ddz16 commited on
Commit
464739a
·
verified ·
1 Parent(s): 3650a98

Upload CamDistill-8B v5 (checkpoint-1326)

Browse files
README.md ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ base_model: Qwen/Qwen3-VL-8B-Instruct
4
+ tags:
5
+ - camera-movement
6
+ - video-understanding
7
+ - qwen3-vl
8
+ - sft
9
+ - distillation
10
+ ---
11
+
12
+ # CamDistill-8B
13
+
14
+ Camera-movement distillation SFT model, based on `Qwen/Qwen3-VL-8B-Instruct`.
15
+
16
+ - Checkpoint: v5-20260724-054032 / checkpoint-1326
17
+ - Training framework: [ms-swift](https://github.com/modelscope/ms-swift)
18
+ - CamDistill config: DEPTH=6, EXTRACT_LAYERS=1,3,5,9,11,13, LAMBDA=0.05
19
+
20
+ ## Usage
21
+
22
+ ```python
23
+ from transformers import AutoModelForCausalLM, AutoProcessor
24
+ model = AutoModelForCausalLM.from_pretrained("ddz16/CamDistill-8B", torch_dtype="bfloat16", device_map="auto")
25
+ processor = AutoProcessor.from_pretrained("ddz16/CamDistill-8B")
26
+ ```
args.json ADDED
@@ -0,0 +1,380 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "output_dir": "/efs/ddz/CamDistill/output/camera_sft_qwen3vl_8b_camdistill/v5-20260724-054032",
3
+ "per_device_train_batch_size": 2,
4
+ "num_train_epochs": 2.0,
5
+ "max_steps": -1,
6
+ "learning_rate": 1.5e-05,
7
+ "lr_scheduler_type": "cosine",
8
+ "lr_scheduler_kwargs": null,
9
+ "warmup_steps": 0,
10
+ "optim": "adamw_torch_fused",
11
+ "optim_args": null,
12
+ "weight_decay": 0.01,
13
+ "adam_beta1": 0.9,
14
+ "adam_beta2": 0.95,
15
+ "adam_epsilon": 1e-08,
16
+ "optim_target_modules": null,
17
+ "gradient_accumulation_steps": 4,
18
+ "average_tokens_across_devices": true,
19
+ "max_grad_norm": 1.0,
20
+ "label_smoothing_factor": 0.0,
21
+ "bf16": true,
22
+ "fp16": false,
23
+ "bf16_full_eval": false,
24
+ "fp16_full_eval": false,
25
+ "tf32": null,
26
+ "gradient_checkpointing": true,
27
+ "gradient_checkpointing_kwargs": null,
28
+ "torch_compile": false,
29
+ "torch_compile_backend": null,
30
+ "torch_compile_mode": null,
31
+ "use_liger_kernel": false,
32
+ "liger_kernel_config": null,
33
+ "use_cache": false,
34
+ "neftune_noise_alpha": null,
35
+ "torch_empty_cache_steps": null,
36
+ "auto_find_batch_size": false,
37
+ "logging_strategy": "steps",
38
+ "logging_steps": 10,
39
+ "logging_first_step": true,
40
+ "log_on_each_node": true,
41
+ "logging_nan_inf_filter": true,
42
+ "include_num_input_tokens_seen": false,
43
+ "log_level": "passive",
44
+ "log_level_replica": "warning",
45
+ "disable_tqdm": null,
46
+ "report_to": [
47
+ "wandb"
48
+ ],
49
+ "run_name": "/efs/ddz/CamDistill/output/camera_sft_qwen3vl_8b_camdistill/v5-20260724-054032",
50
+ "project": "huggingface",
51
+ "trackio_space_id": null,
52
+ "trackio_bucket_id": null,
53
+ "trackio_static_space_id": null,
54
+ "eval_strategy": "no",
55
+ "eval_steps": null,
56
+ "eval_delay": 0,
57
+ "per_device_eval_batch_size": 2,
58
+ "prediction_loss_only": false,
59
+ "eval_on_start": false,
60
+ "eval_do_concat_batches": true,
61
+ "eval_use_gather_object": false,
62
+ "eval_accumulation_steps": null,
63
+ "include_for_metrics": [],
64
+ "batch_eval_metrics": false,
65
+ "save_only_model": false,
66
+ "save_strategy": "epoch",
67
+ "save_steps": 500,
68
+ "save_on_each_node": false,
69
+ "save_total_limit": 1,
70
+ "enable_jit_checkpoint": false,
71
+ "push_to_hub": false,
72
+ "hub_token": null,
73
+ "hub_private_repo": null,
74
+ "hub_model_id": null,
75
+ "hub_strategy": "every_save",
76
+ "hub_always_push": false,
77
+ "hub_revision": null,
78
+ "load_best_model_at_end": false,
79
+ "metric_for_best_model": "loss",
80
+ "greater_is_better": false,
81
+ "ignore_data_skip": false,
82
+ "restore_callback_states_from_checkpoint": false,
83
+ "full_determinism": false,
84
+ "seed": 42,
85
+ "data_seed": 42,
86
+ "use_cpu": false,
87
+ "accelerator_config": {
88
+ "dispatch_batches": false
89
+ },
90
+ "parallelism_config": null,
91
+ "dataloader_drop_last": false,
92
+ "dataloader_num_workers": 8,
93
+ "dataloader_pin_memory": true,
94
+ "dataloader_persistent_workers": false,
95
+ "dataloader_prefetch_factor": null,
96
+ "remove_unused_columns": true,
97
+ "label_names": null,
98
+ "train_sampling_strategy": "random",
99
+ "length_column_name": "length",
100
+ "ddp_find_unused_parameters": null,
101
+ "ddp_bucket_cap_mb": null,
102
+ "ddp_broadcast_buffers": null,
103
+ "ddp_static_graph": null,
104
+ "ddp_backend": null,
105
+ "ddp_timeout": 18000000,
106
+ "fsdp": [],
107
+ "fsdp_config": null,
108
+ "deepspeed": {
109
+ "fp16": {
110
+ "enabled": "auto",
111
+ "loss_scale": 0,
112
+ "loss_scale_window": 1000,
113
+ "initial_scale_power": 16,
114
+ "hysteresis": 2,
115
+ "min_loss_scale": 1
116
+ },
117
+ "bf16": {
118
+ "enabled": "auto"
119
+ },
120
+ "zero_optimization": {
121
+ "stage": 2,
122
+ "offload_optimizer": {
123
+ "device": "none",
124
+ "pin_memory": true
125
+ },
126
+ "allgather_partitions": true,
127
+ "allgather_bucket_size": 200000000.0,
128
+ "overlap_comm": false,
129
+ "reduce_scatter": true,
130
+ "reduce_bucket_size": 200000000.0,
131
+ "contiguous_gradients": true
132
+ },
133
+ "gradient_accumulation_steps": "auto",
134
+ "gradient_clipping": "auto",
135
+ "steps_per_print": 2000,
136
+ "train_batch_size": "auto",
137
+ "train_micro_batch_size_per_gpu": "auto",
138
+ "wall_clock_breakdown": false
139
+ },
140
+ "debug": null,
141
+ "skip_memory_metrics": true,
142
+ "do_train": false,
143
+ "do_eval": false,
144
+ "do_predict": false,
145
+ "resume_from_checkpoint": null,
146
+ "warmup_ratio": 0.05,
147
+ "logging_dir": "/efs/ddz/CamDistill/output/camera_sft_qwen3vl_8b_camdistill/v5-20260724-054032/runs",
148
+ "local_rank": 0,
149
+ "sortish_sampler": false,
150
+ "predict_with_generate": false,
151
+ "generation_max_length": null,
152
+ "generation_num_beams": null,
153
+ "generation_config": null,
154
+ "tuner_backend": "peft",
155
+ "vit_gradient_checkpointing": false,
156
+ "router_aux_loss_coef": 0.0,
157
+ "enable_dft_loss": false,
158
+ "enable_channel_loss": false,
159
+ "safe_serialization": true,
160
+ "max_shard_size": "5GB",
161
+ "check_model": true,
162
+ "acc_strategy": "token",
163
+ "train_dataloader_shuffle": true,
164
+ "group_by_length": false,
165
+ "max_epochs": null,
166
+ "aligner_lr": null,
167
+ "vit_lr": null,
168
+ "use_logits_to_keep": null,
169
+ "ds3_gather_for_generation": true,
170
+ "resume_only_model": false,
171
+ "optimizer": null,
172
+ "loss_type": "camdistill",
173
+ "eval_metric": null,
174
+ "callbacks": [],
175
+ "early_stop_interval": null,
176
+ "eval_use_evalscope": false,
177
+ "eval_dataset": [],
178
+ "eval_dataset_args": null,
179
+ "eval_limit": null,
180
+ "eval_generation_config": null,
181
+ "extra_eval_args": null,
182
+ "tuner_type": "full",
183
+ "use_galore": false,
184
+ "galore_target_modules": null,
185
+ "galore_rank": 128,
186
+ "galore_update_proj_gap": 50,
187
+ "galore_scale": 1.0,
188
+ "galore_proj_type": "std",
189
+ "galore_optim_per_parameter": false,
190
+ "galore_with_embedding": false,
191
+ "galore_quantization": false,
192
+ "galore_proj_quant": false,
193
+ "galore_proj_bits": 4,
194
+ "galore_proj_group_size": 256,
195
+ "galore_cos_threshold": 0.4,
196
+ "galore_gamma_proj": 2,
197
+ "galore_queue_size": 5,
198
+ "lisa_activated_layers": 0,
199
+ "lisa_step_interval": 20,
200
+ "use_flash_ckpt": false,
201
+ "use_ray": false,
202
+ "ray_exp_name": null,
203
+ "device_groups": null,
204
+ "model": "Qwen/Qwen3-VL-8B-Instruct",
205
+ "model_type": "qwen3_vl_camdistill",
206
+ "model_revision": null,
207
+ "task_type": "causal_lm",
208
+ "torch_dtype": "bfloat16",
209
+ "attn_impl": "flash_attention_2",
210
+ "experts_impl": null,
211
+ "new_special_tokens": [],
212
+ "num_labels": null,
213
+ "problem_type": null,
214
+ "rope_scaling": null,
215
+ "device_map": null,
216
+ "max_memory": {},
217
+ "max_model_len": null,
218
+ "local_repo_path": null,
219
+ "init_strategy": null,
220
+ "template": "qwen3_vl",
221
+ "system": null,
222
+ "max_length": 16384,
223
+ "truncation_strategy": "delete",
224
+ "max_pixels": null,
225
+ "agent_template": null,
226
+ "norm_bbox": null,
227
+ "use_chat_template": true,
228
+ "padding_side": "right",
229
+ "padding_free": false,
230
+ "loss_scale": "default",
231
+ "sequence_parallel_size": 1,
232
+ "template_backend": "swift",
233
+ "response_prefix": null,
234
+ "enable_thinking": null,
235
+ "preserve_thinking": null,
236
+ "add_non_thinking_prefix": true,
237
+ "dataset": [
238
+ "/s3-ovbu-d1-us-west-2/dazhaodu/camera_movement_train_diverse_50k_en.jsonl"
239
+ ],
240
+ "val_dataset": [],
241
+ "cached_dataset": [],
242
+ "cached_val_dataset": [],
243
+ "split_dataset_ratio": 0.0,
244
+ "dataset_num_proc": 8,
245
+ "load_from_cache_file": true,
246
+ "dataset_shuffle": true,
247
+ "val_dataset_shuffle": false,
248
+ "streaming": false,
249
+ "interleave_prob": null,
250
+ "stopping_strategy": "first_exhausted",
251
+ "shuffle_buffer_size": 1000,
252
+ "download_mode": "reuse_dataset_if_exists",
253
+ "columns": {},
254
+ "strict": false,
255
+ "disable_auto_column_mapping": false,
256
+ "model_name": null,
257
+ "model_author": null,
258
+ "custom_dataset_info": [],
259
+ "quant_method": null,
260
+ "quant_bits": null,
261
+ "hqq_axis": null,
262
+ "bnb_4bit_compute_dtype": "bfloat16",
263
+ "bnb_4bit_quant_type": "nf4",
264
+ "bnb_4bit_use_double_quant": true,
265
+ "bnb_4bit_quant_storage": null,
266
+ "max_new_tokens": 64,
267
+ "temperature": 0.0,
268
+ "top_k": null,
269
+ "top_p": null,
270
+ "repetition_penalty": null,
271
+ "num_beams": 1,
272
+ "stream": false,
273
+ "stop_words": [],
274
+ "logprobs": false,
275
+ "top_logprobs": null,
276
+ "structured_outputs_regex": null,
277
+ "adapters": [],
278
+ "external_plugins": [
279
+ "/efs/ddz/CamDistill/camera_movement_sft/plugins/camdistill_plugin.py"
280
+ ],
281
+ "custom_register_path": [],
282
+ "model_kwargs": {},
283
+ "enable_npu_model_patch": true,
284
+ "load_args": false,
285
+ "load_data_args": false,
286
+ "packing": false,
287
+ "packing_length": null,
288
+ "packing_num_proc": 1,
289
+ "lazy_tokenize": true,
290
+ "use_hf": true,
291
+ "ignore_args_error": false,
292
+ "use_swift_lora": false,
293
+ "freeze_parameters": [],
294
+ "freeze_parameters_regex": null,
295
+ "freeze_parameters_ratio": 0.0,
296
+ "trainable_parameters": [],
297
+ "trainable_parameters_regex": null,
298
+ "freeze_llm": false,
299
+ "freeze_vit": true,
300
+ "freeze_aligner": true,
301
+ "target_modules": [
302
+ "all-linear"
303
+ ],
304
+ "target_regex": null,
305
+ "target_parameters": null,
306
+ "modules_to_save": [],
307
+ "lora_rank": 8,
308
+ "lora_alpha": 32,
309
+ "lora_dropout": 0.05,
310
+ "lora_bias": "none",
311
+ "lora_dtype": null,
312
+ "lorap_lr_ratio": null,
313
+ "use_rslora": false,
314
+ "use_dora": false,
315
+ "lora_ga_batch_size": 2,
316
+ "lora_ga_iters": 2,
317
+ "lora_ga_max_length": 1024,
318
+ "lora_ga_direction": "ArB2r",
319
+ "lora_ga_scale": "stable",
320
+ "lora_ga_stable_gamma": 16,
321
+ "init_weights": true,
322
+ "fourier_n_frequency": 2000,
323
+ "fourier_scaling": 300.0,
324
+ "boft_block_size": 4,
325
+ "boft_block_num": 0,
326
+ "boft_n_butterfly_factor": 1,
327
+ "boft_dropout": 0.0,
328
+ "vera_rank": 256,
329
+ "vera_projection_prng_key": 0,
330
+ "vera_dropout": 0.0,
331
+ "vera_d_initial": 0.1,
332
+ "adapter_act": "gelu",
333
+ "adapter_length": 128,
334
+ "adalora_target_r": 8,
335
+ "adalora_init_r": 12,
336
+ "adalora_tinit": 0,
337
+ "adalora_tfinal": 0,
338
+ "adalora_deltaT": 1,
339
+ "adalora_beta1": 0.85,
340
+ "adalora_beta2": 0.85,
341
+ "adalora_orth_reg_weight": 0.5,
342
+ "llamapro_num_new_blocks": 4,
343
+ "llamapro_num_groups": null,
344
+ "reft_layer_key": null,
345
+ "reft_layers": null,
346
+ "reft_rank": 4,
347
+ "reft_intervention_type": "LoreftIntervention",
348
+ "reft_args": null,
349
+ "swanlab_token": null,
350
+ "swanlab_project": "ms-swift",
351
+ "swanlab_workspace": null,
352
+ "swanlab_exp_name": null,
353
+ "swanlab_notification_method": null,
354
+ "swanlab_webhook_url": null,
355
+ "swanlab_secret": null,
356
+ "swanlab_sender_email": null,
357
+ "swanlab_receiver_email": null,
358
+ "swanlab_smtp_server": null,
359
+ "swanlab_smtp_port": null,
360
+ "swanlab_email_language": "zh",
361
+ "swanlab_mode": "cloud",
362
+ "add_version": true,
363
+ "create_checkpoint_symlink": false,
364
+ "zero_hpz_partition_size": null,
365
+ "deepspeed_autotp_size": null,
366
+ "swift_version": "4.3.0.dev0",
367
+ "ckpt_dir": null,
368
+ "rank": 0,
369
+ "global_world_size": 8,
370
+ "local_world_size": 8,
371
+ "model_suffix": "Qwen3-VL-8B-Instruct",
372
+ "model_info": "ModelInfo(model_type='qwen3_vl_camdistill', model_dir='/efs/ddz/hf/hub/models--Qwen--Qwen3-VL-8B-Instruct/snapshots/0c351dd01ed87e9c1b53cbc748cba10e6187ff3b', torch_dtype=torch.bfloat16, max_model_len=262144, quant_method=None, quant_bits=None, rope_scaling=None, is_moe_model=False, is_multimodal=True, config=None, task_type='causal_lm', num_labels=None)",
373
+ "model_meta": "ModelMeta(model_type='qwen3_vl_camdistill', model_groups=[ModelGroup(models=[Model(ms_model_id='Qwen/Qwen3-VL-4B-Instruct', hf_model_id='Qwen/Qwen3-VL-4B-Instruct', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-VL-8B-Instruct', hf_model_id='Qwen/Qwen3-VL-8B-Instruct', model_path=None, ms_revision=None, hf_revision=None)], template='qwen3_vl', ignore_patterns=None, requires=None, tags=[])], loader=<class 'camdistill_plugin.CamDistillQwen3VLLoader'>, template=None, model_arch=None, mcore_model_type=None, architectures=['Qwen3VLForConditionalGeneration'], additional_saved_files=[], torch_dtype=None, is_multimodal=True, is_reward=False, task_type=None, ignore_patterns=None, requires=[], tags=[])",
374
+ "model_dir": "/efs/ddz/hf/hub/models--Qwen--Qwen3-VL-8B-Instruct/snapshots/0c351dd01ed87e9c1b53cbc748cba10e6187ff3b",
375
+ "template_meta": "QwenTemplateMeta(template_type='qwen3_vl', prefix=[], prompt=['<|im_start|>user\\n{{QUERY}}<|im_end|>\\n<|im_start|>assistant\\n'], chat_sep=['<|im_end|>\\n'], suffix=['<|im_end|>\\n'], template_cls=<class 'swift.template.templates.qwen.Qwen3VLTemplate'>, system_prefix=['<|im_start|>system\\n{{SYSTEM}}<|im_end|>\\n'], default_system=None, auto_add_bos=False, stop_words=['<|endoftext|>'], agent_template='hermes', is_thinking=False, thinking_prefix='<think>\\n', non_thinking_prefix='', history_thinking_prefix='')",
376
+ "_val_dataset_exists": true,
377
+ "hub": "<class 'swift.hub.hub.HFHub'>",
378
+ "evaluation_strategy": "no",
379
+ "training_args": "Seq2SeqTrainingArguments(output_dir='/efs/ddz/CamDistill/output/camera_sft_qwen3vl_8b_camdistill/v5-20260724-054032', per_device_train_batch_size=2, num_train_epochs=2.0, max_steps=-1, learning_rate=1.5e-05, lr_scheduler_type=<SchedulerType.COSINE: 'cosine'>, lr_scheduler_kwargs=None, warmup_steps=0.05, optim=<OptimizerNames.ADAMW_TORCH_FUSED: 'adamw_torch_fused'>, optim_args=None, weight_decay=0.01, adam_beta1=0.9, adam_beta2=0.95, adam_epsilon=1e-08, optim_target_modules=None, gradient_accumulation_steps=4, average_tokens_across_devices=None, max_grad_norm=1.0, label_smoothing_factor=0.0, bf16=True, fp16=False, bf16_full_eval=False, fp16_full_eval=False, tf32=None, gradient_checkpointing=True, gradient_checkpointing_kwargs=None, torch_compile=False, torch_compile_backend=None, torch_compile_mode=None, use_liger_kernel=False, liger_kernel_config=None, use_cache=False, neftune_noise_alpha=None, torch_empty_cache_steps=None, auto_find_batch_size=False, logging_strategy=<IntervalStrategy.STEPS: 'steps'>, logging_steps=10, logging_first_step=True, log_on_each_node=True, logging_nan_inf_filter=True, include_num_input_tokens_seen=None, log_level='passive', log_level_replica='warning', disable_tqdm=False, report_to=['wandb'], run_name='/efs/ddz/CamDistill/output/camera_sft_qwen3vl_8b_camdistill/v5-20260724-054032', project='huggingface', trackio_space_id=None, trackio_bucket_id=None, trackio_static_space_id=None, eval_strategy=<IntervalStrategy.NO: 'no'>, eval_steps=None, eval_delay=0, per_device_eval_batch_size=2, prediction_loss_only=False, eval_on_start=False, eval_do_concat_batches=True, eval_use_gather_object=False, eval_accumulation_steps=None, include_for_metrics=[], batch_eval_metrics=False, save_only_model=False, save_strategy=<SaveStrategy.EPOCH: 'epoch'>, save_steps=500, save_on_each_node=False, save_total_limit=1, enable_jit_checkpoint=False, push_to_hub=False, hub_token=None, hub_private_repo=None, hub_model_id=None, hub_strategy=<HubStrategy.EVERY_SAVE: 'every_save'>, hub_always_push=False, hub_revision=None, load_best_model_at_end=False, metric_for_best_model='loss', greater_is_better=False, ignore_data_skip=False, restore_callback_states_from_checkpoint=False, full_determinism=False, seed=42, data_seed=42, use_cpu=False, accelerator_config=AcceleratorConfig(split_batches=False, dispatch_batches=False, even_batches=True, use_seedable_sampler=True, non_blocking=False, gradient_accumulation_kwargs=None, use_configured_state=False), parallelism_config=None, dataloader_drop_last=False, dataloader_num_workers=8, dataloader_pin_memory=True, dataloader_persistent_workers=False, dataloader_prefetch_factor=2, remove_unused_columns=False, label_names=None, train_sampling_strategy='random', length_column_name='length', ddp_find_unused_parameters=None, ddp_bucket_cap_mb=None, ddp_broadcast_buffers=None, ddp_static_graph=None, ddp_backend=None, ddp_timeout=18000000, fsdp=[], fsdp_config={'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}, deepspeed={'fp16': {'enabled': 'auto', 'loss_scale': 0, 'loss_scale_window': 1000, 'initial_scale_power': 16, 'hysteresis': 2, 'min_loss_scale': 1}, 'bf16': {'enabled': 'auto'}, 'zero_optimization': {'stage': 2, 'offload_optimizer': {'device': 'none', 'pin_memory': True}, 'allgather_partitions': True, 'allgather_bucket_size': 200000000.0, 'overlap_comm': False, 'reduce_scatter': True, 'reduce_bucket_size': 200000000.0, 'contiguous_gradients': True}, 'gradient_accumulation_steps': 'auto', 'gradient_clipping': 'auto', 'steps_per_print': 2000, 'train_batch_size': 'auto', 'train_micro_batch_size_per_gpu': 'auto', 'wall_clock_breakdown': False}, debug=[], skip_memory_metrics=True, do_train=False, do_eval=False, do_predict=False, resume_from_checkpoint=None, warmup_ratio=0.05, logging_dir='/efs/ddz/CamDistill/output/camera_sft_qwen3vl_8b_camdistill/v5-20260724-054032/runs', local_rank=0, sortish_sampler=False, predict_with_generate=False, generation_max_length=None, generation_num_beams=None, generation_config=None, tuner_backend='peft', vit_gradient_checkpointing=False, router_aux_loss_coef=0.0, enable_dft_loss=False, enable_channel_loss=False, safe_serialization=True, max_shard_size='5GB', check_model=True, acc_strategy='token', train_dataloader_shuffle=True, group_by_length=False, max_epochs=None, aligner_lr=None, vit_lr=None, use_logits_to_keep=None, ds3_gather_for_generation=True, resume_only_model=False, optimizer=None, loss_type='camdistill', eval_metric=None, callbacks=[], early_stop_interval=None, eval_use_evalscope=False, eval_dataset=[], eval_dataset_args=None, eval_limit=None, eval_generation_config=None, extra_eval_args=None, tuner_type='full', use_galore=False, galore_target_modules=None, galore_rank=128, galore_update_proj_gap=50, galore_scale=1.0, galore_proj_type='std', galore_optim_per_parameter=False, galore_with_embedding=False, galore_quantization=False, galore_proj_quant=False, galore_proj_bits=4, galore_proj_group_size=256, galore_cos_threshold=0.4, galore_gamma_proj=2, galore_queue_size=5, lisa_activated_layers=0, lisa_step_interval=20, use_flash_ckpt=False)"
380
+ }
chat_template.jinja ADDED
@@ -0,0 +1,120 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {%- if tools %}
2
+ {{- '<|im_start|>system\n' }}
3
+ {%- if messages[0].role == 'system' %}
4
+ {%- if messages[0].content is string %}
5
+ {{- messages[0].content }}
6
+ {%- else %}
7
+ {%- for content in messages[0].content %}
8
+ {%- if 'text' in content %}
9
+ {{- content.text }}
10
+ {%- endif %}
11
+ {%- endfor %}
12
+ {%- endif %}
13
+ {{- '\n\n' }}
14
+ {%- endif %}
15
+ {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
16
+ {%- for tool in tools %}
17
+ {{- "\n" }}
18
+ {{- tool | tojson }}
19
+ {%- endfor %}
20
+ {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
21
+ {%- else %}
22
+ {%- if messages[0].role == 'system' %}
23
+ {{- '<|im_start|>system\n' }}
24
+ {%- if messages[0].content is string %}
25
+ {{- messages[0].content }}
26
+ {%- else %}
27
+ {%- for content in messages[0].content %}
28
+ {%- if 'text' in content %}
29
+ {{- content.text }}
30
+ {%- endif %}
31
+ {%- endfor %}
32
+ {%- endif %}
33
+ {{- '<|im_end|>\n' }}
34
+ {%- endif %}
35
+ {%- endif %}
36
+ {%- set image_count = namespace(value=0) %}
37
+ {%- set video_count = namespace(value=0) %}
38
+ {%- for message in messages %}
39
+ {%- if message.role == "user" %}
40
+ {{- '<|im_start|>' + message.role + '\n' }}
41
+ {%- if message.content is string %}
42
+ {{- message.content }}
43
+ {%- else %}
44
+ {%- for content in message.content %}
45
+ {%- if content.type == 'image' or 'image' in content or 'image_url' in content %}
46
+ {%- set image_count.value = image_count.value + 1 %}
47
+ {%- if add_vision_id %}Picture {{ image_count.value }}: {% endif -%}
48
+ <|vision_start|><|image_pad|><|vision_end|>
49
+ {%- elif content.type == 'video' or 'video' in content %}
50
+ {%- set video_count.value = video_count.value + 1 %}
51
+ {%- if add_vision_id %}Video {{ video_count.value }}: {% endif -%}
52
+ <|vision_start|><|video_pad|><|vision_end|>
53
+ {%- elif 'text' in content %}
54
+ {{- content.text }}
55
+ {%- endif %}
56
+ {%- endfor %}
57
+ {%- endif %}
58
+ {{- '<|im_end|>\n' }}
59
+ {%- elif message.role == "assistant" %}
60
+ {{- '<|im_start|>' + message.role + '\n' }}
61
+ {%- if message.content is string %}
62
+ {{- message.content }}
63
+ {%- else %}
64
+ {%- for content_item in message.content %}
65
+ {%- if 'text' in content_item %}
66
+ {{- content_item.text }}
67
+ {%- endif %}
68
+ {%- endfor %}
69
+ {%- endif %}
70
+ {%- if message.tool_calls %}
71
+ {%- for tool_call in message.tool_calls %}
72
+ {%- if (loop.first and message.content) or (not loop.first) %}
73
+ {{- '\n' }}
74
+ {%- endif %}
75
+ {%- if tool_call.function %}
76
+ {%- set tool_call = tool_call.function %}
77
+ {%- endif %}
78
+ {{- '<tool_call>\n{"name": "' }}
79
+ {{- tool_call.name }}
80
+ {{- '", "arguments": ' }}
81
+ {%- if tool_call.arguments is string %}
82
+ {{- tool_call.arguments }}
83
+ {%- else %}
84
+ {{- tool_call.arguments | tojson }}
85
+ {%- endif %}
86
+ {{- '}\n</tool_call>' }}
87
+ {%- endfor %}
88
+ {%- endif %}
89
+ {{- '<|im_end|>\n' }}
90
+ {%- elif message.role == "tool" %}
91
+ {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
92
+ {{- '<|im_start|>user' }}
93
+ {%- endif %}
94
+ {{- '\n<tool_response>\n' }}
95
+ {%- if message.content is string %}
96
+ {{- message.content }}
97
+ {%- else %}
98
+ {%- for content in message.content %}
99
+ {%- if content.type == 'image' or 'image' in content or 'image_url' in content %}
100
+ {%- set image_count.value = image_count.value + 1 %}
101
+ {%- if add_vision_id %}Picture {{ image_count.value }}: {% endif -%}
102
+ <|vision_start|><|image_pad|><|vision_end|>
103
+ {%- elif content.type == 'video' or 'video' in content %}
104
+ {%- set video_count.value = video_count.value + 1 %}
105
+ {%- if add_vision_id %}Video {{ video_count.value }}: {% endif -%}
106
+ <|vision_start|><|video_pad|><|vision_end|>
107
+ {%- elif 'text' in content %}
108
+ {{- content.text }}
109
+ {%- endif %}
110
+ {%- endfor %}
111
+ {%- endif %}
112
+ {{- '\n</tool_response>' }}
113
+ {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
114
+ {{- '<|im_end|>\n' }}
115
+ {%- endif %}
116
+ {%- endif %}
117
+ {%- endfor %}
118
+ {%- if add_generation_prompt %}
119
+ {{- '<|im_start|>assistant\n' }}
120
+ {%- endif %}
config.json ADDED
@@ -0,0 +1,70 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Qwen3VLForConditionalGeneration"
4
+ ],
5
+ "bos_token_id": null,
6
+ "dtype": "bfloat16",
7
+ "eos_token_id": 151645,
8
+ "hidden_size": 4096,
9
+ "image_token_id": 151655,
10
+ "model_type": "qwen3_vl",
11
+ "pad_token_id": 151643,
12
+ "text_config": {
13
+ "attention_bias": false,
14
+ "attention_dropout": 0.0,
15
+ "bos_token_id": 151643,
16
+ "dtype": "bfloat16",
17
+ "eos_token_id": 151645,
18
+ "head_dim": 128,
19
+ "hidden_act": "silu",
20
+ "hidden_size": 4096,
21
+ "initializer_range": 0.02,
22
+ "intermediate_size": 12288,
23
+ "max_position_embeddings": 262144,
24
+ "model_type": "qwen3_vl_text",
25
+ "num_attention_heads": 32,
26
+ "num_hidden_layers": 36,
27
+ "num_key_value_heads": 8,
28
+ "pad_token_id": 151643,
29
+ "rms_norm_eps": 1e-06,
30
+ "rope_parameters": {
31
+ "mrope_interleaved": true,
32
+ "mrope_section": [
33
+ 24,
34
+ 20,
35
+ 20
36
+ ],
37
+ "rope_theta": 5000000,
38
+ "rope_type": "default"
39
+ },
40
+ "use_cache": false,
41
+ "vocab_size": 151936
42
+ },
43
+ "tie_word_embeddings": false,
44
+ "transformers_version": "5.8.1",
45
+ "use_cache": false,
46
+ "video_token_id": 151656,
47
+ "vision_config": {
48
+ "deepstack_visual_indexes": [
49
+ 8,
50
+ 16,
51
+ 24
52
+ ],
53
+ "depth": 27,
54
+ "dtype": "bfloat16",
55
+ "hidden_act": "gelu_pytorch_tanh",
56
+ "hidden_size": 1152,
57
+ "in_channels": 3,
58
+ "initializer_range": 0.02,
59
+ "intermediate_size": 4304,
60
+ "model_type": "qwen3_vl_vision",
61
+ "num_heads": 16,
62
+ "num_position_embeddings": 2304,
63
+ "out_hidden_size": 4096,
64
+ "patch_size": 16,
65
+ "spatial_merge_size": 2,
66
+ "temporal_patch_size": 2
67
+ },
68
+ "vision_end_token_id": 151653,
69
+ "vision_start_token_id": 151652
70
+ }
generation_config.json ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token_id": 151643,
3
+ "do_sample": true,
4
+ "eos_token_id": [
5
+ 151645,
6
+ 151643
7
+ ],
8
+ "pad_token_id": 151643,
9
+ "repetition_penalty": 1.0,
10
+ "temperature": 0.7,
11
+ "top_k": 20,
12
+ "top_p": 0.8,
13
+ "transformers_version": "5.8.1"
14
+ }
model-00001-of-00004.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5df449d69419494a8458d2787d73bfcb4f0ee95c5a9b4fdea409389f50363a81
3
+ size 4998798776
model-00002-of-00004.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b88bbcfd021b78adc481d7d69eb42b1ba680fe0d72fdd54407960bb55ddeb0cd
3
+ size 4915962432
model-00003-of-00004.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3f230b1d9239c8880f215278ea69d8d0f45c525d84eda4388a6e391110feb8ee
3
+ size 4974674312
model-00004-of-00004.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ab17b18ff3b9eafef69201b8c6d804db34775e155d943a98200beff201ec4c17
3
+ size 3510134112
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
preprocessor_config.json ADDED
@@ -0,0 +1,21 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "size": {
3
+ "longest_edge": 16777216,
4
+ "shortest_edge": 65536
5
+ },
6
+ "patch_size": 16,
7
+ "temporal_patch_size": 2,
8
+ "merge_size": 2,
9
+ "image_mean": [
10
+ 0.5,
11
+ 0.5,
12
+ 0.5
13
+ ],
14
+ "image_std": [
15
+ 0.5,
16
+ 0.5,
17
+ 0.5
18
+ ],
19
+ "processor_class": "Qwen3VLProcessor",
20
+ "image_processor_type": "Qwen2VLImageProcessorFast"
21
+ }
processor_config.json ADDED
@@ -0,0 +1,60 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "image_processor": {
3
+ "do_convert_rgb": true,
4
+ "do_normalize": true,
5
+ "do_rescale": true,
6
+ "do_resize": true,
7
+ "image_mean": [
8
+ 0.5,
9
+ 0.5,
10
+ 0.5
11
+ ],
12
+ "image_processor_type": "Qwen2VLImageProcessor",
13
+ "image_std": [
14
+ 0.5,
15
+ 0.5,
16
+ 0.5
17
+ ],
18
+ "merge_size": 2,
19
+ "patch_size": 16,
20
+ "resample": 3,
21
+ "rescale_factor": 0.00392156862745098,
22
+ "size": {
23
+ "longest_edge": 16777216,
24
+ "shortest_edge": 65536
25
+ },
26
+ "temporal_patch_size": 2
27
+ },
28
+ "processor_class": "Qwen3VLProcessor",
29
+ "video_processor": {
30
+ "do_convert_rgb": true,
31
+ "do_normalize": true,
32
+ "do_rescale": true,
33
+ "do_resize": true,
34
+ "do_sample_frames": true,
35
+ "fps": 2,
36
+ "image_mean": [
37
+ 0.5,
38
+ 0.5,
39
+ 0.5
40
+ ],
41
+ "image_std": [
42
+ 0.5,
43
+ 0.5,
44
+ 0.5
45
+ ],
46
+ "max_frames": 768,
47
+ "merge_size": 2,
48
+ "min_frames": 4,
49
+ "patch_size": 16,
50
+ "resample": 3,
51
+ "rescale_factor": 0.00392156862745098,
52
+ "return_metadata": false,
53
+ "size": {
54
+ "longest_edge": 25165824,
55
+ "shortest_edge": 4096
56
+ },
57
+ "temporal_patch_size": 2,
58
+ "video_processor_type": "Qwen3VLVideoProcessor"
59
+ }
60
+ }
tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:be75606093db2094d7cd20f3c2f385c212750648bd6ea4fb2bf507a6a4c55506
3
+ size 11422650
tokenizer_config.json ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_prefix_space": false,
3
+ "backend": "tokenizers",
4
+ "bos_token": null,
5
+ "clean_up_tokenization_spaces": false,
6
+ "eos_token": "<|im_end|>",
7
+ "errors": "replace",
8
+ "extra_special_tokens": [
9
+ "<|im_start|>",
10
+ "<|im_end|>",
11
+ "<|object_ref_start|>",
12
+ "<|object_ref_end|>",
13
+ "<|box_start|>",
14
+ "<|box_end|>",
15
+ "<|quad_start|>",
16
+ "<|quad_end|>",
17
+ "<|vision_start|>",
18
+ "<|vision_end|>",
19
+ "<|vision_pad|>",
20
+ "<|image_pad|>",
21
+ "<|video_pad|>"
22
+ ],
23
+ "is_local": true,
24
+ "local_files_only": false,
25
+ "model_max_length": 262144,
26
+ "pad_token": "<|endoftext|>",
27
+ "processor_class": "Qwen3VLProcessor",
28
+ "split_special_tokens": false,
29
+ "tokenizer_class": "Qwen2Tokenizer",
30
+ "unk_token": null
31
+ }
trainer_state.json ADDED
@@ -0,0 +1,1630 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 2.0,
6
+ "eval_steps": 500,
7
+ "global_step": 1326,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "cos_sim": -0.0038788579404354095,
14
+ "distill_lambda": 0.05000000447034836,
15
+ "distill_loss": 1.0038789510726929,
16
+ "epoch": 0.0015082956259426848,
17
+ "grad_norm": 58.99883270263672,
18
+ "learning_rate": 2.2388059701492537e-07,
19
+ "loss": 2.520024538040161,
20
+ "sft_loss": 0.02855682373046875,
21
+ "step": 1,
22
+ "token_acc": 0.8815537084398977
23
+ },
24
+ {
25
+ "cos_sim": 0.00022824517347746424,
26
+ "distill_lambda": 0.050000001986821495,
27
+ "distill_loss": 0.9997717539469401,
28
+ "epoch": 0.015082956259426848,
29
+ "grad_norm": 45.729373931884766,
30
+ "learning_rate": 2.2388059701492535e-06,
31
+ "loss": 2.368698332044813,
32
+ "sft_loss": 0.024033228556315105,
33
+ "step": 10,
34
+ "token_acc": 0.8868504628691127
35
+ },
36
+ {
37
+ "cos_sim": 0.09473565220832825,
38
+ "distill_lambda": 0.05,
39
+ "distill_loss": 0.9052643775939941,
40
+ "epoch": 0.030165912518853696,
41
+ "grad_norm": 25.559162139892578,
42
+ "learning_rate": 4.477611940298507e-06,
43
+ "loss": 1.6844642639160157,
44
+ "sft_loss": 0.007376289367675782,
45
+ "step": 20,
46
+ "token_acc": 0.9129723341716918
47
+ },
48
+ {
49
+ "cos_sim": 0.46269831657409666,
50
+ "distill_lambda": 0.05,
51
+ "distill_loss": 0.5373017311096191,
52
+ "epoch": 0.04524886877828054,
53
+ "grad_norm": 11.83865737915039,
54
+ "learning_rate": 6.716417910447762e-06,
55
+ "loss": 1.0358287811279296,
56
+ "sft_loss": 0.0055045604705810545,
57
+ "step": 30,
58
+ "token_acc": 0.9293570753230381
59
+ },
60
+ {
61
+ "cos_sim": 0.7524478912353516,
62
+ "distill_lambda": 0.05,
63
+ "distill_loss": 0.24755215644836426,
64
+ "epoch": 0.06033182503770739,
65
+ "grad_norm": 8.810158729553223,
66
+ "learning_rate": 8.955223880597014e-06,
67
+ "loss": 0.5503463745117188,
68
+ "sft_loss": 0.004820716381072998,
69
+ "step": 40,
70
+ "token_acc": 0.936442432636975
71
+ },
72
+ {
73
+ "cos_sim": 0.8826584815979004,
74
+ "distill_lambda": 0.05,
75
+ "distill_loss": 0.1173414945602417,
76
+ "epoch": 0.07541478129713423,
77
+ "grad_norm": 6.174355506896973,
78
+ "learning_rate": 1.119402985074627e-05,
79
+ "loss": 0.3322910308837891,
80
+ "sft_loss": 0.004517018795013428,
81
+ "step": 50,
82
+ "token_acc": 0.9420423475551599
83
+ },
84
+ {
85
+ "cos_sim": 0.9243446350097656,
86
+ "distill_lambda": 0.05,
87
+ "distill_loss": 0.0756553053855896,
88
+ "epoch": 0.09049773755656108,
89
+ "grad_norm": 4.474364757537842,
90
+ "learning_rate": 1.3432835820895523e-05,
91
+ "loss": 0.26489078998565674,
92
+ "sft_loss": 0.004495072364807129,
93
+ "step": 60,
94
+ "token_acc": 0.941977054343285
95
+ },
96
+ {
97
+ "cos_sim": 0.9318131446838379,
98
+ "distill_lambda": 0.05,
99
+ "distill_loss": 0.06818689107894897,
100
+ "epoch": 0.10558069381598793,
101
+ "grad_norm": 2.775825023651123,
102
+ "learning_rate": 1.4999789854529206e-05,
103
+ "loss": 0.2519868850708008,
104
+ "sft_loss": 0.004465246200561523,
105
+ "step": 70,
106
+ "token_acc": 0.9433893684688778
107
+ },
108
+ {
109
+ "cos_sim": 0.9455140113830567,
110
+ "distill_lambda": 0.05,
111
+ "distill_loss": 0.054486030340194704,
112
+ "epoch": 0.12066365007541478,
113
+ "grad_norm": 4.228744029998779,
114
+ "learning_rate": 1.4996054262643387e-05,
115
+ "loss": 0.22749404907226561,
116
+ "sft_loss": 0.004384887218475342,
117
+ "step": 80,
118
+ "token_acc": 0.9445034140777238
119
+ },
120
+ {
121
+ "cos_sim": 0.9479646682739258,
122
+ "distill_lambda": 0.05,
123
+ "distill_loss": 0.052035284042358396,
124
+ "epoch": 0.13574660633484162,
125
+ "grad_norm": 6.396712779998779,
126
+ "learning_rate": 1.498765144861652e-05,
127
+ "loss": 0.22304110527038573,
128
+ "sft_loss": 0.0043682694435119625,
129
+ "step": 90,
130
+ "token_acc": 0.9444644410450224
131
+ },
132
+ {
133
+ "cos_sim": 0.9392847061157227,
134
+ "distill_lambda": 0.05,
135
+ "distill_loss": 0.060715246200561526,
136
+ "epoch": 0.15082956259426847,
137
+ "grad_norm": 1.407213568687439,
138
+ "learning_rate": 1.4974586644241e-05,
139
+ "loss": 0.23490571975708008,
140
+ "sft_loss": 0.004305040836334229,
141
+ "step": 100,
142
+ "token_acc": 0.9454039405258917
143
+ },
144
+ {
145
+ "cos_sim": 0.9490450859069824,
146
+ "distill_lambda": 0.05,
147
+ "distill_loss": 0.05095493793487549,
148
+ "epoch": 0.16591251885369532,
149
+ "grad_norm": 2.289855480194092,
150
+ "learning_rate": 1.4956867983975587e-05,
151
+ "loss": 0.21652469635009766,
152
+ "sft_loss": 0.004218649864196777,
153
+ "step": 110,
154
+ "token_acc": 0.947564196739759
155
+ },
156
+ {
157
+ "cos_sim": 0.9502164840698242,
158
+ "distill_lambda": 0.05,
159
+ "distill_loss": 0.04978338181972504,
160
+ "epoch": 0.18099547511312217,
161
+ "grad_norm": 1.5713039636611938,
162
+ "learning_rate": 1.493450649988069e-05,
163
+ "loss": 0.2129663944244385,
164
+ "sft_loss": 0.004166030883789062,
165
+ "step": 120,
166
+ "token_acc": 0.948197482197355
167
+ },
168
+ {
169
+ "cos_sim": 0.9450735092163086,
170
+ "distill_lambda": 0.05,
171
+ "distill_loss": 0.054926556348800656,
172
+ "epoch": 0.19607843137254902,
173
+ "grad_norm": 1.064908504486084,
174
+ "learning_rate": 1.4907516114749562e-05,
175
+ "loss": 0.22033605575561524,
176
+ "sft_loss": 0.004139173030853272,
177
+ "step": 130,
178
+ "token_acc": 0.9483397101983743
179
+ },
180
+ {
181
+ "cos_sim": 0.9489644050598145,
182
+ "distill_lambda": 0.05,
183
+ "distill_loss": 0.05103568434715271,
184
+ "epoch": 0.21116138763197587,
185
+ "grad_norm": 1.0259686708450317,
186
+ "learning_rate": 1.487591363343962e-05,
187
+ "loss": 0.2132336139678955,
188
+ "sft_loss": 0.004111766815185547,
189
+ "step": 140,
190
+ "token_acc": 0.9494511370894669
191
+ },
192
+ {
193
+ "cos_sim": 0.9520336151123047,
194
+ "distill_lambda": 0.05,
195
+ "distill_loss": 0.04796643853187561,
196
+ "epoch": 0.22624434389140272,
197
+ "grad_norm": 1.421288013458252,
198
+ "learning_rate": 1.4839718732409347e-05,
199
+ "loss": 0.20410797595977784,
200
+ "sft_loss": 0.0039800524711608885,
201
+ "step": 150,
202
+ "token_acc": 0.9502197768419759
203
+ },
204
+ {
205
+ "cos_sim": 0.949534797668457,
206
+ "distill_lambda": 0.05,
207
+ "distill_loss": 0.050465166568756104,
208
+ "epoch": 0.24132730015082957,
209
+ "grad_norm": 1.156361699104309,
210
+ "learning_rate": 1.4798953947467258e-05,
211
+ "loss": 0.21003410816192628,
212
+ "sft_loss": 0.004040306806564331,
213
+ "step": 160,
214
+ "token_acc": 0.9498282518162751
215
+ },
216
+ {
217
+ "cos_sim": 0.948972511291504,
218
+ "distill_lambda": 0.05,
219
+ "distill_loss": 0.051027512550354,
220
+ "epoch": 0.2564102564102564,
221
+ "grad_norm": 1.0698307752609253,
222
+ "learning_rate": 1.4753644659740567e-05,
223
+ "loss": 0.20925650596618653,
224
+ "sft_loss": 0.003987890481948852,
225
+ "step": 170,
226
+ "token_acc": 0.9506242473208433
227
+ },
228
+ {
229
+ "cos_sim": 0.9544513702392579,
230
+ "distill_lambda": 0.05,
231
+ "distill_loss": 0.045548611879348756,
232
+ "epoch": 0.27149321266968324,
233
+ "grad_norm": 1.2081609964370728,
234
+ "learning_rate": 1.4703819079872297e-05,
235
+ "loss": 0.19707401990890502,
236
+ "sft_loss": 0.0038811326026916503,
237
+ "step": 180,
238
+ "token_acc": 0.951166674188744
239
+ },
240
+ {
241
+ "cos_sim": 0.9517936706542969,
242
+ "distill_lambda": 0.05,
243
+ "distill_loss": 0.04820637702941895,
244
+ "epoch": 0.2865761689291101,
245
+ "grad_norm": 1.3792004585266113,
246
+ "learning_rate": 1.4649508230456655e-05,
247
+ "loss": 0.1991652250289917,
248
+ "sft_loss": 0.0038135945796966553,
249
+ "step": 190,
250
+ "token_acc": 0.9515554316167814
251
+ },
252
+ {
253
+ "cos_sim": 0.9557013511657715,
254
+ "distill_lambda": 0.05,
255
+ "distill_loss": 0.044298702478408815,
256
+ "epoch": 0.30165912518853694,
257
+ "grad_norm": 1.2754738330841064,
258
+ "learning_rate": 1.459074592672363e-05,
259
+ "loss": 0.1925476908683777,
260
+ "sft_loss": 0.003802180290222168,
261
+ "step": 200,
262
+ "token_acc": 0.9524128472554461
263
+ },
264
+ {
265
+ "cos_sim": 0.9551361083984375,
266
+ "distill_lambda": 0.05,
267
+ "distill_loss": 0.04486394822597504,
268
+ "epoch": 0.3167420814479638,
269
+ "grad_norm": 1.6434708833694458,
270
+ "learning_rate": 1.4527568755484822e-05,
271
+ "loss": 0.18934518098831177,
272
+ "sft_loss": 0.003673839569091797,
273
+ "step": 210,
274
+ "token_acc": 0.9543265859473611
275
+ },
276
+ {
277
+ "cos_sim": 0.9531267166137696,
278
+ "distill_lambda": 0.05,
279
+ "distill_loss": 0.04687337875366211,
280
+ "epoch": 0.33182503770739064,
281
+ "grad_norm": 0.9669171571731567,
282
+ "learning_rate": 1.4460016052353615e-05,
283
+ "loss": 0.19388854503631592,
284
+ "sft_loss": 0.0037153482437133787,
285
+ "step": 220,
286
+ "token_acc": 0.9548988874546683
287
+ },
288
+ {
289
+ "cos_sim": 0.9562634468078614,
290
+ "distill_lambda": 0.05,
291
+ "distill_loss": 0.043736580014228824,
292
+ "epoch": 0.3469079939668175,
293
+ "grad_norm": 1.262874960899353,
294
+ "learning_rate": 1.43881298772539e-05,
295
+ "loss": 0.18832703828811645,
296
+ "sft_loss": 0.00369839072227478,
297
+ "step": 230,
298
+ "token_acc": 0.9528906529589746
299
+ },
300
+ {
301
+ "cos_sim": 0.9578750610351563,
302
+ "distill_lambda": 0.05,
303
+ "distill_loss": 0.04212484657764435,
304
+ "epoch": 0.36199095022624433,
305
+ "grad_norm": 0.8858125805854797,
306
+ "learning_rate": 1.4311954988232549e-05,
307
+ "loss": 0.18181778192520143,
308
+ "sft_loss": 0.003575563430786133,
309
+ "step": 240,
310
+ "token_acc": 0.9552487502975482
311
+ },
312
+ {
313
+ "cos_sim": 0.9587590217590332,
314
+ "distill_lambda": 0.05,
315
+ "distill_loss": 0.04124096930027008,
316
+ "epoch": 0.3770739064856712,
317
+ "grad_norm": 0.7233920693397522,
318
+ "learning_rate": 1.4231538813591998e-05,
319
+ "loss": 0.18377578258514404,
320
+ "sft_loss": 0.0036809444427490234,
321
+ "step": 250,
322
+ "token_acc": 0.954867890147078
323
+ },
324
+ {
325
+ "cos_sim": 0.957796859741211,
326
+ "distill_lambda": 0.05,
327
+ "distill_loss": 0.04220314919948578,
328
+ "epoch": 0.39215686274509803,
329
+ "grad_norm": 0.83754563331604,
330
+ "learning_rate": 1.4146931422360251e-05,
331
+ "loss": 0.18479570150375366,
332
+ "sft_loss": 0.003664708137512207,
333
+ "step": 260,
334
+ "token_acc": 0.954435550482827
335
+ },
336
+ {
337
+ "cos_sim": 0.9538604736328125,
338
+ "distill_lambda": 0.05,
339
+ "distill_loss": 0.0461395263671875,
340
+ "epoch": 0.4072398190045249,
341
+ "grad_norm": 0.7190311551094055,
342
+ "learning_rate": 1.4058185493116703e-05,
343
+ "loss": 0.19155261516571045,
344
+ "sft_loss": 0.0036790430545806884,
345
+ "step": 270,
346
+ "token_acc": 0.9548158618225874
347
+ },
348
+ {
349
+ "cos_sim": 0.9538644790649414,
350
+ "distill_lambda": 0.05,
351
+ "distill_loss": 0.04613546431064606,
352
+ "epoch": 0.42232277526395173,
353
+ "grad_norm": 0.8648778200149536,
354
+ "learning_rate": 1.396535628119321e-05,
355
+ "loss": 0.18670086860656737,
356
+ "sft_loss": 0.003527629375457764,
357
+ "step": 280,
358
+ "token_acc": 0.9557889084454521
359
+ },
360
+ {
361
+ "cos_sim": 0.9595279693603516,
362
+ "distill_lambda": 0.05,
363
+ "distill_loss": 0.04047197699546814,
364
+ "epoch": 0.4374057315233786,
365
+ "grad_norm": 0.8901738524436951,
366
+ "learning_rate": 1.3868501584270802e-05,
367
+ "loss": 0.1772393226623535,
368
+ "sft_loss": 0.003515130281448364,
369
+ "step": 290,
370
+ "token_acc": 0.9564196791374593
371
+ },
372
+ {
373
+ "cos_sim": 0.9556629180908203,
374
+ "distill_lambda": 0.05,
375
+ "distill_loss": 0.04433700740337372,
376
+ "epoch": 0.45248868778280543,
377
+ "grad_norm": 0.7970823645591736,
378
+ "learning_rate": 1.3767681706393457e-05,
379
+ "loss": 0.1882307529449463,
380
+ "sft_loss": 0.0036653608083724977,
381
+ "step": 300,
382
+ "token_acc": 0.9546085057403417
383
+ },
384
+ {
385
+ "cos_sim": 0.9546260833740234,
386
+ "distill_lambda": 0.05,
387
+ "distill_loss": 0.04537382423877716,
388
+ "epoch": 0.4675716440422323,
389
+ "grad_norm": 0.849625825881958,
390
+ "learning_rate": 1.3662959420421375e-05,
391
+ "loss": 0.1848926067352295,
392
+ "sft_loss": 0.0035092025995254517,
393
+ "step": 310,
394
+ "token_acc": 0.9562830728057904
395
+ },
396
+ {
397
+ "cos_sim": 0.9584080696105957,
398
+ "distill_lambda": 0.05,
399
+ "distill_loss": 0.0415919303894043,
400
+ "epoch": 0.48265460030165913,
401
+ "grad_norm": 1.1224209070205688,
402
+ "learning_rate": 1.35543999289471e-05,
403
+ "loss": 0.17938269376754762,
404
+ "sft_loss": 0.00352611243724823,
405
+ "step": 320,
406
+ "token_acc": 0.9556174728604272
407
+ },
408
+ {
409
+ "cos_sim": 0.9568581581115723,
410
+ "distill_lambda": 0.05,
411
+ "distill_loss": 0.04314180314540863,
412
+ "epoch": 0.497737556561086,
413
+ "grad_norm": 0.6940692067146301,
414
+ "learning_rate": 1.3442070823698822e-05,
415
+ "loss": 0.18326592445373535,
416
+ "sft_loss": 0.0035699695348739625,
417
+ "step": 330,
418
+ "token_acc": 0.9559384414466557
419
+ },
420
+ {
421
+ "cos_sim": 0.9564298629760742,
422
+ "distill_lambda": 0.05,
423
+ "distill_loss": 0.04357010424137116,
424
+ "epoch": 0.5128205128205128,
425
+ "grad_norm": 0.6807535290718079,
426
+ "learning_rate": 1.332604204345616e-05,
427
+ "loss": 0.18083639144897462,
428
+ "sft_loss": 0.0034726321697235107,
429
+ "step": 340,
430
+ "token_acc": 0.957401203499357
431
+ },
432
+ {
433
+ "cos_sim": 0.9551042556762696,
434
+ "distill_lambda": 0.05,
435
+ "distill_loss": 0.04489580094814301,
436
+ "epoch": 0.5279034690799397,
437
+ "grad_norm": 0.9952406883239746,
438
+ "learning_rate": 1.3206385830504614e-05,
439
+ "loss": 0.18671003580093384,
440
+ "sft_loss": 0.0035898983478546143,
441
+ "step": 350,
442
+ "token_acc": 0.9559120709602776
443
+ },
444
+ {
445
+ "cos_sim": 0.9583676338195801,
446
+ "distill_lambda": 0.05,
447
+ "distill_loss": 0.04163242280483246,
448
+ "epoch": 0.5429864253393665,
449
+ "grad_norm": 0.7764410972595215,
450
+ "learning_rate": 1.3083176685655786e-05,
451
+ "loss": 0.17547743320465087,
452
+ "sft_loss": 0.0034020483493804933,
453
+ "step": 360,
454
+ "token_acc": 0.9579539837683563
455
+ },
456
+ {
457
+ "cos_sim": 0.9563582420349122,
458
+ "distill_lambda": 0.05,
459
+ "distill_loss": 0.04364180862903595,
460
+ "epoch": 0.5580693815987934,
461
+ "grad_norm": 0.6657865643501282,
462
+ "learning_rate": 1.295649132186141e-05,
463
+ "loss": 0.18018112182617188,
464
+ "sft_loss": 0.0034485697746276856,
465
+ "step": 370,
466
+ "token_acc": 0.9584288103053199
467
+ },
468
+ {
469
+ "cos_sim": 0.9531916618347168,
470
+ "distill_lambda": 0.05,
471
+ "distill_loss": 0.04680837094783783,
472
+ "epoch": 0.5731523378582202,
473
+ "grad_norm": 0.5820720195770264,
474
+ "learning_rate": 1.2826408616450027e-05,
475
+ "loss": 0.1861934781074524,
476
+ "sft_loss": 0.003478127717971802,
477
+ "step": 380,
478
+ "token_acc": 0.9566955266955267
479
+ },
480
+ {
481
+ "cos_sim": 0.9560998916625977,
482
+ "distill_lambda": 0.05,
483
+ "distill_loss": 0.0439001590013504,
484
+ "epoch": 0.5882352941176471,
485
+ "grad_norm": 0.8649207949638367,
486
+ "learning_rate": 1.2693009562016084e-05,
487
+ "loss": 0.182230806350708,
488
+ "sft_loss": 0.0034997045993804933,
489
+ "step": 390,
490
+ "token_acc": 0.9569879802026868
491
+ },
492
+ {
493
+ "cos_sim": 0.9577288627624512,
494
+ "distill_lambda": 0.05,
495
+ "distill_loss": 0.042271199822425845,
496
+ "epoch": 0.6033182503770739,
497
+ "grad_norm": 1.019384503364563,
498
+ "learning_rate": 1.2556377215992019e-05,
499
+ "loss": 0.17852908372879028,
500
+ "sft_loss": 0.0034654736518859863,
501
+ "step": 400,
502
+ "token_acc": 0.9581241117953576
503
+ },
504
+ {
505
+ "cos_sim": 0.9568939208984375,
506
+ "distill_lambda": 0.05,
507
+ "distill_loss": 0.043106073141098024,
508
+ "epoch": 0.6184012066365008,
509
+ "grad_norm": 0.8741854429244995,
510
+ "learning_rate": 1.2416596648934716e-05,
511
+ "loss": 0.17971467971801758,
512
+ "sft_loss": 0.0034607797861099243,
513
+ "step": 410,
514
+ "token_acc": 0.957463580274986
515
+ },
516
+ {
517
+ "cos_sim": 0.9562771797180176,
518
+ "distill_lambda": 0.05,
519
+ "distill_loss": 0.043722710013389586,
520
+ "epoch": 0.6334841628959276,
521
+ "grad_norm": 0.9732675552368164,
522
+ "learning_rate": 1.2273754891558542e-05,
523
+ "loss": 0.17926173210144042,
524
+ "sft_loss": 0.0034157931804656982,
525
+ "step": 420,
526
+ "token_acc": 0.9579150638323959
527
+ },
528
+ {
529
+ "cos_sim": 0.9562744140625,
530
+ "distill_lambda": 0.05,
531
+ "distill_loss": 0.043725624680519104,
532
+ "epoch": 0.6485671191553545,
533
+ "grad_norm": 0.8504221439361572,
534
+ "learning_rate": 1.212794088054797e-05,
535
+ "loss": 0.17900450229644777,
536
+ "sft_loss": 0.003407609462738037,
537
+ "step": 430,
538
+ "token_acc": 0.9575951205344176
539
+ },
540
+ {
541
+ "cos_sim": 0.9573652267456054,
542
+ "distill_lambda": 0.05,
543
+ "distill_loss": 0.04263473153114319,
544
+ "epoch": 0.6636500754147813,
545
+ "grad_norm": 0.6911851167678833,
546
+ "learning_rate": 1.1979245403183451e-05,
547
+ "loss": 0.1755327343940735,
548
+ "sft_loss": 0.003353661298751831,
549
+ "step": 440,
550
+ "token_acc": 0.9577667524194536
551
+ },
552
+ {
553
+ "cos_sim": 0.9589430809020996,
554
+ "distill_lambda": 0.05,
555
+ "distill_loss": 0.041057008504867556,
556
+ "epoch": 0.6787330316742082,
557
+ "grad_norm": 0.8256673216819763,
558
+ "learning_rate": 1.182776104081511e-05,
559
+ "loss": 0.17140830755233766,
560
+ "sft_loss": 0.0033036589622497557,
561
+ "step": 450,
562
+ "token_acc": 0.9586852351003294
563
+ },
564
+ {
565
+ "cos_sim": 0.9559874534606934,
566
+ "distill_lambda": 0.05,
567
+ "distill_loss": 0.04401255249977112,
568
+ "epoch": 0.693815987933635,
569
+ "grad_norm": 0.9640001654624939,
570
+ "learning_rate": 1.167358211121938e-05,
571
+ "loss": 0.17596397399902344,
572
+ "sft_loss": 0.0032982468605041505,
573
+ "step": 460,
574
+ "token_acc": 0.9592674651007855
575
+ },
576
+ {
577
+ "cos_sim": 0.9567716598510743,
578
+ "distill_lambda": 0.05,
579
+ "distill_loss": 0.043228423595428465,
580
+ "epoch": 0.7088989441930619,
581
+ "grad_norm": 0.8248547315597534,
582
+ "learning_rate": 1.1516804609874507e-05,
583
+ "loss": 0.17867534160614013,
584
+ "sft_loss": 0.00342218279838562,
585
+ "step": 470,
586
+ "token_acc": 0.9576837082276677
587
+ },
588
+ {
589
+ "cos_sim": 0.9568825721740722,
590
+ "distill_lambda": 0.05,
591
+ "distill_loss": 0.04311748445034027,
592
+ "epoch": 0.7239819004524887,
593
+ "grad_norm": 0.9829240441322327,
594
+ "learning_rate": 1.1357526150191471e-05,
595
+ "loss": 0.174727201461792,
596
+ "sft_loss": 0.0033043503761291506,
597
+ "step": 480,
598
+ "token_acc": 0.9591350273419852
599
+ },
600
+ {
601
+ "cos_sim": 0.9550409317016602,
602
+ "distill_lambda": 0.05,
603
+ "distill_loss": 0.04495901167392731,
604
+ "epoch": 0.7390648567119156,
605
+ "grad_norm": 0.7202755212783813,
606
+ "learning_rate": 1.1195845902737558e-05,
607
+ "loss": 0.1809255599975586,
608
+ "sft_loss": 0.0034059733152389526,
609
+ "step": 490,
610
+ "token_acc": 0.9573326718243693
611
+ },
612
+ {
613
+ "cos_sim": 0.9572129249572754,
614
+ "distill_lambda": 0.05,
615
+ "distill_loss": 0.0427870899438858,
616
+ "epoch": 0.7541478129713424,
617
+ "grad_norm": 0.667961835861206,
618
+ "learning_rate": 1.1031864533490397e-05,
619
+ "loss": 0.17608168125152587,
620
+ "sft_loss": 0.0033631980419158935,
621
+ "step": 500,
622
+ "token_acc": 0.9583889765691392
623
+ },
624
+ {
625
+ "cos_sim": 0.9579586029052735,
626
+ "distill_lambda": 0.05,
627
+ "distill_loss": 0.04204140901565552,
628
+ "epoch": 0.7692307692307693,
629
+ "grad_norm": 0.7028427124023438,
630
+ "learning_rate": 1.0865684141160912e-05,
631
+ "loss": 0.17287062406539916,
632
+ "sft_loss": 0.00330013632774353,
633
+ "step": 510,
634
+ "token_acc": 0.958580794446587
635
+ },
636
+ {
637
+ "cos_sim": 0.9577391624450684,
638
+ "distill_lambda": 0.05,
639
+ "distill_loss": 0.04226081669330597,
640
+ "epoch": 0.7843137254901961,
641
+ "grad_norm": 0.8001043796539307,
642
+ "learning_rate": 1.0697408193624248e-05,
643
+ "loss": 0.17270078659057617,
644
+ "sft_loss": 0.003283858299255371,
645
+ "step": 520,
646
+ "token_acc": 0.9596215576955567
647
+ },
648
+ {
649
+ "cos_sim": 0.9563358306884766,
650
+ "distill_lambda": 0.05,
651
+ "distill_loss": 0.04366412460803985,
652
+ "epoch": 0.799396681749623,
653
+ "grad_norm": 0.6969711780548096,
654
+ "learning_rate": 1.0527141463498186e-05,
655
+ "loss": 0.1776696562767029,
656
+ "sft_loss": 0.00336897075176239,
657
+ "step": 530,
658
+ "token_acc": 0.9582172289085196
659
+ },
660
+ {
661
+ "cos_sim": 0.9588777542114257,
662
+ "distill_lambda": 0.05,
663
+ "distill_loss": 0.041122305393219,
664
+ "epoch": 0.8144796380090498,
665
+ "grad_norm": 0.6015124320983887,
666
+ "learning_rate": 1.0354989962909223e-05,
667
+ "loss": 0.1682189702987671,
668
+ "sft_loss": 0.003200727701187134,
669
+ "step": 540,
670
+ "token_acc": 0.9603295649302662
671
+ },
672
+ {
673
+ "cos_sim": 0.9579531669616699,
674
+ "distill_lambda": 0.05,
675
+ "distill_loss": 0.0420468658208847,
676
+ "epoch": 0.8295625942684767,
677
+ "grad_norm": 0.6517713665962219,
678
+ "learning_rate": 1.0181060877486896e-05,
679
+ "loss": 0.17154401540756226,
680
+ "sft_loss": 0.0032584071159362793,
681
+ "step": 550,
682
+ "token_acc": 0.9595002540574469
683
+ },
684
+ {
685
+ "cos_sim": 0.9611032485961915,
686
+ "distill_lambda": 0.05,
687
+ "distill_loss": 0.03889682590961456,
688
+ "epoch": 0.8446455505279035,
689
+ "grad_norm": 0.791344165802002,
690
+ "learning_rate": 1.0005462499627443e-05,
691
+ "loss": 0.16251996755599976,
692
+ "sft_loss": 0.003133907914161682,
693
+ "step": 560,
694
+ "token_acc": 0.961120714961349
695
+ },
696
+ {
697
+ "cos_sim": 0.9566350936889648,
698
+ "distill_lambda": 0.05,
699
+ "distill_loss": 0.043365010619163515,
700
+ "epoch": 0.8597285067873304,
701
+ "grad_norm": 0.594643771648407,
702
+ "learning_rate": 9.82830416106839e-06,
703
+ "loss": 0.17398338317871093,
704
+ "sft_loss": 0.0032687306404113768,
705
+ "step": 570,
706
+ "token_acc": 0.9590278884462151
707
+ },
708
+ {
709
+ "cos_sim": 0.9601497650146484,
710
+ "distill_lambda": 0.05,
711
+ "distill_loss": 0.03985028862953186,
712
+ "epoch": 0.8748114630467572,
713
+ "grad_norm": 0.5415398478507996,
714
+ "learning_rate": 9.649696164815983e-06,
715
+ "loss": 0.16367310285568237,
716
+ "sft_loss": 0.003122270107269287,
717
+ "step": 580,
718
+ "token_acc": 0.9611103391845368
719
+ },
720
+ {
721
+ "cos_sim": 0.9599147796630859,
722
+ "distill_lambda": 0.05,
723
+ "distill_loss": 0.04008529782295227,
724
+ "epoch": 0.889894419306184,
725
+ "grad_norm": 0.6488097310066223,
726
+ "learning_rate": 9.469749716467922e-06,
727
+ "loss": 0.16612260341644286,
728
+ "sft_loss": 0.003187066316604614,
729
+ "step": 590,
730
+ "token_acc": 0.9604768853426061
731
+ },
732
+ {
733
+ "cos_sim": 0.9572768211364746,
734
+ "distill_lambda": 0.05,
735
+ "distill_loss": 0.04272326231002808,
736
+ "epoch": 0.9049773755656109,
737
+ "grad_norm": 0.6474986672401428,
738
+ "learning_rate": 9.288576854974107e-06,
739
+ "loss": 0.17227413654327392,
740
+ "sft_loss": 0.0032474040985107423,
741
+ "step": 600,
742
+ "token_acc": 0.9598335339239882
743
+ },
744
+ {
745
+ "cos_sim": 0.9604640960693359,
746
+ "distill_lambda": 0.05,
747
+ "distill_loss": 0.03953596353530884,
748
+ "epoch": 0.9200603318250377,
749
+ "grad_norm": 0.6288496851921082,
750
+ "learning_rate": 9.106290382878496e-06,
751
+ "loss": 0.16618789434432985,
752
+ "sft_loss": 0.0032165735960006715,
753
+ "step": 610,
754
+ "token_acc": 0.9597207796175231
755
+ },
756
+ {
757
+ "cos_sim": 0.9617090225219727,
758
+ "distill_lambda": 0.05,
759
+ "distill_loss": 0.038291019201278684,
760
+ "epoch": 0.9351432880844646,
761
+ "grad_norm": 0.6721214652061462,
762
+ "learning_rate": 8.923003796085571e-06,
763
+ "loss": 0.16136680841445922,
764
+ "sft_loss": 0.0031281620264053344,
765
+ "step": 620,
766
+ "token_acc": 0.961438223938224
767
+ },
768
+ {
769
+ "cos_sim": 0.9553379058837891,
770
+ "distill_lambda": 0.05,
771
+ "distill_loss": 0.04466201663017273,
772
+ "epoch": 0.9502262443438914,
773
+ "grad_norm": 0.5502645969390869,
774
+ "learning_rate": 8.738831213195076e-06,
775
+ "loss": 0.17655622959136963,
776
+ "sft_loss": 0.0032842814922332763,
777
+ "step": 630,
778
+ "token_acc": 0.9586743281676243
779
+ },
780
+ {
781
+ "cos_sim": 0.958277702331543,
782
+ "distill_lambda": 0.05,
783
+ "distill_loss": 0.041722333431243895,
784
+ "epoch": 0.9653092006033183,
785
+ "grad_norm": 0.6368167996406555,
786
+ "learning_rate": 8.553887304449051e-06,
787
+ "loss": 0.16567234992980956,
788
+ "sft_loss": 0.0030911445617675783,
789
+ "step": 640,
790
+ "token_acc": 0.9614270749892169
791
+ },
792
+ {
793
+ "cos_sim": 0.9605836868286133,
794
+ "distill_lambda": 0.05,
795
+ "distill_loss": 0.039416351914405824,
796
+ "epoch": 0.9803921568627451,
797
+ "grad_norm": 0.6817902326583862,
798
+ "learning_rate": 8.368287220335405e-06,
799
+ "loss": 0.16319282054901124,
800
+ "sft_loss": 0.0031289577484130858,
801
+ "step": 650,
802
+ "token_acc": 0.9607986668120299
803
+ },
804
+ {
805
+ "cos_sim": 0.9586556434631348,
806
+ "distill_lambda": 0.05,
807
+ "distill_loss": 0.0413442999124527,
808
+ "epoch": 0.995475113122172,
809
+ "grad_norm": 0.7272047400474548,
810
+ "learning_rate": 8.18214651989248e-06,
811
+ "loss": 0.16760696172714235,
812
+ "sft_loss": 0.003170502185821533,
813
+ "step": 660,
814
+ "token_acc": 0.9606587660660332
815
+ },
816
+ {
817
+ "cos_sim": 0.961158561706543,
818
+ "distill_lambda": 0.05,
819
+ "distill_loss": 0.03884148597717285,
820
+ "epoch": 1.0105580693815988,
821
+ "grad_norm": 0.4904056191444397,
822
+ "learning_rate": 7.99558109875924e-06,
823
+ "loss": 0.1609166979789734,
824
+ "sft_loss": 0.0030865728855133056,
825
+ "step": 670,
826
+ "token_acc": 0.9618842421799919
827
+ },
828
+ {
829
+ "cos_sim": 0.9569988250732422,
830
+ "distill_lambda": 0.05,
831
+ "distill_loss": 0.04300126731395722,
832
+ "epoch": 1.0256410256410255,
833
+ "grad_norm": 0.7804086804389954,
834
+ "learning_rate": 7.808707117015871e-06,
835
+ "loss": 0.1645439624786377,
836
+ "sft_loss": 0.002991935610771179,
837
+ "step": 680,
838
+ "token_acc": 0.9634873880474275
839
+ },
840
+ {
841
+ "cos_sim": 0.9613571166992188,
842
+ "distill_lambda": 0.05,
843
+ "distill_loss": 0.03864296674728394,
844
+ "epoch": 1.0407239819004526,
845
+ "grad_norm": 0.9397616982460022,
846
+ "learning_rate": 7.621640926859767e-06,
847
+ "loss": 0.1597062587738037,
848
+ "sft_loss": 0.0030586719512939453,
849
+ "step": 690,
850
+ "token_acc": 0.9618396829854285
851
+ },
852
+ {
853
+ "cos_sim": 0.9568060874938965,
854
+ "distill_lambda": 0.05,
855
+ "distill_loss": 0.04319390058517456,
856
+ "epoch": 1.0558069381598794,
857
+ "grad_norm": 0.723336398601532,
858
+ "learning_rate": 7.434499000161847e-06,
859
+ "loss": 0.16262754201889038,
860
+ "sft_loss": 0.0029224157333374023,
861
+ "step": 700,
862
+ "token_acc": 0.9633835241413639
863
+ },
864
+ {
865
+ "cos_sim": 0.9574602127075196,
866
+ "distill_lambda": 0.05,
867
+ "distill_loss": 0.04253982305526734,
868
+ "epoch": 1.0708898944193062,
869
+ "grad_norm": 0.6005541086196899,
870
+ "learning_rate": 7.247397855948417e-06,
871
+ "loss": 0.1620495080947876,
872
+ "sft_loss": 0.0029370561242103577,
873
+ "step": 710,
874
+ "token_acc": 0.9636207473832851
875
+ },
876
+ {
877
+ "cos_sim": 0.9617099761962891,
878
+ "distill_lambda": 0.05,
879
+ "distill_loss": 0.03829004764556885,
880
+ "epoch": 1.085972850678733,
881
+ "grad_norm": 0.7606406211853027,
882
+ "learning_rate": 7.060453987853635e-06,
883
+ "loss": 0.15295138359069824,
884
+ "sft_loss": 0.002865228056907654,
885
+ "step": 720,
886
+ "token_acc": 0.9648746675263964
887
+ },
888
+ {
889
+ "cos_sim": 0.957455825805664,
890
+ "distill_lambda": 0.05,
891
+ "distill_loss": 0.04254419207572937,
892
+ "epoch": 1.10105580693816,
893
+ "grad_norm": 0.542752206325531,
894
+ "learning_rate": 6.873783791587794e-06,
895
+ "loss": 0.1618901252746582,
896
+ "sft_loss": 0.0029318571090698243,
897
+ "step": 730,
898
+ "token_acc": 0.9640453202570585
899
+ },
900
+ {
901
+ "cos_sim": 0.9582402229309082,
902
+ "distill_lambda": 0.05,
903
+ "distill_loss": 0.04175975024700165,
904
+ "epoch": 1.1161387631975868,
905
+ "grad_norm": 0.5904515981674194,
906
+ "learning_rate": 6.6875034924665825e-06,
907
+ "loss": 0.15857551097869874,
908
+ "sft_loss": 0.002867497503757477,
909
+ "step": 740,
910
+ "token_acc": 0.9638847578222961
911
+ },
912
+ {
913
+ "cos_sim": 0.9587013244628906,
914
+ "distill_lambda": 0.05,
915
+ "distill_loss": 0.041298767924308775,
916
+ "epoch": 1.1312217194570136,
917
+ "grad_norm": 0.5564361214637756,
918
+ "learning_rate": 6.501729073046421e-06,
919
+ "loss": 0.15704817771911622,
920
+ "sft_loss": 0.002842816710472107,
921
+ "step": 750,
922
+ "token_acc": 0.96492840783385
923
+ },
924
+ {
925
+ "cos_sim": 0.9581723213195801,
926
+ "distill_lambda": 0.05,
927
+ "distill_loss": 0.041827693581581116,
928
+ "epoch": 1.1463046757164403,
929
+ "grad_norm": 0.4957558214664459,
930
+ "learning_rate": 6.316576200910982e-06,
931
+ "loss": 0.15571348667144774,
932
+ "sft_loss": 0.0027746617794036864,
933
+ "step": 760,
934
+ "token_acc": 0.965580921450632
935
+ },
936
+ {
937
+ "cos_sim": 0.9609840393066407,
938
+ "distill_lambda": 0.05,
939
+ "distill_loss": 0.039015993475914,
940
+ "epoch": 1.1613876319758674,
941
+ "grad_norm": 0.5658223032951355,
942
+ "learning_rate": 6.1321601566537614e-06,
943
+ "loss": 0.15808570384979248,
944
+ "sft_loss": 0.0029893785715103148,
945
+ "step": 770,
946
+ "token_acc": 0.963050224112171
947
+ },
948
+ {
949
+ "cos_sim": 0.9590757369995118,
950
+ "distill_lambda": 0.05,
951
+ "distill_loss": 0.040924301743507384,
952
+ "epoch": 1.1764705882352942,
953
+ "grad_norm": 0.8099390268325806,
954
+ "learning_rate": 5.94859576210165e-06,
955
+ "loss": 0.1593359351158142,
956
+ "sft_loss": 0.002933032810688019,
957
+ "step": 780,
958
+ "token_acc": 0.9637189262909925
959
+ },
960
+ {
961
+ "cos_sim": 0.9608880043029785,
962
+ "distill_lambda": 0.05,
963
+ "distill_loss": 0.03911195397377014,
964
+ "epoch": 1.191553544494721,
965
+ "grad_norm": 0.594108521938324,
966
+ "learning_rate": 5.765997308824127e-06,
967
+ "loss": 0.15664927959442138,
968
+ "sft_loss": 0.0029396921396255492,
969
+ "step": 790,
970
+ "token_acc": 0.9627525942108137
971
+ },
972
+ {
973
+ "cos_sim": 0.962888240814209,
974
+ "distill_lambda": 0.05,
975
+ "distill_loss": 0.03711185157299042,
976
+ "epoch": 1.2066365007541477,
977
+ "grad_norm": 0.5782893300056458,
978
+ "learning_rate": 5.584478486972594e-06,
979
+ "loss": 0.1509794235229492,
980
+ "sft_loss": 0.0028625145554542542,
981
+ "step": 800,
982
+ "token_acc": 0.9644990612732548
983
+ },
984
+ {
985
+ "cos_sim": 0.960543441772461,
986
+ "distill_lambda": 0.05,
987
+ "distill_loss": 0.03945652544498444,
988
+ "epoch": 1.2217194570135748,
989
+ "grad_norm": 0.6433768272399902,
990
+ "learning_rate": 5.404152314494185e-06,
991
+ "loss": 0.15497575998306273,
992
+ "sft_loss": 0.002870166301727295,
993
+ "step": 810,
994
+ "token_acc": 0.9647309912787126
995
+ },
996
+ {
997
+ "cos_sim": 0.9622140884399414,
998
+ "distill_lambda": 0.05,
999
+ "distill_loss": 0.03778596818447113,
1000
+ "epoch": 1.2368024132730016,
1001
+ "grad_norm": 0.6467146873474121,
1002
+ "learning_rate": 5.225131066764096e-06,
1003
+ "loss": 0.15006879568099976,
1004
+ "sft_loss": 0.00280035138130188,
1005
+ "step": 820,
1006
+ "token_acc": 0.9652448178082399
1007
+ },
1008
+ {
1009
+ "cos_sim": 0.9583595275878907,
1010
+ "distill_lambda": 0.05,
1011
+ "distill_loss": 0.04164048433303833,
1012
+ "epoch": 1.2518853695324283,
1013
+ "grad_norm": 0.5411479473114014,
1014
+ "learning_rate": 5.04752620668025e-06,
1015
+ "loss": 0.1611725091934204,
1016
+ "sft_loss": 0.002954617142677307,
1017
+ "step": 830,
1018
+ "token_acc": 0.9629328772979906
1019
+ },
1020
+ {
1021
+ "cos_sim": 0.9565027236938477,
1022
+ "distill_lambda": 0.05,
1023
+ "distill_loss": 0.043497335910797116,
1024
+ "epoch": 1.2669683257918551,
1025
+ "grad_norm": 0.7859143614768982,
1026
+ "learning_rate": 4.871448315263849e-06,
1027
+ "loss": 0.15971229076385499,
1028
+ "sft_loss": 0.0028161421418190003,
1029
+ "step": 840,
1030
+ "token_acc": 0.9648385764536664
1031
+ },
1032
+ {
1033
+ "cos_sim": 0.9596900939941406,
1034
+ "distill_lambda": 0.05,
1035
+ "distill_loss": 0.04030995666980743,
1036
+ "epoch": 1.282051282051282,
1037
+ "grad_norm": 0.606879711151123,
1038
+ "learning_rate": 4.697007022808989e-06,
1039
+ "loss": 0.15423134565353394,
1040
+ "sft_loss": 0.0028042316436767576,
1041
+ "step": 850,
1042
+ "token_acc": 0.9645077260079332
1043
+ },
1044
+ {
1045
+ "cos_sim": 0.9622067451477051,
1046
+ "distill_lambda": 0.05,
1047
+ "distill_loss": 0.037793251872062686,
1048
+ "epoch": 1.297134238310709,
1049
+ "grad_norm": 0.7979101538658142,
1050
+ "learning_rate": 4.52431094062422e-06,
1051
+ "loss": 0.15060091018676758,
1052
+ "sft_loss": 0.0028166159987449645,
1053
+ "step": 860,
1054
+ "token_acc": 0.9646602300549729
1055
+ },
1056
+ {
1057
+ "cos_sim": 0.9580653190612793,
1058
+ "distill_lambda": 0.05,
1059
+ "distill_loss": 0.041934633255004884,
1060
+ "epoch": 1.3122171945701357,
1061
+ "grad_norm": 0.6560086607933044,
1062
+ "learning_rate": 4.353467593408566e-06,
1063
+ "loss": 0.15815997123718262,
1064
+ "sft_loss": 0.002845767140388489,
1065
+ "step": 870,
1066
+ "token_acc": 0.964312429871587
1067
+ },
1068
+ {
1069
+ "cos_sim": 0.9604076385498047,
1070
+ "distill_lambda": 0.05,
1071
+ "distill_loss": 0.03959237039089203,
1072
+ "epoch": 1.3273001508295625,
1073
+ "grad_norm": 1.0168495178222656,
1074
+ "learning_rate": 4.1845833523040775e-06,
1075
+ "loss": 0.15515239238739015,
1076
+ "sft_loss": 0.0028688937425613405,
1077
+ "step": 880,
1078
+ "token_acc": 0.9635724852071006
1079
+ },
1080
+ {
1081
+ "cos_sim": 0.961815071105957,
1082
+ "distill_lambda": 0.05,
1083
+ "distill_loss": 0.03818497359752655,
1084
+ "epoch": 1.3423831070889896,
1085
+ "grad_norm": 0.5989195704460144,
1086
+ "learning_rate": 4.0177633686666225e-06,
1087
+ "loss": 0.14794611930847168,
1088
+ "sft_loss": 0.0027140676975250244,
1089
+ "step": 890,
1090
+ "token_acc": 0.966539761725032
1091
+ },
1092
+ {
1093
+ "cos_sim": 0.9618725776672363,
1094
+ "distill_lambda": 0.05,
1095
+ "distill_loss": 0.03812746405601501,
1096
+ "epoch": 1.3574660633484164,
1097
+ "grad_norm": 0.6104589700698853,
1098
+ "learning_rate": 3.853111508596134e-06,
1099
+ "loss": 0.14966075420379638,
1100
+ "sft_loss": 0.0027705252170562742,
1101
+ "step": 900,
1102
+ "token_acc": 0.9650014664190047
1103
+ },
1104
+ {
1105
+ "cos_sim": 0.9612789154052734,
1106
+ "distill_lambda": 0.05,
1107
+ "distill_loss": 0.03872108459472656,
1108
+ "epoch": 1.3725490196078431,
1109
+ "grad_norm": 0.667165219783783,
1110
+ "learning_rate": 3.6907302882671214e-06,
1111
+ "loss": 0.15173587799072266,
1112
+ "sft_loss": 0.0028056919574737547,
1113
+ "step": 910,
1114
+ "token_acc": 0.9655432747478407
1115
+ },
1116
+ {
1117
+ "cos_sim": 0.957188606262207,
1118
+ "distill_lambda": 0.05,
1119
+ "distill_loss": 0.042811453342437744,
1120
+ "epoch": 1.38763197586727,
1121
+ "grad_norm": 0.6276723146438599,
1122
+ "learning_rate": 3.5307208100996376e-06,
1123
+ "loss": 0.1576075553894043,
1124
+ "sft_loss": 0.002784663438796997,
1125
+ "step": 920,
1126
+ "token_acc": 0.9652169856794116
1127
+ },
1128
+ {
1129
+ "cos_sim": 0.9586119651794434,
1130
+ "distill_lambda": 0.05,
1131
+ "distill_loss": 0.041388103365898134,
1132
+ "epoch": 1.4027149321266967,
1133
+ "grad_norm": 0.7700992822647095,
1134
+ "learning_rate": 3.373182699810508e-06,
1135
+ "loss": 0.15658562183380126,
1136
+ "sft_loss": 0.002823895215988159,
1137
+ "step": 930,
1138
+ "token_acc": 0.9648340535074009
1139
+ },
1140
+ {
1141
+ "cos_sim": 0.9586081504821777,
1142
+ "distill_lambda": 0.05,
1143
+ "distill_loss": 0.04139181971549988,
1144
+ "epoch": 1.4177978883861238,
1145
+ "grad_norm": 0.5908275246620178,
1146
+ "learning_rate": 3.2182140443839834e-06,
1147
+ "loss": 0.15772662162780762,
1148
+ "sft_loss": 0.002859365940093994,
1149
+ "step": 940,
1150
+ "token_acc": 0.9637626109152789
1151
+ },
1152
+ {
1153
+ "cos_sim": 0.956392478942871,
1154
+ "distill_lambda": 0.05,
1155
+ "distill_loss": 0.043607497215271,
1156
+ "epoch": 1.4328808446455505,
1157
+ "grad_norm": 0.6041402816772461,
1158
+ "learning_rate": 3.0659113310004437e-06,
1159
+ "loss": 0.16355013847351074,
1160
+ "sft_loss": 0.002930566668510437,
1161
+ "step": 950,
1162
+ "token_acc": 0.9637660697035
1163
+ },
1164
+ {
1165
+ "cos_sim": 0.9604990959167481,
1166
+ "distill_lambda": 0.05,
1167
+ "distill_loss": 0.03950095772743225,
1168
+ "epoch": 1.4479638009049773,
1169
+ "grad_norm": 0.5494619011878967,
1170
+ "learning_rate": 2.9163693869611806e-06,
1171
+ "loss": 0.15139245986938477,
1172
+ "sft_loss": 0.002755966782569885,
1173
+ "step": 960,
1174
+ "token_acc": 0.9654504656810008
1175
+ },
1176
+ {
1177
+ "cos_sim": 0.9605469703674316,
1178
+ "distill_lambda": 0.05,
1179
+ "distill_loss": 0.03945311903953552,
1180
+ "epoch": 1.4630467571644044,
1181
+ "grad_norm": 0.5176012516021729,
1182
+ "learning_rate": 2.7696813206466494e-06,
1183
+ "loss": 0.1496511936187744,
1184
+ "sft_loss": 0.0027039438486099244,
1185
+ "step": 970,
1186
+ "token_acc": 0.967034284344282
1187
+ },
1188
+ {
1189
+ "cos_sim": 0.9582103729248047,
1190
+ "distill_lambda": 0.05,
1191
+ "distill_loss": 0.04178963303565979,
1192
+ "epoch": 1.4781297134238311,
1193
+ "grad_norm": 0.5375176668167114,
1194
+ "learning_rate": 2.6259384635449823e-06,
1195
+ "loss": 0.15224893093109132,
1196
+ "sft_loss": 0.002668297290802002,
1197
+ "step": 980,
1198
+ "token_acc": 0.966408741968469
1199
+ },
1200
+ {
1201
+ "cos_sim": 0.9583855628967285,
1202
+ "distill_lambda": 0.05,
1203
+ "distill_loss": 0.041614392399787904,
1204
+ "epoch": 1.493212669683258,
1205
+ "grad_norm": 0.6213409900665283,
1206
+ "learning_rate": 2.48523031338683e-06,
1207
+ "loss": 0.15863910913467408,
1208
+ "sft_loss": 0.0028767526149749757,
1209
+ "step": 990,
1210
+ "token_acc": 0.964980719890771
1211
+ },
1212
+ {
1213
+ "cos_sim": 0.9614453315734863,
1214
+ "distill_lambda": 0.05,
1215
+ "distill_loss": 0.03855462372303009,
1216
+ "epoch": 1.5082956259426847,
1217
+ "grad_norm": 0.6084774732589722,
1218
+ "learning_rate": 2.347644478421916e-06,
1219
+ "loss": 0.1520156145095825,
1220
+ "sft_loss": 0.0028227567672729492,
1221
+ "step": 1000,
1222
+ "token_acc": 0.9643499144340423
1223
+ },
1224
+ {
1225
+ "cos_sim": 0.9563823699951172,
1226
+ "distill_lambda": 0.05,
1227
+ "distill_loss": 0.043617653846740725,
1228
+ "epoch": 1.5233785822021115,
1229
+ "grad_norm": 0.5677129030227661,
1230
+ "learning_rate": 2.2132666228720783e-06,
1231
+ "loss": 0.16164510250091552,
1232
+ "sft_loss": 0.002870526909828186,
1233
+ "step": 1010,
1234
+ "token_acc": 0.9640724946695096
1235
+ },
1236
+ {
1237
+ "cos_sim": 0.9657289505004882,
1238
+ "distill_lambda": 0.05,
1239
+ "distill_loss": 0.03427100479602814,
1240
+ "epoch": 1.5384615384615383,
1241
+ "grad_norm": 0.5602020621299744,
1242
+ "learning_rate": 2.0821804135946685e-06,
1243
+ "loss": 0.14286451339721679,
1244
+ "sft_loss": 0.00275096595287323,
1245
+ "step": 1020,
1246
+ "token_acc": 0.9654056305981509
1247
+ },
1248
+ {
1249
+ "cos_sim": 0.9586674690246582,
1250
+ "distill_lambda": 0.05,
1251
+ "distill_loss": 0.04133264124393463,
1252
+ "epoch": 1.5535444947209653,
1253
+ "grad_norm": 0.5280652642250061,
1254
+ "learning_rate": 1.954467467989575e-06,
1255
+ "loss": 0.15333621501922606,
1256
+ "sft_loss": 0.0027251243591308594,
1257
+ "step": 1030,
1258
+ "token_acc": 0.9665621259639713
1259
+ },
1260
+ {
1261
+ "cos_sim": 0.959110164642334,
1262
+ "distill_lambda": 0.05,
1263
+ "distill_loss": 0.04088979959487915,
1264
+ "epoch": 1.5686274509803921,
1265
+ "grad_norm": 0.5409430265426636,
1266
+ "learning_rate": 1.8302073031822736e-06,
1267
+ "loss": 0.15340681076049806,
1268
+ "sft_loss": 0.0027494728565216064,
1269
+ "step": 1040,
1270
+ "token_acc": 0.9654812506617858
1271
+ },
1272
+ {
1273
+ "cos_sim": 0.961603832244873,
1274
+ "distill_lambda": 0.05,
1275
+ "distill_loss": 0.03839621543884277,
1276
+ "epoch": 1.5837104072398192,
1277
+ "grad_norm": 0.5610837340354919,
1278
+ "learning_rate": 1.7094772865145908e-06,
1279
+ "loss": 0.14765048027038574,
1280
+ "sft_loss": 0.00269426703453064,
1281
+ "step": 1050,
1282
+ "token_acc": 0.9670046289622475
1283
+ },
1284
+ {
1285
+ "cos_sim": 0.9625849723815918,
1286
+ "distill_lambda": 0.05,
1287
+ "distill_loss": 0.03741500377655029,
1288
+ "epoch": 1.598793363499246,
1289
+ "grad_norm": 0.6360013484954834,
1290
+ "learning_rate": 1.5923525873739312e-06,
1291
+ "loss": 0.14554901123046876,
1292
+ "sft_loss": 0.0026776567101478577,
1293
+ "step": 1060,
1294
+ "token_acc": 0.9659362770727379
1295
+ },
1296
+ {
1297
+ "cos_sim": 0.9599601745605468,
1298
+ "distill_lambda": 0.05,
1299
+ "distill_loss": 0.040039819478988645,
1300
+ "epoch": 1.6138763197586727,
1301
+ "grad_norm": 0.6609951257705688,
1302
+ "learning_rate": 1.4789061303910304e-06,
1303
+ "loss": 0.15372539758682252,
1304
+ "sft_loss": 0.002801927924156189,
1305
+ "step": 1070,
1306
+ "token_acc": 0.9649936158569952
1307
+ },
1308
+ {
1309
+ "cos_sim": 0.9597688674926758,
1310
+ "distill_lambda": 0.05,
1311
+ "distill_loss": 0.04023122191429138,
1312
+ "epoch": 1.6289592760180995,
1313
+ "grad_norm": 0.6310714483261108,
1314
+ "learning_rate": 1.3692085500353444e-06,
1315
+ "loss": 0.1543022871017456,
1316
+ "sft_loss": 0.0028103858232498167,
1317
+ "step": 1080,
1318
+ "token_acc": 0.9652636474623568
1319
+ },
1320
+ {
1321
+ "cos_sim": 0.9623384475708008,
1322
+ "distill_lambda": 0.05,
1323
+ "distill_loss": 0.037661558389663695,
1324
+ "epoch": 1.6440422322775263,
1325
+ "grad_norm": 0.5179605484008789,
1326
+ "learning_rate": 1.263328146636342e-06,
1327
+ "loss": 0.1467289924621582,
1328
+ "sft_loss": 0.0027022033929824827,
1329
+ "step": 1090,
1330
+ "token_acc": 0.9658165947211648
1331
+ },
1332
+ {
1333
+ "cos_sim": 0.9614273071289062,
1334
+ "distill_lambda": 0.05,
1335
+ "distill_loss": 0.03857271373271942,
1336
+ "epoch": 1.6591251885369531,
1337
+ "grad_norm": 0.47959592938423157,
1338
+ "learning_rate": 1.1613308438580924e-06,
1339
+ "loss": 0.149735426902771,
1340
+ "sft_loss": 0.0027505964040756225,
1341
+ "step": 1100,
1342
+ "token_acc": 0.9648848166735731
1343
+ },
1344
+ {
1345
+ "cos_sim": 0.9636540412902832,
1346
+ "distill_lambda": 0.05,
1347
+ "distill_loss": 0.03634599447250366,
1348
+ "epoch": 1.6742081447963801,
1349
+ "grad_norm": 0.527296781539917,
1350
+ "learning_rate": 1.063280147653618e-06,
1351
+ "loss": 0.14220519065856935,
1352
+ "sft_loss": 0.0026266127824783325,
1353
+ "step": 1110,
1354
+ "token_acc": 0.9672569546358804
1355
+ },
1356
+ {
1357
+ "cos_sim": 0.9571693420410157,
1358
+ "distill_lambda": 0.05,
1359
+ "distill_loss": 0.04283071160316467,
1360
+ "epoch": 1.689291101055807,
1361
+ "grad_norm": 0.4759458303451538,
1362
+ "learning_rate": 9.692371067245726e-07,
1363
+ "loss": 0.15967360734939576,
1364
+ "sft_loss": 0.0028482645750045775,
1365
+ "step": 1120,
1366
+ "token_acc": 0.9646659011203873
1367
+ },
1368
+ {
1369
+ "cos_sim": 0.9603818893432617,
1370
+ "distill_lambda": 0.05,
1371
+ "distill_loss": 0.03961814045906067,
1372
+ "epoch": 1.704374057315234,
1373
+ "grad_norm": 0.4827066659927368,
1374
+ "learning_rate": 8.792602745108685e-07,
1375
+ "loss": 0.14749515056610107,
1376
+ "sft_loss": 0.002628315985202789,
1377
+ "step": 1130,
1378
+ "token_acc": 0.9664777353291163
1379
+ },
1380
+ {
1381
+ "cos_sim": 0.9596490859985352,
1382
+ "distill_lambda": 0.05,
1383
+ "distill_loss": 0.040350842475891116,
1384
+ "epoch": 1.7194570135746607,
1385
+ "grad_norm": 0.5767756700515747,
1386
+ "learning_rate": 7.93405672733902e-07,
1387
+ "loss": 0.15450209379196167,
1388
+ "sft_loss": 0.0028106480836868285,
1389
+ "step": 1140,
1390
+ "token_acc": 0.9643665243125796
1391
+ },
1392
+ {
1393
+ "cos_sim": 0.9617944717407226,
1394
+ "distill_lambda": 0.05,
1395
+ "distill_loss": 0.03820553421974182,
1396
+ "epoch": 1.7345399698340875,
1397
+ "grad_norm": 0.5325683951377869,
1398
+ "learning_rate": 7.117267565160989e-07,
1399
+ "loss": 0.14688180685043334,
1400
+ "sft_loss": 0.002679780125617981,
1401
+ "step": 1150,
1402
+ "token_acc": 0.9659744184959571
1403
+ },
1404
+ {
1405
+ "cos_sim": 0.9576915740966797,
1406
+ "distill_lambda": 0.05,
1407
+ "distill_loss": 0.04230844974517822,
1408
+ "epoch": 1.7496229260935143,
1409
+ "grad_norm": 0.5952834486961365,
1410
+ "learning_rate": 6.342743810984836e-07,
1411
+ "loss": 0.1566309928894043,
1412
+ "sft_loss": 0.002779296040534973,
1413
+ "step": 1160,
1414
+ "token_acc": 0.9652321327662612
1415
+ },
1416
+ {
1417
+ "cos_sim": 0.9652015686035156,
1418
+ "distill_lambda": 0.05,
1419
+ "distill_loss": 0.03479845523834228,
1420
+ "epoch": 1.7647058823529411,
1421
+ "grad_norm": 0.563335120677948,
1422
+ "learning_rate": 5.610967701769814e-07,
1423
+ "loss": 0.141930890083313,
1424
+ "sft_loss": 0.0026954174041748046,
1425
+ "step": 1170,
1426
+ "token_acc": 0.9666916752633717
1427
+ },
1428
+ {
1429
+ "cos_sim": 0.9610443115234375,
1430
+ "distill_lambda": 0.05,
1431
+ "distill_loss": 0.03895579874515533,
1432
+ "epoch": 1.779788838612368,
1433
+ "grad_norm": 0.776590883731842,
1434
+ "learning_rate": 4.922394858772067e-07,
1435
+ "loss": 0.15091207027435302,
1436
+ "sft_loss": 0.0027682125568389893,
1437
+ "step": 1180,
1438
+ "token_acc": 0.9653806080889086
1439
+ },
1440
+ {
1441
+ "cos_sim": 0.9644537925720215,
1442
+ "distill_lambda": 0.05,
1443
+ "distill_loss": 0.035546216368675235,
1444
+ "epoch": 1.7948717948717947,
1445
+ "grad_norm": 0.482070654630661,
1446
+ "learning_rate": 4.2774540038639286e-07,
1447
+ "loss": 0.1390576958656311,
1448
+ "sft_loss": 0.0025682419538497923,
1449
+ "step": 1190,
1450
+ "token_acc": 0.967479937349631
1451
+ },
1452
+ {
1453
+ "cos_sim": 0.9611939430236817,
1454
+ "distill_lambda": 0.05,
1455
+ "distill_loss": 0.03880613148212433,
1456
+ "epoch": 1.8099547511312217,
1457
+ "grad_norm": 0.5717806816101074,
1458
+ "learning_rate": 3.676546692601465e-07,
1459
+ "loss": 0.14678553342819214,
1460
+ "sft_loss": 0.0026467412710189818,
1461
+ "step": 1200,
1462
+ "token_acc": 0.9670608902970794
1463
+ },
1464
+ {
1465
+ "cos_sim": 0.9606647491455078,
1466
+ "distill_lambda": 0.05,
1467
+ "distill_loss": 0.039335194230079654,
1468
+ "epoch": 1.8250377073906485,
1469
+ "grad_norm": 0.4626607298851013,
1470
+ "learning_rate": 3.1200470642063716e-07,
1471
+ "loss": 0.14852668046951295,
1472
+ "sft_loss": 0.002674698829650879,
1473
+ "step": 1210,
1474
+ "token_acc": 0.9666944341601876
1475
+ },
1476
+ {
1477
+ "cos_sim": 0.9623315811157227,
1478
+ "distill_lambda": 0.05,
1479
+ "distill_loss": 0.03766842782497406,
1480
+ "epoch": 1.8401206636500755,
1481
+ "grad_norm": 0.5528730154037476,
1482
+ "learning_rate": 2.6083016086180715e-07,
1483
+ "loss": 0.14226431846618653,
1484
+ "sft_loss": 0.00256233811378479,
1485
+ "step": 1220,
1486
+ "token_acc": 0.9690826869327639
1487
+ },
1488
+ {
1489
+ "cos_sim": 0.9570770263671875,
1490
+ "distill_lambda": 0.05,
1491
+ "distill_loss": 0.04292306005954742,
1492
+ "epoch": 1.8552036199095023,
1493
+ "grad_norm": 0.4962717592716217,
1494
+ "learning_rate": 2.14162895076071e-07,
1495
+ "loss": 0.1580639362335205,
1496
+ "sft_loss": 0.002793344855308533,
1497
+ "step": 1230,
1498
+ "token_acc": 0.9648224657206156
1499
+ },
1500
+ {
1501
+ "cos_sim": 0.9596912384033203,
1502
+ "distill_lambda": 0.05,
1503
+ "distill_loss": 0.040308722853660585,
1504
+ "epoch": 1.8702865761689291,
1505
+ "grad_norm": 0.5444379448890686,
1506
+ "learning_rate": 1.7203196521597354e-07,
1507
+ "loss": 0.14951192140579223,
1508
+ "sft_loss": 0.0026568114757537843,
1509
+ "step": 1240,
1510
+ "token_acc": 0.9667831067472307
1511
+ },
1512
+ {
1513
+ "cos_sim": 0.9610977172851562,
1514
+ "distill_lambda": 0.05,
1515
+ "distill_loss": 0.03890231251716614,
1516
+ "epoch": 1.885369532428356,
1517
+ "grad_norm": 0.4733203947544098,
1518
+ "learning_rate": 1.3446360300313375e-07,
1519
+ "loss": 0.14659295082092286,
1520
+ "sft_loss": 0.0026359140872955324,
1521
+ "step": 1250,
1522
+ "token_acc": 0.9674186654001424
1523
+ },
1524
+ {
1525
+ "cos_sim": 0.9640268325805664,
1526
+ "distill_lambda": 0.05,
1527
+ "distill_loss": 0.03597318828105926,
1528
+ "epoch": 1.9004524886877827,
1529
+ "grad_norm": 0.5802148580551147,
1530
+ "learning_rate": 1.0148119939575207e-07,
1531
+ "loss": 0.14078178405761718,
1532
+ "sft_loss": 0.0026007711887359617,
1533
+ "step": 1260,
1534
+ "token_acc": 0.9674666666666667
1535
+ },
1536
+ {
1537
+ "cos_sim": 0.9594644546508789,
1538
+ "distill_lambda": 0.05,
1539
+ "distill_loss": 0.04053559899330139,
1540
+ "epoch": 1.9155354449472095,
1541
+ "grad_norm": 0.4988519847393036,
1542
+ "learning_rate": 7.310529002484079e-08,
1543
+ "loss": 0.15229123830795288,
1544
+ "sft_loss": 0.002732321619987488,
1545
+ "step": 1270,
1546
+ "token_acc": 0.9661774555196413
1547
+ },
1548
+ {
1549
+ "cos_sim": 0.960145378112793,
1550
+ "distill_lambda": 0.05,
1551
+ "distill_loss": 0.03985467553138733,
1552
+ "epoch": 1.9306184012066365,
1553
+ "grad_norm": 0.5619922280311584,
1554
+ "learning_rate": 4.9353542408249465e-08,
1555
+ "loss": 0.15132608413696289,
1556
+ "sft_loss": 0.002736206352710724,
1557
+ "step": 1280,
1558
+ "token_acc": 0.965960607138989
1559
+ },
1560
+ {
1561
+ "cos_sim": 0.9620681762695312,
1562
+ "distill_lambda": 0.05,
1563
+ "distill_loss": 0.03793186545372009,
1564
+ "epoch": 1.9457013574660633,
1565
+ "grad_norm": 0.5861108303070068,
1566
+ "learning_rate": 3.024074495044826e-08,
1567
+ "loss": 0.1475370168685913,
1568
+ "sft_loss": 0.0027139380574226378,
1569
+ "step": 1290,
1570
+ "token_acc": 0.966789667896679
1571
+ },
1572
+ {
1573
+ "cos_sim": 0.9627554893493653,
1574
+ "distill_lambda": 0.05,
1575
+ "distill_loss": 0.03724450767040253,
1576
+ "epoch": 1.9607843137254903,
1577
+ "grad_norm": 0.5047047734260559,
1578
+ "learning_rate": 1.5778797734913307e-08,
1579
+ "loss": 0.1461432695388794,
1580
+ "sft_loss": 0.0027047514915466307,
1581
+ "step": 1300,
1582
+ "token_acc": 0.9653572346246277
1583
+ },
1584
+ {
1585
+ "cos_sim": 0.9619269371032715,
1586
+ "distill_lambda": 0.05,
1587
+ "distill_loss": 0.038073021173477176,
1588
+ "epoch": 1.9758672699849171,
1589
+ "grad_norm": 0.5808618664741516,
1590
+ "learning_rate": 5.976705114852654e-09,
1591
+ "loss": 0.14662744998931884,
1592
+ "sft_loss": 0.0026784569025039675,
1593
+ "step": 1310,
1594
+ "token_acc": 0.9660818900662675
1595
+ },
1596
+ {
1597
+ "cos_sim": 0.9581977844238281,
1598
+ "distill_lambda": 0.05,
1599
+ "distill_loss": 0.04180223047733307,
1600
+ "epoch": 1.990950226244344,
1601
+ "grad_norm": 0.513159990310669,
1602
+ "learning_rate": 8.40570106880878e-10,
1603
+ "loss": 0.15337657928466797,
1604
+ "sft_loss": 0.0027029067277908325,
1605
+ "step": 1320,
1606
+ "token_acc": 0.9666289507430037
1607
+ }
1608
+ ],
1609
+ "logging_steps": 10,
1610
+ "max_steps": 1326,
1611
+ "num_input_tokens_seen": 0,
1612
+ "num_train_epochs": 2,
1613
+ "save_steps": 500,
1614
+ "stateful_callbacks": {
1615
+ "TrainerControl": {
1616
+ "args": {
1617
+ "should_epoch_stop": false,
1618
+ "should_evaluate": false,
1619
+ "should_log": false,
1620
+ "should_save": true,
1621
+ "should_training_stop": true
1622
+ },
1623
+ "attributes": {}
1624
+ }
1625
+ },
1626
+ "total_flos": 1.6265711502005109e+19,
1627
+ "train_batch_size": 2,
1628
+ "trial_name": null,
1629
+ "trial_params": null
1630
+ }