rdtand's picture
v5: max-not-sum sibling aggregation, kernel shape mask, joint input_global — validator: ppl=4.16, mean_NLL=1.43, MTP P0=89.5%
09de726 verified
Raw
History Blame
60.5 kB
{
"architectures": [
"Qwen3_5ForConditionalGeneration"
],
"image_token_id": 248056,
"language_model_only": false,
"model_type": "qwen3_5",
"text_config": {
"attention_bias": false,
"attention_dropout": 0.0,
"attn_output_gate": true,
"bos_token_id": 248044,
"dtype": "bfloat16",
"eos_token_id": 248044,
"full_attention_interval": 4,
"head_dim": 256,
"hidden_act": "silu",
"hidden_size": 5120,
"initializer_range": 0.02,
"intermediate_size": 17408,
"layer_types": [
"linear_attention",
"linear_attention",
"linear_attention",
"full_attention",
"linear_attention",
"linear_attention",
"linear_attention",
"full_attention",
"linear_attention",
"linear_attention",
"linear_attention",
"full_attention",
"linear_attention",
"linear_attention",
"linear_attention",
"full_attention",
"linear_attention",
"linear_attention",
"linear_attention",
"full_attention",
"linear_attention",
"linear_attention",
"linear_attention",
"full_attention",
"linear_attention",
"linear_attention",
"linear_attention",
"full_attention",
"linear_attention",
"linear_attention",
"linear_attention",
"full_attention",
"linear_attention",
"linear_attention",
"linear_attention",
"full_attention",
"linear_attention",
"linear_attention",
"linear_attention",
"full_attention",
"linear_attention",
"linear_attention",
"linear_attention",
"full_attention",
"linear_attention",
"linear_attention",
"linear_attention",
"full_attention",
"linear_attention",
"linear_attention",
"linear_attention",
"full_attention",
"linear_attention",
"linear_attention",
"linear_attention",
"full_attention",
"linear_attention",
"linear_attention",
"linear_attention",
"full_attention",
"linear_attention",
"linear_attention",
"linear_attention",
"full_attention"
],
"linear_conv_kernel_dim": 4,
"linear_key_head_dim": 128,
"linear_num_key_heads": 16,
"linear_num_value_heads": 48,
"linear_value_head_dim": 128,
"mamba_ssm_dtype": "float32",
"max_position_embeddings": 262144,
"model_type": "qwen3_5_text",
"mtp_num_hidden_layers": 1,
"mtp_use_dedicated_embeddings": false,
"num_attention_heads": 24,
"num_hidden_layers": 64,
"num_key_value_heads": 4,
"output_gate_type": "swish",
"pad_token_id": null,
"partial_rotary_factor": 0.25,
"rms_norm_eps": 1e-06,
"rope_parameters": {
"mrope_interleaved": true,
"mrope_section": [
11,
11,
10
],
"partial_rotary_factor": 0.25,
"rope_theta": 10000000,
"rope_type": "default"
},
"tie_word_embeddings": false,
"use_cache": true,
"vocab_size": 248320
},
"tie_word_embeddings": false,
"transformers_version": "4.57.1",
"video_token_id": 248057,
"vision_config": {
"deepstack_visual_indexes": [],
"depth": 27,
"hidden_act": "gelu_pytorch_tanh",
"hidden_size": 1152,
"in_channels": 3,
"initializer_range": 0.02,
"intermediate_size": 4304,
"model_type": "qwen3_5",
"num_heads": 16,
"num_position_embeddings": 2304,
"out_hidden_size": 5120,
"patch_size": 16,
"spatial_merge_size": 2,
"temporal_patch_size": 2
},
"vision_end_token_id": 248054,
"vision_start_token_id": 248053,
"quantization_config": {
"quant_method": "compressed-tensors",
"format": "mixed-precision",
"config_groups": {
"group_0": {
"format": "mxfp8-quantized",
"weights": {
"num_bits": 8,
"type": "float",
"strategy": "group",
"group_size": 32,
"symmetric": true,
"dynamic": false,
"scale_dtype": "torch.uint8",
"zp_dtype": "torch.uint8",
"observer": "memoryless_minmax"
},
"input_activations": {
"num_bits": 8,
"type": "float",
"strategy": "group",
"group_size": 32,
"symmetric": true,
"dynamic": true,
"scale_dtype": "torch.uint8",
"zp_dtype": "torch.uint8"
},
"targets": [
"re:^language_model[.]model[.]layers[.]23[.]self_attn[.]k_proj$",
"re:^language_model[.]model[.]layers[.]23[.]self_attn[.]q_proj$",
"re:^language_model[.]model[.]layers[.]23[.]self_attn[.]qkv_proj$",
"re:^language_model[.]model[.]layers[.]23[.]self_attn[.]v_proj$",
"re:^language_model[.]model[.]layers[.]29[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]29[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]29[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]31[.]self_attn[.]o_proj$",
"re:^language_model[.]model[.]layers[.]39[.]self_attn[.]k_proj$",
"re:^language_model[.]model[.]layers[.]39[.]self_attn[.]q_proj$",
"re:^language_model[.]model[.]layers[.]39[.]self_attn[.]qkv_proj$",
"re:^language_model[.]model[.]layers[.]39[.]self_attn[.]v_proj$",
"re:^language_model[.]model[.]layers[.]43[.]self_attn[.]o_proj$",
"re:^language_model[.]model[.]layers[.]44[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]44[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]44[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]46[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]46[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]46[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]47[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]47[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]47[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]51[.]self_attn[.]k_proj$",
"re:^language_model[.]model[.]layers[.]51[.]self_attn[.]q_proj$",
"re:^language_model[.]model[.]layers[.]51[.]self_attn[.]qkv_proj$",
"re:^language_model[.]model[.]layers[.]51[.]self_attn[.]v_proj$",
"re:^language_model[.]model[.]layers[.]56[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]56[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]56[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]63[.]self_attn[.]k_proj$",
"re:^language_model[.]model[.]layers[.]63[.]self_attn[.]q_proj$",
"re:^language_model[.]model[.]layers[.]63[.]self_attn[.]qkv_proj$",
"re:^language_model[.]model[.]layers[.]63[.]self_attn[.]v_proj$",
"re:^visual[.]blocks[.]13[.]attn[.]qkv$",
"re:^visual[.]blocks[.]6[.]attn[.]qkv$",
"re:^visual[.]blocks[.]7[.]attn[.]proj$",
"re:^visual[.]blocks[.]8[.]attn[.]qkv$",
"re:^visual[.]blocks[.]9[.]attn[.]proj$",
"re:^visual[.]blocks[.]9[.]attn[.]qkv$"
]
},
"group_1": {
"format": "nvfp4-pack-quantized",
"weights": {
"num_bits": 4,
"type": "float",
"strategy": "tensor_group",
"group_size": 16,
"symmetric": true,
"dynamic": false,
"scale_dtype": "torch.float8_e4m3fn",
"zp_dtype": "torch.float8_e4m3fn",
"observer": "memoryless_minmax"
},
"input_activations": {
"num_bits": 4,
"type": "float",
"strategy": "tensor_group",
"group_size": 16,
"symmetric": true,
"dynamic": "local",
"observer": "static_minmax",
"scale_dtype": "torch.float8_e4m3fn",
"zp_dtype": "torch.float8_e4m3fn"
},
"targets": [
"re:^language_model[.]lm_head$",
"re:^language_model[.]model[.]layers[.]0[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]0[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]0[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]0[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]0[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]0[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]0[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]1[.]linear_attn[.]in_proj_a$",
"re:^language_model[.]model[.]layers[.]1[.]linear_attn[.]in_proj_b$",
"re:^language_model[.]model[.]layers[.]1[.]linear_attn[.]in_proj_ba$",
"re:^language_model[.]model[.]layers[.]1[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]1[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]1[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]1[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]1[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]1[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]1[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]1[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]10[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]10[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]10[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]10[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]10[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]10[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]10[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]10[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]11[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]11[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]11[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]11[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]11[.]self_attn[.]o_proj$",
"re:^language_model[.]model[.]layers[.]12[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]12[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]12[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]12[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]12[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]12[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]12[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]12[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]13[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]13[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]13[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]13[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]13[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]13[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]13[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]13[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]14[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]14[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]14[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]14[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]14[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]14[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]14[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]14[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]15[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]15[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]15[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]15[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]15[.]self_attn[.]o_proj$",
"re:^language_model[.]model[.]layers[.]16[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]16[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]16[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]16[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]16[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]16[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]16[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]16[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]17[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]17[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]17[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]17[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]17[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]17[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]17[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]17[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]18[.]linear_attn[.]in_proj_a$",
"re:^language_model[.]model[.]layers[.]18[.]linear_attn[.]in_proj_b$",
"re:^language_model[.]model[.]layers[.]18[.]linear_attn[.]in_proj_ba$",
"re:^language_model[.]model[.]layers[.]18[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]18[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]18[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]18[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]18[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]18[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]18[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]18[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]19[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]19[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]19[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]19[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]19[.]self_attn[.]o_proj$",
"re:^language_model[.]model[.]layers[.]2[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]2[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]2[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]2[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]2[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]2[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]2[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]2[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]20[.]linear_attn[.]in_proj_a$",
"re:^language_model[.]model[.]layers[.]20[.]linear_attn[.]in_proj_b$",
"re:^language_model[.]model[.]layers[.]20[.]linear_attn[.]in_proj_ba$",
"re:^language_model[.]model[.]layers[.]20[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]20[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]20[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]20[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]20[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]20[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]20[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]20[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]21[.]linear_attn[.]in_proj_a$",
"re:^language_model[.]model[.]layers[.]21[.]linear_attn[.]in_proj_b$",
"re:^language_model[.]model[.]layers[.]21[.]linear_attn[.]in_proj_ba$",
"re:^language_model[.]model[.]layers[.]21[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]21[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]21[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]21[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]21[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]21[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]21[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]21[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]22[.]linear_attn[.]in_proj_a$",
"re:^language_model[.]model[.]layers[.]22[.]linear_attn[.]in_proj_b$",
"re:^language_model[.]model[.]layers[.]22[.]linear_attn[.]in_proj_ba$",
"re:^language_model[.]model[.]layers[.]22[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]22[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]22[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]22[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]22[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]22[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]22[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]22[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]23[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]23[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]23[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]23[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]23[.]self_attn[.]o_proj$",
"re:^language_model[.]model[.]layers[.]24[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]24[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]24[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]24[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]24[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]24[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]24[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]24[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]25[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]25[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]25[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]25[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]25[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]25[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]25[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]25[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]26[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]26[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]26[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]26[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]26[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]26[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]26[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]26[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]27[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]27[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]27[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]27[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]27[.]self_attn[.]o_proj$",
"re:^language_model[.]model[.]layers[.]28[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]28[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]28[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]28[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]28[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]28[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]28[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]28[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]29[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]29[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]29[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]29[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]29[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]3[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]3[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]3[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]3[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]3[.]self_attn[.]o_proj$",
"re:^language_model[.]model[.]layers[.]30[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]30[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]30[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]30[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]30[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]31[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]31[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]31[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]31[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]32[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]32[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]32[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]32[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]32[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]33[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]33[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]33[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]33[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]33[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]34[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]34[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]34[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]34[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]34[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]35[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]35[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]35[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]35[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]35[.]self_attn[.]o_proj$",
"re:^language_model[.]model[.]layers[.]36[.]linear_attn[.]in_proj_a$",
"re:^language_model[.]model[.]layers[.]36[.]linear_attn[.]in_proj_b$",
"re:^language_model[.]model[.]layers[.]36[.]linear_attn[.]in_proj_ba$",
"re:^language_model[.]model[.]layers[.]36[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]36[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]36[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]36[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]36[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]36[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]36[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]36[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]37[.]linear_attn[.]in_proj_a$",
"re:^language_model[.]model[.]layers[.]37[.]linear_attn[.]in_proj_b$",
"re:^language_model[.]model[.]layers[.]37[.]linear_attn[.]in_proj_ba$",
"re:^language_model[.]model[.]layers[.]37[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]37[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]37[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]37[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]37[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]37[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]37[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]37[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]38[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]38[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]38[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]38[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]38[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]38[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]38[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]38[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]39[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]39[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]39[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]39[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]39[.]self_attn[.]o_proj$",
"re:^language_model[.]model[.]layers[.]4[.]linear_attn[.]in_proj_a$",
"re:^language_model[.]model[.]layers[.]4[.]linear_attn[.]in_proj_b$",
"re:^language_model[.]model[.]layers[.]4[.]linear_attn[.]in_proj_ba$",
"re:^language_model[.]model[.]layers[.]4[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]4[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]4[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]4[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]4[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]4[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]4[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]4[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]40[.]linear_attn[.]in_proj_a$",
"re:^language_model[.]model[.]layers[.]40[.]linear_attn[.]in_proj_b$",
"re:^language_model[.]model[.]layers[.]40[.]linear_attn[.]in_proj_ba$",
"re:^language_model[.]model[.]layers[.]40[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]40[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]40[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]40[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]40[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]40[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]40[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]40[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]41[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]41[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]41[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]41[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]41[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]41[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]41[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]41[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]42[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]42[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]42[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]42[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]42[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]43[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]43[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]43[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]43[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]44[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]44[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]44[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]44[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]44[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]45[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]45[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]45[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]45[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]45[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]46[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]46[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]47[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]47[.]self_attn[.]o_proj$",
"re:^language_model[.]model[.]layers[.]48[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]48[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]48[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]48[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]48[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]49[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]49[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]49[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]49[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]49[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]5[.]linear_attn[.]in_proj_a$",
"re:^language_model[.]model[.]layers[.]5[.]linear_attn[.]in_proj_b$",
"re:^language_model[.]model[.]layers[.]5[.]linear_attn[.]in_proj_ba$",
"re:^language_model[.]model[.]layers[.]5[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]5[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]5[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]5[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]5[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]5[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]5[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]5[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]50[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]50[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]50[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]50[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]50[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]51[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]51[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]51[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]51[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]51[.]self_attn[.]o_proj$",
"re:^language_model[.]model[.]layers[.]52[.]linear_attn[.]in_proj_a$",
"re:^language_model[.]model[.]layers[.]52[.]linear_attn[.]in_proj_b$",
"re:^language_model[.]model[.]layers[.]52[.]linear_attn[.]in_proj_ba$",
"re:^language_model[.]model[.]layers[.]52[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]52[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]52[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]52[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]52[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]52[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]52[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]52[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]53[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]53[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]53[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]53[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]53[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]53[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]53[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]53[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]54[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]54[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]54[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]54[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]54[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]54[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]54[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]54[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]55[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]55[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]55[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]55[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]55[.]self_attn[.]k_proj$",
"re:^language_model[.]model[.]layers[.]55[.]self_attn[.]o_proj$",
"re:^language_model[.]model[.]layers[.]55[.]self_attn[.]q_proj$",
"re:^language_model[.]model[.]layers[.]55[.]self_attn[.]qkv_proj$",
"re:^language_model[.]model[.]layers[.]55[.]self_attn[.]v_proj$",
"re:^language_model[.]model[.]layers[.]56[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]56[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]56[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]56[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]56[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]57[.]linear_attn[.]in_proj_a$",
"re:^language_model[.]model[.]layers[.]57[.]linear_attn[.]in_proj_b$",
"re:^language_model[.]model[.]layers[.]57[.]linear_attn[.]in_proj_ba$",
"re:^language_model[.]model[.]layers[.]57[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]57[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]57[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]57[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]57[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]57[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]57[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]57[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]58[.]linear_attn[.]in_proj_a$",
"re:^language_model[.]model[.]layers[.]58[.]linear_attn[.]in_proj_b$",
"re:^language_model[.]model[.]layers[.]58[.]linear_attn[.]in_proj_ba$",
"re:^language_model[.]model[.]layers[.]58[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]58[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]58[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]58[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]58[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]58[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]58[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]58[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]59[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]59[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]59[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]59[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]59[.]self_attn[.]o_proj$",
"re:^language_model[.]model[.]layers[.]6[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]6[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]6[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]6[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]6[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]6[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]6[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]6[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]60[.]linear_attn[.]in_proj_a$",
"re:^language_model[.]model[.]layers[.]60[.]linear_attn[.]in_proj_b$",
"re:^language_model[.]model[.]layers[.]60[.]linear_attn[.]in_proj_ba$",
"re:^language_model[.]model[.]layers[.]60[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]60[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]60[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]60[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]60[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]60[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]60[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]60[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]61[.]linear_attn[.]in_proj_a$",
"re:^language_model[.]model[.]layers[.]61[.]linear_attn[.]in_proj_b$",
"re:^language_model[.]model[.]layers[.]61[.]linear_attn[.]in_proj_ba$",
"re:^language_model[.]model[.]layers[.]61[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]61[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]61[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]61[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]61[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]61[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]61[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]61[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]62[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]62[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]62[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]62[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]62[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]62[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]62[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]62[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]63[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]63[.]self_attn[.]o_proj$",
"re:^language_model[.]model[.]layers[.]7[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]7[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]7[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]7[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]7[.]self_attn[.]o_proj$",
"re:^language_model[.]model[.]layers[.]8[.]linear_attn[.]in_proj_a$",
"re:^language_model[.]model[.]layers[.]8[.]linear_attn[.]in_proj_b$",
"re:^language_model[.]model[.]layers[.]8[.]linear_attn[.]in_proj_ba$",
"re:^language_model[.]model[.]layers[.]8[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]8[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]8[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]8[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]8[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]8[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]8[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]8[.]mlp[.]up_proj$",
"re:^language_model[.]model[.]layers[.]9[.]linear_attn[.]in_proj_a$",
"re:^language_model[.]model[.]layers[.]9[.]linear_attn[.]in_proj_b$",
"re:^language_model[.]model[.]layers[.]9[.]linear_attn[.]in_proj_ba$",
"re:^language_model[.]model[.]layers[.]9[.]linear_attn[.]in_proj_qkv$",
"re:^language_model[.]model[.]layers[.]9[.]linear_attn[.]in_proj_qkvz$",
"re:^language_model[.]model[.]layers[.]9[.]linear_attn[.]in_proj_z$",
"re:^language_model[.]model[.]layers[.]9[.]linear_attn[.]out_proj$",
"re:^language_model[.]model[.]layers[.]9[.]mlp[.]down_proj$",
"re:^language_model[.]model[.]layers[.]9[.]mlp[.]gate_proj$",
"re:^language_model[.]model[.]layers[.]9[.]mlp[.]gate_up_proj$",
"re:^language_model[.]model[.]layers[.]9[.]mlp[.]up_proj$",
"re:^mtp[.]layers[.]0[.]mlp[.]down_proj$",
"re:^mtp[.]layers[.]0[.]mlp[.]gate_proj$",
"re:^mtp[.]layers[.]0[.]mlp[.]gate_up_proj$",
"re:^mtp[.]layers[.]0[.]mlp[.]up_proj$",
"re:^mtp[.]layers[.]0[.]self_attn[.]k_proj$",
"re:^mtp[.]layers[.]0[.]self_attn[.]o_proj$",
"re:^mtp[.]layers[.]0[.]self_attn[.]q_proj$",
"re:^mtp[.]layers[.]0[.]self_attn[.]qkv_proj$",
"re:^mtp[.]layers[.]0[.]self_attn[.]v_proj$",
"re:^visual[.]blocks[.]10[.]attn[.]proj$",
"re:^visual[.]blocks[.]10[.]attn[.]qkv$",
"re:^visual[.]blocks[.]10[.]mlp[.]linear_fc1$",
"re:^visual[.]blocks[.]10[.]mlp[.]linear_fc2$",
"re:^visual[.]blocks[.]11[.]attn[.]proj$",
"re:^visual[.]blocks[.]11[.]attn[.]qkv$",
"re:^visual[.]blocks[.]11[.]mlp[.]linear_fc1$",
"re:^visual[.]blocks[.]11[.]mlp[.]linear_fc2$",
"re:^visual[.]blocks[.]12[.]attn[.]proj$",
"re:^visual[.]blocks[.]12[.]attn[.]qkv$",
"re:^visual[.]blocks[.]12[.]mlp[.]linear_fc1$",
"re:^visual[.]blocks[.]12[.]mlp[.]linear_fc2$",
"re:^visual[.]blocks[.]13[.]attn[.]proj$",
"re:^visual[.]blocks[.]13[.]mlp[.]linear_fc1$",
"re:^visual[.]blocks[.]13[.]mlp[.]linear_fc2$",
"re:^visual[.]blocks[.]14[.]attn[.]proj$",
"re:^visual[.]blocks[.]14[.]attn[.]qkv$",
"re:^visual[.]blocks[.]14[.]mlp[.]linear_fc1$",
"re:^visual[.]blocks[.]14[.]mlp[.]linear_fc2$",
"re:^visual[.]blocks[.]15[.]attn[.]proj$",
"re:^visual[.]blocks[.]15[.]attn[.]qkv$",
"re:^visual[.]blocks[.]15[.]mlp[.]linear_fc1$",
"re:^visual[.]blocks[.]15[.]mlp[.]linear_fc2$",
"re:^visual[.]blocks[.]16[.]attn[.]proj$",
"re:^visual[.]blocks[.]16[.]attn[.]qkv$",
"re:^visual[.]blocks[.]16[.]mlp[.]linear_fc1$",
"re:^visual[.]blocks[.]16[.]mlp[.]linear_fc2$",
"re:^visual[.]blocks[.]17[.]attn[.]proj$",
"re:^visual[.]blocks[.]17[.]attn[.]qkv$",
"re:^visual[.]blocks[.]17[.]mlp[.]linear_fc1$",
"re:^visual[.]blocks[.]17[.]mlp[.]linear_fc2$",
"re:^visual[.]blocks[.]18[.]attn[.]proj$",
"re:^visual[.]blocks[.]18[.]attn[.]qkv$",
"re:^visual[.]blocks[.]18[.]mlp[.]linear_fc1$",
"re:^visual[.]blocks[.]18[.]mlp[.]linear_fc2$",
"re:^visual[.]blocks[.]19[.]attn[.]proj$",
"re:^visual[.]blocks[.]19[.]attn[.]qkv$",
"re:^visual[.]blocks[.]19[.]mlp[.]linear_fc1$",
"re:^visual[.]blocks[.]19[.]mlp[.]linear_fc2$",
"re:^visual[.]blocks[.]20[.]attn[.]proj$",
"re:^visual[.]blocks[.]20[.]attn[.]qkv$",
"re:^visual[.]blocks[.]20[.]mlp[.]linear_fc1$",
"re:^visual[.]blocks[.]20[.]mlp[.]linear_fc2$",
"re:^visual[.]blocks[.]21[.]attn[.]proj$",
"re:^visual[.]blocks[.]21[.]attn[.]qkv$",
"re:^visual[.]blocks[.]21[.]mlp[.]linear_fc1$",
"re:^visual[.]blocks[.]22[.]attn[.]proj$",
"re:^visual[.]blocks[.]22[.]attn[.]qkv$",
"re:^visual[.]blocks[.]22[.]mlp[.]linear_fc1$",
"re:^visual[.]blocks[.]22[.]mlp[.]linear_fc2$",
"re:^visual[.]blocks[.]23[.]attn[.]proj$",
"re:^visual[.]blocks[.]23[.]attn[.]qkv$",
"re:^visual[.]blocks[.]23[.]mlp[.]linear_fc1$",
"re:^visual[.]blocks[.]24[.]attn[.]proj$",
"re:^visual[.]blocks[.]24[.]attn[.]qkv$",
"re:^visual[.]blocks[.]24[.]mlp[.]linear_fc1$",
"re:^visual[.]blocks[.]24[.]mlp[.]linear_fc2$",
"re:^visual[.]blocks[.]25[.]attn[.]proj$",
"re:^visual[.]blocks[.]25[.]attn[.]qkv$",
"re:^visual[.]blocks[.]25[.]mlp[.]linear_fc1$",
"re:^visual[.]blocks[.]26[.]attn[.]proj$",
"re:^visual[.]blocks[.]26[.]attn[.]qkv$",
"re:^visual[.]blocks[.]26[.]mlp[.]linear_fc1$",
"re:^visual[.]blocks[.]5[.]mlp[.]linear_fc1$",
"re:^visual[.]blocks[.]5[.]mlp[.]linear_fc2$",
"re:^visual[.]blocks[.]6[.]mlp[.]linear_fc1$",
"re:^visual[.]blocks[.]6[.]mlp[.]linear_fc2$",
"re:^visual[.]blocks[.]7[.]attn[.]qkv$",
"re:^visual[.]blocks[.]7[.]mlp[.]linear_fc1$",
"re:^visual[.]blocks[.]7[.]mlp[.]linear_fc2$",
"re:^visual[.]blocks[.]8[.]mlp[.]linear_fc1$",
"re:^visual[.]blocks[.]8[.]mlp[.]linear_fc2$",
"re:^visual[.]blocks[.]9[.]mlp[.]linear_fc1$",
"re:^visual[.]blocks[.]9[.]mlp[.]linear_fc2$",
"re:^visual[.]merger[.]linear_fc1$",
"re:^visual[.]merger[.]linear_fc2$"
]
}
},
"ignore": [
"language_model.lm_head",
"language_model.model.embed_tokens",
"language_model.model.layers.0.linear_attn.in_proj_a",
"language_model.model.layers.0.linear_attn.in_proj_b",
"language_model.model.layers.0.linear_attn.in_proj_ba",
"language_model.model.layers.0.linear_attn.out_proj",
"language_model.model.layers.10.linear_attn.in_proj_a",
"language_model.model.layers.10.linear_attn.in_proj_b",
"language_model.model.layers.10.linear_attn.in_proj_ba",
"language_model.model.layers.11.self_attn.k_proj",
"language_model.model.layers.11.self_attn.q_proj",
"language_model.model.layers.11.self_attn.qkv_proj",
"language_model.model.layers.11.self_attn.v_proj",
"language_model.model.layers.12.linear_attn.in_proj_a",
"language_model.model.layers.12.linear_attn.in_proj_b",
"language_model.model.layers.12.linear_attn.in_proj_ba",
"language_model.model.layers.13.linear_attn.in_proj_a",
"language_model.model.layers.13.linear_attn.in_proj_b",
"language_model.model.layers.13.linear_attn.in_proj_ba",
"language_model.model.layers.14.linear_attn.in_proj_a",
"language_model.model.layers.14.linear_attn.in_proj_b",
"language_model.model.layers.14.linear_attn.in_proj_ba",
"language_model.model.layers.15.self_attn.k_proj",
"language_model.model.layers.15.self_attn.q_proj",
"language_model.model.layers.15.self_attn.qkv_proj",
"language_model.model.layers.15.self_attn.v_proj",
"language_model.model.layers.16.linear_attn.in_proj_a",
"language_model.model.layers.16.linear_attn.in_proj_b",
"language_model.model.layers.16.linear_attn.in_proj_ba",
"language_model.model.layers.17.linear_attn.in_proj_a",
"language_model.model.layers.17.linear_attn.in_proj_b",
"language_model.model.layers.17.linear_attn.in_proj_ba",
"language_model.model.layers.19.self_attn.k_proj",
"language_model.model.layers.19.self_attn.q_proj",
"language_model.model.layers.19.self_attn.qkv_proj",
"language_model.model.layers.19.self_attn.v_proj",
"language_model.model.layers.2.linear_attn.in_proj_a",
"language_model.model.layers.2.linear_attn.in_proj_b",
"language_model.model.layers.2.linear_attn.in_proj_ba",
"language_model.model.layers.24.linear_attn.in_proj_a",
"language_model.model.layers.24.linear_attn.in_proj_b",
"language_model.model.layers.24.linear_attn.in_proj_ba",
"language_model.model.layers.25.linear_attn.in_proj_a",
"language_model.model.layers.25.linear_attn.in_proj_b",
"language_model.model.layers.25.linear_attn.in_proj_ba",
"language_model.model.layers.26.linear_attn.in_proj_a",
"language_model.model.layers.26.linear_attn.in_proj_b",
"language_model.model.layers.26.linear_attn.in_proj_ba",
"language_model.model.layers.27.self_attn.k_proj",
"language_model.model.layers.27.self_attn.q_proj",
"language_model.model.layers.27.self_attn.qkv_proj",
"language_model.model.layers.27.self_attn.v_proj",
"language_model.model.layers.28.linear_attn.in_proj_a",
"language_model.model.layers.28.linear_attn.in_proj_b",
"language_model.model.layers.28.linear_attn.in_proj_ba",
"language_model.model.layers.29.linear_attn.in_proj_a",
"language_model.model.layers.29.linear_attn.in_proj_b",
"language_model.model.layers.29.linear_attn.in_proj_ba",
"language_model.model.layers.3.self_attn.k_proj",
"language_model.model.layers.3.self_attn.q_proj",
"language_model.model.layers.3.self_attn.qkv_proj",
"language_model.model.layers.3.self_attn.v_proj",
"language_model.model.layers.30.linear_attn.in_proj_a",
"language_model.model.layers.30.linear_attn.in_proj_b",
"language_model.model.layers.30.linear_attn.in_proj_ba",
"language_model.model.layers.30.linear_attn.in_proj_qkv",
"language_model.model.layers.30.linear_attn.in_proj_qkvz",
"language_model.model.layers.30.linear_attn.in_proj_z",
"language_model.model.layers.31.self_attn.k_proj",
"language_model.model.layers.31.self_attn.q_proj",
"language_model.model.layers.31.self_attn.qkv_proj",
"language_model.model.layers.31.self_attn.v_proj",
"language_model.model.layers.32.linear_attn.in_proj_a",
"language_model.model.layers.32.linear_attn.in_proj_b",
"language_model.model.layers.32.linear_attn.in_proj_ba",
"language_model.model.layers.32.linear_attn.in_proj_qkv",
"language_model.model.layers.32.linear_attn.in_proj_qkvz",
"language_model.model.layers.32.linear_attn.in_proj_z",
"language_model.model.layers.33.linear_attn.in_proj_a",
"language_model.model.layers.33.linear_attn.in_proj_b",
"language_model.model.layers.33.linear_attn.in_proj_ba",
"language_model.model.layers.33.linear_attn.in_proj_qkv",
"language_model.model.layers.33.linear_attn.in_proj_qkvz",
"language_model.model.layers.33.linear_attn.in_proj_z",
"language_model.model.layers.34.linear_attn.in_proj_a",
"language_model.model.layers.34.linear_attn.in_proj_b",
"language_model.model.layers.34.linear_attn.in_proj_ba",
"language_model.model.layers.34.linear_attn.in_proj_qkv",
"language_model.model.layers.34.linear_attn.in_proj_qkvz",
"language_model.model.layers.34.linear_attn.in_proj_z",
"language_model.model.layers.35.self_attn.k_proj",
"language_model.model.layers.35.self_attn.q_proj",
"language_model.model.layers.35.self_attn.qkv_proj",
"language_model.model.layers.35.self_attn.v_proj",
"language_model.model.layers.38.linear_attn.in_proj_a",
"language_model.model.layers.38.linear_attn.in_proj_b",
"language_model.model.layers.38.linear_attn.in_proj_ba",
"language_model.model.layers.41.linear_attn.in_proj_a",
"language_model.model.layers.41.linear_attn.in_proj_b",
"language_model.model.layers.41.linear_attn.in_proj_ba",
"language_model.model.layers.42.linear_attn.in_proj_a",
"language_model.model.layers.42.linear_attn.in_proj_b",
"language_model.model.layers.42.linear_attn.in_proj_ba",
"language_model.model.layers.42.linear_attn.in_proj_qkv",
"language_model.model.layers.42.linear_attn.in_proj_qkvz",
"language_model.model.layers.42.linear_attn.in_proj_z",
"language_model.model.layers.43.self_attn.k_proj",
"language_model.model.layers.43.self_attn.q_proj",
"language_model.model.layers.43.self_attn.qkv_proj",
"language_model.model.layers.43.self_attn.v_proj",
"language_model.model.layers.44.linear_attn.in_proj_a",
"language_model.model.layers.44.linear_attn.in_proj_b",
"language_model.model.layers.44.linear_attn.in_proj_ba",
"language_model.model.layers.45.linear_attn.in_proj_a",
"language_model.model.layers.45.linear_attn.in_proj_b",
"language_model.model.layers.45.linear_attn.in_proj_ba",
"language_model.model.layers.45.linear_attn.in_proj_qkv",
"language_model.model.layers.45.linear_attn.in_proj_qkvz",
"language_model.model.layers.45.linear_attn.in_proj_z",
"language_model.model.layers.46.linear_attn.in_proj_a",
"language_model.model.layers.46.linear_attn.in_proj_b",
"language_model.model.layers.46.linear_attn.in_proj_ba",
"language_model.model.layers.46.linear_attn.in_proj_qkv",
"language_model.model.layers.46.linear_attn.in_proj_qkvz",
"language_model.model.layers.46.linear_attn.in_proj_z",
"language_model.model.layers.47.self_attn.k_proj",
"language_model.model.layers.47.self_attn.q_proj",
"language_model.model.layers.47.self_attn.qkv_proj",
"language_model.model.layers.47.self_attn.v_proj",
"language_model.model.layers.48.linear_attn.in_proj_a",
"language_model.model.layers.48.linear_attn.in_proj_b",
"language_model.model.layers.48.linear_attn.in_proj_ba",
"language_model.model.layers.48.linear_attn.in_proj_qkv",
"language_model.model.layers.48.linear_attn.in_proj_qkvz",
"language_model.model.layers.48.linear_attn.in_proj_z",
"language_model.model.layers.49.linear_attn.in_proj_a",
"language_model.model.layers.49.linear_attn.in_proj_b",
"language_model.model.layers.49.linear_attn.in_proj_ba",
"language_model.model.layers.49.linear_attn.in_proj_qkv",
"language_model.model.layers.49.linear_attn.in_proj_qkvz",
"language_model.model.layers.49.linear_attn.in_proj_z",
"language_model.model.layers.50.linear_attn.in_proj_a",
"language_model.model.layers.50.linear_attn.in_proj_b",
"language_model.model.layers.50.linear_attn.in_proj_ba",
"language_model.model.layers.50.linear_attn.in_proj_qkv",
"language_model.model.layers.50.linear_attn.in_proj_qkvz",
"language_model.model.layers.50.linear_attn.in_proj_z",
"language_model.model.layers.53.linear_attn.in_proj_a",
"language_model.model.layers.53.linear_attn.in_proj_b",
"language_model.model.layers.53.linear_attn.in_proj_ba",
"language_model.model.layers.54.linear_attn.in_proj_a",
"language_model.model.layers.54.linear_attn.in_proj_b",
"language_model.model.layers.54.linear_attn.in_proj_ba",
"language_model.model.layers.56.linear_attn.in_proj_a",
"language_model.model.layers.56.linear_attn.in_proj_b",
"language_model.model.layers.56.linear_attn.in_proj_ba",
"language_model.model.layers.59.self_attn.k_proj",
"language_model.model.layers.59.self_attn.q_proj",
"language_model.model.layers.59.self_attn.qkv_proj",
"language_model.model.layers.59.self_attn.v_proj",
"language_model.model.layers.6.linear_attn.in_proj_a",
"language_model.model.layers.6.linear_attn.in_proj_b",
"language_model.model.layers.6.linear_attn.in_proj_ba",
"language_model.model.layers.62.linear_attn.in_proj_a",
"language_model.model.layers.62.linear_attn.in_proj_b",
"language_model.model.layers.62.linear_attn.in_proj_ba",
"language_model.model.layers.63.mlp.gate_proj",
"language_model.model.layers.63.mlp.gate_up_proj",
"language_model.model.layers.63.mlp.up_proj",
"language_model.model.layers.7.self_attn.k_proj",
"language_model.model.layers.7.self_attn.q_proj",
"language_model.model.layers.7.self_attn.qkv_proj",
"language_model.model.layers.7.self_attn.v_proj",
"mtp.fc",
"visual.blocks.0.attn.proj",
"visual.blocks.0.attn.qkv",
"visual.blocks.0.mlp.linear_fc1",
"visual.blocks.0.mlp.linear_fc2",
"visual.blocks.1.attn.proj",
"visual.blocks.1.attn.qkv",
"visual.blocks.1.mlp.linear_fc1",
"visual.blocks.1.mlp.linear_fc2",
"visual.blocks.2.attn.proj",
"visual.blocks.2.attn.qkv",
"visual.blocks.2.mlp.linear_fc1",
"visual.blocks.2.mlp.linear_fc2",
"visual.blocks.21.mlp.linear_fc2",
"visual.blocks.23.mlp.linear_fc2",
"visual.blocks.25.mlp.linear_fc2",
"visual.blocks.26.mlp.linear_fc2",
"visual.blocks.3.attn.proj",
"visual.blocks.3.attn.qkv",
"visual.blocks.3.mlp.linear_fc1",
"visual.blocks.3.mlp.linear_fc2",
"visual.blocks.4.attn.proj",
"visual.blocks.4.attn.qkv",
"visual.blocks.4.mlp.linear_fc1",
"visual.blocks.4.mlp.linear_fc2",
"visual.blocks.5.attn.proj",
"visual.blocks.5.attn.qkv",
"visual.blocks.6.attn.proj",
"visual.blocks.8.attn.proj",
"visual.pos_embed"
],
"quantization_status": "compressed"
}
}