4b scale-up charter SFT checkpoint 48

.gitattributes

@@ -57,3 +57,4 @@ sft_4epoch/charter/checkpoint-4/tokenizer.json filter=lfs diff=lfs merge=lfs -te
 sft_4epoch/charter/checkpoint-12/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 sft_4epoch/charter/checkpoint-24/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 sft_4epoch/charter/checkpoint-36/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+sft_4epoch/charter/checkpoint-48/tokenizer.json filter=lfs diff=lfs merge=lfs -text

sft_4epoch/charter/checkpoint-48/chat_template.jinja

@@ -0,0 +1,47 @@
+{{ bos_token }}
+{%- if messages[0]['role'] == 'system' -%}
+    {%- if messages[0]['content'] is string -%}
+        {%- set first_user_prefix = messages[0]['content'] + '
+
+' -%}
+    {%- else -%}
+        {%- set first_user_prefix = messages[0]['content'][0]['text'] + '
+
+' -%}
+    {%- endif -%}
+    {%- set loop_messages = messages[1:] -%}
+{%- else -%}
+    {%- set first_user_prefix = "" -%}
+    {%- set loop_messages = messages -%}
+{%- endif -%}
+{%- for message in loop_messages -%}
+    {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}
+        {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }}
+    {%- endif -%}
+    {%- if (message['role'] == 'assistant') -%}
+        {%- set role = "model" -%}
+    {%- else -%}
+        {%- set role = message['role'] -%}
+    {%- endif -%}
+    {{ '<start_of_turn>' + role + '
+' + (first_user_prefix if loop.first else "") }}
+    {%- if message['content'] is string -%}
+        {{ message['content'] | trim }}
+    {%- elif message['content'] is iterable -%}
+        {%- for item in message['content'] -%}
+            {%- if item['type'] == 'image' -%}
+                {{ '<start_of_image>' }}
+            {%- elif item['type'] == 'text' -%}
+                {{ item['text'] | trim }}
+            {%- endif -%}
+        {%- endfor -%}
+    {%- else -%}
+        {{ raise_exception("Invalid content type") }}
+    {%- endif -%}
+    {{ '<end_of_turn>
+' }}
+{%- endfor -%}
+{%- if add_generation_prompt -%}
+    {{'<start_of_turn>model
+'}}
+{%- endif -%}

sft_4epoch/charter/checkpoint-48/config.json

@@ -0,0 +1,111 @@
+{
+  "architectures": [
+    "Gemma3ForConditionalGeneration"
+  ],
+  "boi_token_index": 255999,
+  "bos_token_id": 2,
+  "dtype": "bfloat16",
+  "eoi_token_index": 256000,
+  "eos_token_id": 1,
+  "image_token_index": 262144,
+  "initializer_range": 0.02,
+  "mm_tokens_per_image": 256,
+  "model_type": "gemma3",
+  "pad_token_id": 0,
+  "text_config": {
+    "_sliding_window_pattern": 6,
+    "attention_bias": false,
+    "attention_dropout": 0.0,
+    "attn_logit_softcapping": null,
+    "bos_token_id": 2,
+    "cache_implementation": "hybrid",
+    "dtype": "bfloat16",
+    "eos_token_id": 1,
+    "final_logit_softcapping": null,
+    "head_dim": 256,
+    "hidden_activation": "gelu_pytorch_tanh",
+    "hidden_size": 2560,
+    "initializer_range": 0.02,
+    "intermediate_size": 10240,
+    "layer_types": [
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention"
+    ],
+    "max_position_embeddings": 131072,
+    "model_type": "gemma3_text",
+    "num_attention_heads": 8,
+    "num_hidden_layers": 34,
+    "num_key_value_heads": 4,
+    "pad_token_id": 0,
+    "query_pre_attn_scalar": 256,
+    "rms_norm_eps": 1e-06,
+    "rope_parameters": {
+      "full_attention": {
+        "factor": 8.0,
+        "rope_theta": 1000000.0,
+        "rope_type": "linear"
+      },
+      "sliding_attention": {
+        "rope_theta": 10000.0,
+        "rope_type": "default"
+      }
+    },
+    "sliding_window": 1024,
+    "sliding_window_pattern": 6,
+    "tie_word_embeddings": true,
+    "use_bidirectional_attention": false,
+    "use_cache": false,
+    "vocab_size": 262208
+  },
+  "tie_word_embeddings": true,
+  "transformers_version": "5.9.0",
+  "unsloth_fixed": true,
+  "use_cache": false,
+  "vision_config": {
+    "attention_dropout": 0.0,
+    "dtype": "bfloat16",
+    "hidden_act": "gelu_pytorch_tanh",
+    "hidden_size": 1152,
+    "image_size": 896,
+    "intermediate_size": 4304,
+    "layer_norm_eps": 1e-06,
+    "model_type": "siglip_vision_model",
+    "num_attention_heads": 16,
+    "num_channels": 3,
+    "num_hidden_layers": 27,
+    "patch_size": 14,
+    "vision_use_head": false
+  }
+}

sft_4epoch/charter/checkpoint-48/generation_config.json

@@ -0,0 +1,13 @@
+{
+  "bos_token_id": 2,
+  "cache_implementation": "hybrid",
+  "do_sample": true,
+  "eos_token_id": [
+    1,
+    106
+  ],
+  "pad_token_id": 0,
+  "top_k": 64,
+  "top_p": 0.95,
+  "transformers_version": "5.9.0"
+}

sft_4epoch/charter/checkpoint-48/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:544b6ec660a37442b519e667b6758ff876fa40f265e1c3069325d106ad9bd679
+size 9942783064

sft_4epoch/charter/checkpoint-48/optimizer.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f4e5e6de705f1f5db543d57b295297c91076ba869a51edcfaaaeed055d811452
+size 15521485995

sft_4epoch/charter/checkpoint-48/preprocessor_config.json

@@ -0,0 +1,29 @@
+{
+  "do_convert_rgb": null,
+  "do_normalize": true,
+  "do_pan_and_scan": null,
+  "do_rescale": true,
+  "do_resize": true,
+  "image_mean": [
+    0.5,
+    0.5,
+    0.5
+  ],
+  "image_processor_type": "Gemma3ImageProcessor",
+  "image_seq_length": 256,
+  "image_std": [
+    0.5,
+    0.5,
+    0.5
+  ],
+  "pan_and_scan_max_num_crops": null,
+  "pan_and_scan_min_crop_size": null,
+  "pan_and_scan_min_ratio_to_activate": null,
+  "processor_class": "Gemma3Processor",
+  "resample": 2,
+  "rescale_factor": 0.00392156862745098,
+  "size": {
+    "height": 896,
+    "width": 896
+  }
+}

sft_4epoch/charter/checkpoint-48/processor_config.json

@@ -0,0 +1,4 @@
+{
+  "image_seq_length": 256,
+  "processor_class": "Gemma3Processor"
+}

sft_4epoch/charter/checkpoint-48/pytorch_model_fsdp.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:18136af7e1ed98523a8775b99302c5f22f952032e5936ab5db985506a5ab9ac1
+size 9942970813

sft_4epoch/charter/checkpoint-48/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d6ae654ed628e46a69a3185a2008a760ecb6fa76a6ed8ea36091fbf51572a9dc
+size 14917

sft_4epoch/charter/checkpoint-48/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aca90864f21237f3b2ef7719c284655f031e849a610ff91881a6d9e258037405
+size 14917

sft_4epoch/charter/checkpoint-48/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ec3414e5c8b4d3b47de7f61d5314acd64635f9517a355779879e73c2585549d1
+size 1465

sft_4epoch/charter/checkpoint-48/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726
+size 33384567

sft_4epoch/charter/checkpoint-48/tokenizer_config.json

@@ -0,0 +1,26 @@
+{
+  "backend": "tokenizers",
+  "boi_token": "<start_of_image>",
+  "bos_token": "<bos>",
+  "clean_up_tokenization_spaces": false,
+  "eoi_token": "<end_of_image>",
+  "eos_token": "<eos>",
+  "image_token": "<image_soft_token>",
+  "is_local": true,
+  "local_files_only": false,
+  "mask_token": "<mask>",
+  "model_max_length": 131072,
+  "model_specific_special_tokens": {
+    "boi_token": "<start_of_image>",
+    "eoi_token": "<end_of_image>",
+    "image_token": "<image_soft_token>"
+  },
+  "pad_token": "<pad>",
+  "padding_side": "left",
+  "processor_class": "Gemma3Processor",
+  "sp_model_kwargs": null,
+  "spaces_between_special_tokens": false,
+  "tokenizer_class": "GemmaTokenizer",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false
+}

sft_4epoch/charter/checkpoint-48/tokens_state.json

@@ -0,0 +1 @@
+{"total": 100638720, "trainable": 62647352}

sft_4epoch/charter/checkpoint-48/trainer_state.json

@@ -0,0 +1,706 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.08533333333333333,
+  "eval_steps": 500,
+  "global_step": 48,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0017777777777777779,
+      "grad_norm": 2.9375,
+      "learning_rate": 0.0,
+      "loss": 1.1085205078125,
+      "memory/device_reserved (GiB)": 40.7,
+      "memory/max_active (GiB)": 34.07,
+      "memory/max_allocated (GiB)": 34.07,
+      "ppl": 3.02987,
+      "step": 1,
+      "tokens/total": 2097152,
+      "tokens/train_per_sec_per_gpu": 614.88,
+      "tokens/trainable": 1319329
+    },
+    {
+      "epoch": 0.0035555555555555557,
+      "grad_norm": 2.921875,
+      "learning_rate": 3.3333333333333333e-06,
+      "loss": 1.093994140625,
+      "memory/device_reserved (GiB)": 47.77,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.98618,
+      "step": 2,
+      "tokens/total": 4194304,
+      "tokens/train_per_sec_per_gpu": 617.46,
+      "tokens/trainable": 2628278
+    },
+    {
+      "epoch": 0.005333333333333333,
+      "grad_norm": 2.609375,
+      "learning_rate": 6.666666666666667e-06,
+      "loss": 1.0843505859375,
+      "memory/device_reserved (GiB)": 47.77,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.95752,
+      "step": 3,
+      "tokens/total": 6291456,
+      "tokens/train_per_sec_per_gpu": 655.41,
+      "tokens/trainable": 3948806
+    },
+    {
+      "epoch": 0.0071111111111111115,
+      "grad_norm": 1.7421875,
+      "learning_rate": 1e-05,
+      "loss": 1.0223388671875,
+      "memory/device_reserved (GiB)": 47.77,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.77969,
+      "step": 4,
+      "tokens/total": 8388608,
+      "tokens/train_per_sec_per_gpu": 598.2,
+      "tokens/trainable": 5229996
+    },
+    {
+      "epoch": 0.008888888888888889,
+      "grad_norm": 2.03125,
+      "learning_rate": 9.98903822616921e-06,
+      "loss": 1.0355224609375,
+      "memory/device_reserved (GiB)": 47.8,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.81658,
+      "step": 5,
+      "tokens/total": 10485760,
+      "tokens/train_per_sec_per_gpu": 684.37,
+      "tokens/trainable": 6541073
+    },
+    {
+      "epoch": 0.010666666666666666,
+      "grad_norm": 1.703125,
+      "learning_rate": 9.956206309337067e-06,
+      "loss": 1.012939453125,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.75368,
+      "step": 6,
+      "tokens/total": 12582912,
+      "tokens/train_per_sec_per_gpu": 634.23,
+      "tokens/trainable": 7870413
+    },
+    {
+      "epoch": 0.012444444444444444,
+      "grad_norm": 1.28125,
+      "learning_rate": 9.901664203302126e-06,
+      "loss": 0.9525146484375,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.59222,
+      "step": 7,
+      "tokens/total": 14680064,
+      "tokens/train_per_sec_per_gpu": 635.02,
+      "tokens/trainable": 9189777
+    },
+    {
+      "epoch": 0.014222222222222223,
+      "grad_norm": 0.91015625,
+      "learning_rate": 9.825677631722436e-06,
+      "loss": 0.9556884765625,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.60046,
+      "step": 8,
+      "tokens/total": 16777216,
+      "tokens/train_per_sec_per_gpu": 618.16,
+      "tokens/trainable": 10479490
+    },
+    {
+      "epoch": 0.016,
+      "grad_norm": 0.85546875,
+      "learning_rate": 9.728616793536588e-06,
+      "loss": 0.9462890625,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.57613,
+      "step": 9,
+      "tokens/total": 18874368,
+      "tokens/train_per_sec_per_gpu": 652.09,
+      "tokens/trainable": 11784999
+    },
+    {
+      "epoch": 0.017777777777777778,
+      "grad_norm": 0.67578125,
+      "learning_rate": 9.610954559391704e-06,
+      "loss": 0.958740234375,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.60841,
+      "step": 10,
+      "tokens/total": 20971520,
+      "tokens/train_per_sec_per_gpu": 600.22,
+      "tokens/trainable": 13094236
+    },
+    {
+      "epoch": 0.019555555555555555,
+      "grad_norm": 0.87109375,
+      "learning_rate": 9.473264167865172e-06,
+      "loss": 0.943359375,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.5686,
+      "step": 11,
+      "tokens/total": 23068672,
+      "tokens/train_per_sec_per_gpu": 610.29,
+      "tokens/trainable": 14404718
+    },
+    {
+      "epoch": 0.021333333333333333,
+      "grad_norm": 0.6640625,
+      "learning_rate": 9.316216432703918e-06,
+      "loss": 0.9000244140625,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.45966,
+      "step": 12,
+      "tokens/total": 25165824,
+      "tokens/train_per_sec_per_gpu": 667.87,
+      "tokens/trainable": 15735257
+    },
+    {
+      "epoch": 0.02311111111111111,
+      "grad_norm": 0.55859375,
+      "learning_rate": 9.140576474687263e-06,
+      "loss": 0.94091796875,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.56233,
+      "step": 13,
+      "tokens/total": 27262976,
+      "tokens/train_per_sec_per_gpu": 634.12,
+      "tokens/trainable": 17038230
+    },
+    {
+      "epoch": 0.024888888888888887,
+      "grad_norm": 0.76953125,
+      "learning_rate": 8.947199994035402e-06,
+      "loss": 0.921875,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.514,
+      "step": 14,
+      "tokens/total": 29360128,
+      "tokens/train_per_sec_per_gpu": 705.62,
+      "tokens/trainable": 18338536
+    },
+    {
+      "epoch": 0.02666666666666667,
+      "grad_norm": 0.9140625,
+      "learning_rate": 8.737029101523931e-06,
+      "loss": 0.9312744140625,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.53774,
+      "step": 15,
+      "tokens/total": 31457280,
+      "tokens/train_per_sec_per_gpu": 587.74,
+      "tokens/trainable": 19638784
+    },
+    {
+      "epoch": 0.028444444444444446,
+      "grad_norm": 0.494140625,
+      "learning_rate": 8.511087728614863e-06,
+      "loss": 0.9188232421875,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.50634,
+      "step": 16,
+      "tokens/total": 33554432,
+      "tokens/train_per_sec_per_gpu": 625.92,
+      "tokens/trainable": 20962100
+    },
+    {
+      "epoch": 0.030222222222222223,
+      "grad_norm": 0.42578125,
+      "learning_rate": 8.270476638965463e-06,
+      "loss": 0.9246826171875,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.52107,
+      "step": 17,
+      "tokens/total": 35651584,
+      "tokens/train_per_sec_per_gpu": 647.08,
+      "tokens/trainable": 22244548
+    },
+    {
+      "epoch": 0.032,
+      "grad_norm": 0.5390625,
+      "learning_rate": 8.016368065618361e-06,
+      "loss": 0.92431640625,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.52014,
+      "step": 18,
+      "tokens/total": 37748736,
+      "tokens/train_per_sec_per_gpu": 633.07,
+      "tokens/trainable": 23550208
+    },
+    {
+      "epoch": 0.033777777777777775,
+      "grad_norm": 0.439453125,
+      "learning_rate": 7.75e-06,
+      "loss": 0.9212646484375,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.51247,
+      "step": 19,
+      "tokens/total": 39845888,
+      "tokens/train_per_sec_per_gpu": 654.96,
+      "tokens/trainable": 24856052
+    },
+    {
+      "epoch": 0.035555555555555556,
+      "grad_norm": 0.419921875,
+      "learning_rate": 7.472670160550849e-06,
+      "loss": 0.9345703125,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.54612,
+      "step": 20,
+      "tokens/total": 41943040,
+      "tokens/train_per_sec_per_gpu": 613.91,
+      "tokens/trainable": 26140298
+    },
+    {
+      "epoch": 0.037333333333333336,
+      "grad_norm": 0.58984375,
+      "learning_rate": 7.185729670371605e-06,
+      "loss": 0.937255859375,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.55297,
+      "step": 21,
+      "tokens/total": 44040192,
+      "tokens/train_per_sec_per_gpu": 627.12,
+      "tokens/trainable": 27452072
+    },
+    {
+      "epoch": 0.03911111111111111,
+      "grad_norm": 0.435546875,
+      "learning_rate": 6.890576474687264e-06,
+      "loss": 0.938232421875,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.55546,
+      "step": 22,
+      "tokens/total": 46137344,
+      "tokens/train_per_sec_per_gpu": 640.7,
+      "tokens/trainable": 28743186
+    },
+    {
+      "epoch": 0.04088888888888889,
+      "grad_norm": 0.41796875,
+      "learning_rate": 6.588648530198505e-06,
+      "loss": 0.9014892578125,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.46327,
+      "step": 23,
+      "tokens/total": 48234496,
+      "tokens/train_per_sec_per_gpu": 595.72,
+      "tokens/trainable": 30066808
+    },
+    {
+      "epoch": 0.042666666666666665,
+      "grad_norm": 0.443359375,
+      "learning_rate": 6.281416799501188e-06,
+      "loss": 0.9259033203125,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.52415,
+      "step": 24,
+      "tokens/total": 50331648,
+      "tokens/train_per_sec_per_gpu": 594.85,
+      "tokens/trainable": 31368300
+    },
+    {
+      "epoch": 0.044444444444444446,
+      "grad_norm": 0.3984375,
+      "learning_rate": 5.970378084704441e-06,
+      "loss": 0.9315185546875,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.53836,
+      "step": 25,
+      "tokens/total": 52428800,
+      "tokens/train_per_sec_per_gpu": 622.74,
+      "tokens/trainable": 32679264
+    },
+    {
+      "epoch": 0.04622222222222222,
+      "grad_norm": 0.375,
+      "learning_rate": 5.657047735161256e-06,
+      "loss": 0.9356689453125,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.54892,
+      "step": 26,
+      "tokens/total": 54525952,
+      "tokens/train_per_sec_per_gpu": 601.47,
+      "tokens/trainable": 33988396
+    },
+    {
+      "epoch": 0.048,
+      "grad_norm": 0.41796875,
+      "learning_rate": 5.342952264838748e-06,
+      "loss": 0.920166015625,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.50971,
+      "step": 27,
+      "tokens/total": 56623104,
+      "tokens/train_per_sec_per_gpu": 634.93,
+      "tokens/trainable": 35298496
+    },
+    {
+      "epoch": 0.049777777777777775,
+      "grad_norm": 0.390625,
+      "learning_rate": 5.02962191529556e-06,
+      "loss": 0.920654296875,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.51093,
+      "step": 28,
+      "tokens/total": 58720256,
+      "tokens/train_per_sec_per_gpu": 635.98,
+      "tokens/trainable": 36597120
+    },
+    {
+      "epoch": 0.051555555555555556,
+      "grad_norm": 0.50390625,
+      "learning_rate": 4.718583200498814e-06,
+      "loss": 0.9144287109375,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.49535,
+      "step": 29,
+      "tokens/total": 60817408,
+      "tokens/train_per_sec_per_gpu": 661.5,
+      "tokens/trainable": 37907664
+    },
+    {
+      "epoch": 0.05333333333333334,
+      "grad_norm": 0.41015625,
+      "learning_rate": 4.4113514698014955e-06,
+      "loss": 0.906982421875,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.47684,
+      "step": 30,
+      "tokens/total": 62889984,
+      "tokens/train_per_sec_per_gpu": 612.59,
+      "tokens/trainable": 39186144
+    },
+    {
+      "epoch": 0.05511111111111111,
+      "grad_norm": 0.349609375,
+      "learning_rate": 4.109423525312738e-06,
+      "loss": 0.919921875,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.50909,
+      "step": 31,
+      "tokens/total": 64987136,
+      "tokens/train_per_sec_per_gpu": 646.69,
+      "tokens/trainable": 40499084
+    },
+    {
+      "epoch": 0.05688888888888889,
+      "grad_norm": 0.4375,
+      "learning_rate": 3.8142703296283954e-06,
+      "loss": 0.9412841796875,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.56327,
+      "step": 32,
+      "tokens/total": 67084288,
+      "tokens/train_per_sec_per_gpu": 578.71,
+      "tokens/trainable": 41754008
+    },
+    {
+      "epoch": 0.058666666666666666,
+      "grad_norm": 0.353515625,
+      "learning_rate": 3.527329839449152e-06,
+      "loss": 0.921142578125,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.51216,
+      "step": 33,
+      "tokens/total": 69181440,
+      "tokens/train_per_sec_per_gpu": 645.1,
+      "tokens/trainable": 43059700
+    },
+    {
+      "epoch": 0.060444444444444446,
+      "grad_norm": 0.35546875,
+      "learning_rate": 3.2500000000000015e-06,
+      "loss": 0.886962890625,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.42775,
+      "step": 34,
+      "tokens/total": 71278592,
+      "tokens/train_per_sec_per_gpu": 546.77,
+      "tokens/trainable": 44340384
+    },
+    {
+      "epoch": 0.06222222222222222,
+      "grad_norm": 0.341796875,
+      "learning_rate": 2.98363193438164e-06,
+      "loss": 0.9200439453125,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.5094,
+      "step": 35,
+      "tokens/total": 73375744,
+      "tokens/train_per_sec_per_gpu": 633.3,
+      "tokens/trainable": 45649076
+    },
+    {
+      "epoch": 0.064,
+      "grad_norm": 0.380859375,
+      "learning_rate": 2.7295233610345384e-06,
+      "loss": 0.929443359375,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.5331,
+      "step": 36,
+      "tokens/total": 75472896,
+      "tokens/train_per_sec_per_gpu": 600.88,
+      "tokens/trainable": 46969288
+    },
+    {
+      "epoch": 0.06577777777777778,
+      "grad_norm": 0.5859375,
+      "learning_rate": 2.4889122713851397e-06,
+      "loss": 0.9454345703125,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.57393,
+      "step": 37,
+      "tokens/total": 77570048,
+      "tokens/train_per_sec_per_gpu": 660.32,
+      "tokens/trainable": 48301204
+    },
+    {
+      "epoch": 0.06755555555555555,
+      "grad_norm": 0.34765625,
+      "learning_rate": 2.262970898476071e-06,
+      "loss": 0.905029296875,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.472,
+      "step": 38,
+      "tokens/total": 79667200,
+      "tokens/train_per_sec_per_gpu": 601.44,
+      "tokens/trainable": 49615048
+    },
+    {
+      "epoch": 0.06933333333333333,
+      "grad_norm": 0.328125,
+      "learning_rate": 2.0528000059646e-06,
+      "loss": 0.895751953125,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.44918,
+      "step": 39,
+      "tokens/total": 81764352,
+      "tokens/train_per_sec_per_gpu": 643.88,
+      "tokens/trainable": 50907048
+    },
+    {
+      "epoch": 0.07111111111111111,
+      "grad_norm": 0.349609375,
+      "learning_rate": 1.8594235253127373e-06,
+      "loss": 0.9324951171875,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.54084,
+      "step": 40,
+      "tokens/total": 83861504,
+      "tokens/train_per_sec_per_gpu": 604.16,
+      "tokens/trainable": 52206344
+    },
+    {
+      "epoch": 0.07288888888888889,
+      "grad_norm": 0.349609375,
+      "learning_rate": 1.6837835672960834e-06,
+      "loss": 0.929931640625,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.53434,
+      "step": 41,
+      "tokens/total": 85958656,
+      "tokens/train_per_sec_per_gpu": 606.12,
+      "tokens/trainable": 53511636
+    },
+    {
+      "epoch": 0.07466666666666667,
+      "grad_norm": 0.388671875,
+      "learning_rate": 1.526735832134829e-06,
+      "loss": 0.9239501953125,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.51922,
+      "step": 42,
+      "tokens/total": 88055808,
+      "tokens/train_per_sec_per_gpu": 641.12,
+      "tokens/trainable": 54832244
+    },
+    {
+      "epoch": 0.07644444444444444,
+      "grad_norm": 0.34375,
+      "learning_rate": 1.389045440608296e-06,
+      "loss": 0.907958984375,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.47926,
+      "step": 43,
+      "tokens/total": 90152960,
+      "tokens/train_per_sec_per_gpu": 599.45,
+      "tokens/trainable": 56163436
+    },
+    {
+      "epoch": 0.07822222222222222,
+      "grad_norm": 0.349609375,
+      "learning_rate": 1.2713832064634127e-06,
+      "loss": 0.9462890625,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.57613,
+      "step": 44,
+      "tokens/total": 92250112,
+      "tokens/train_per_sec_per_gpu": 641.09,
+      "tokens/trainable": 57456500
+    },
+    {
+      "epoch": 0.08,
+      "grad_norm": 0.3515625,
+      "learning_rate": 1.174322368277565e-06,
+      "loss": 0.9158935546875,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.49901,
+      "step": 45,
+      "tokens/total": 94347264,
+      "tokens/train_per_sec_per_gpu": 617.32,
+      "tokens/trainable": 58768064
+    },
+    {
+      "epoch": 0.08177777777777778,
+      "grad_norm": 1.2421875,
+      "learning_rate": 1.0983357966978747e-06,
+      "loss": 0.9150390625,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.49687,
+      "step": 46,
+      "tokens/total": 96444416,
+      "tokens/train_per_sec_per_gpu": 653.38,
+      "tokens/trainable": 60042444
+    },
+    {
+      "epoch": 0.08355555555555555,
+      "grad_norm": 0.474609375,
+      "learning_rate": 1.0437936906629336e-06,
+      "loss": 0.9189453125,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.50665,
+      "step": 47,
+      "tokens/total": 98541568,
+      "tokens/train_per_sec_per_gpu": 605.99,
+      "tokens/trainable": 61333816
+    },
+    {
+      "epoch": 0.08533333333333333,
+      "grad_norm": 0.515625,
+      "learning_rate": 1.0109617738307914e-06,
+      "loss": 0.919189453125,
+      "memory/device_reserved (GiB)": 47.82,
+      "memory/max_active (GiB)": 41.3,
+      "memory/max_allocated (GiB)": 41.3,
+      "ppl": 2.50726,
+      "step": 48,
+      "tokens/total": 100638720,
+      "tokens/train_per_sec_per_gpu": 617.43,
+      "tokens/trainable": 62647352
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 48,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.1883537665147535e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}

sft_4epoch/charter/checkpoint-48/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb6dc761c85b168655dd8ebc2a80f811c40b68515cf41dadbdf055ee9c1af7fc
+size 8913