn.indexer.wq_b": {
+      "bits": 8
+    },
+    "model.layers.77.self_attn.indexer.wk": {
+      "bits": 8
+    },
+    "model.layers.77.self_attn.indexer.weights_proj": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    "model.layers.77.mlp.shared_experts.gate_proj": {
+      "bits": 8
+    },
+    "model.layers.77.mlp.shared_experts.up_proj": {
+      "bits": 8
+    },
+    "model.layers.77.mlp.shared_experts.down_proj": {
+      "bits": 8
+    },
+    ".*layers\\.0\\.mlp.*": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    ".*layers\\.1\\.mlp.*": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    ".*layers\\.2\\.mlp.*": {
+      "bits": 16,
+      "data_type": "float"
+    },
+    ".*weights_proj.*": {
+      "bits": 16,
+      "data_type": "float"
+    }
+  }
+}
\ No newline at end of file
diff --git a/tokenizer.json b/tokenizer.json
new file mode 100644
index 0000000000000000000000000000000000000000..aba40197a4cdb5607f4ab7a05fb0a4ee8054fd6d
--- /dev/null
+++ b/tokenizer.json
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19e773648cb4e65de8660ea6365e10acca112d42a854923df93db4a6f333a82d
+size 20217442
diff --git a/tokenizer_config.json b/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..1723f7d90e3fb497303ec7b18f88cf5d05928f37
--- /dev/null
+++ b/tokenizer_config.json
@@ -0,0 +1,33 @@
+{
+  "backend": "tokenizers",
+  "clean_up_tokenization_spaces": false,
+  "do_lower_case": false,
+  "eos_token": "<|endoftext|>",
+  "extra_special_tokens": [
+    "<|endoftext|>",
+    "[MASK]",
+    "[gMASK]",
+    "[sMASK]",
+    "<sop>",
+    "<eop>",
+    "<|system|>",
+    "<|user|>",
+    "<|assistant|>",
+    "<|observation|>",
+    "<|begin_of_image|>",
+    "<|end_of_image|>",
+    "<|begin_of_video|>",
+    "<|end_of_video|>",
+    "<|begin_of_audio|>",
+    "<|end_of_audio|>",
+    "<|begin_of_transcription|>",
+    "<|end_of_transcription|>"
+  ],
+  "is_local": true,
+  "model_max_length": 202752,
+  "model_specific_special_tokens": {},
+  "pad_token": "<|endoftext|>",
+  "padding_side": "left",
+  "remove_space": false,
+  "tokenizer_class": "TokenizersBackend"
+}