Replace GRPO with evaluated checkpoint-490

GRPO/args.json

@@ -1,5 +1,5 @@
 {
-  "output_dir": "/home/tione/notebook/users/youqiwang/ForgeryVCR-Test/train/grpo/outputs/forgeryvcr_grpo_threshold_e6_ckpt2550_gbs32/v2-20260726-094546",
+  "output_dir": "/home/tione/notebook/users/youqiwang/FakeAgent/Agent/Zoom-In-Forensics-Agent/GRPO/qwen3-vl/4B_noCoT/train/IMD_FR/outputs/nocot_fullsft_5e-5_128_256_cls_loc_tooluse_20251229_034758/v0-20251229-034816",
   "overwrite_output_dir": false,
   "do_train": false,
   "do_eval": false,
@@ -20,8 +20,8 @@
   "adam_beta2": 0.95,
   "adam_epsilon": 1e-08,
   "max_grad_norm": 1.0,
-  "num_train_epochs": 2.0,
-  "max_steps": -1,
+  "num_train_epochs": 3.0,
+  "max_steps": 776,
   "lr_scheduler_type": "cosine",
   "lr_scheduler_kwargs": null,
   "warmup_ratio": 0.05,
@@ -29,7 +29,7 @@
   "log_level": "passive",
   "log_level_replica": "warning",
   "log_on_each_node": true,
-  "logging_dir": "/home/tione/notebook/users/youqiwang/ForgeryVCR-Test/train/grpo/outputs/forgeryvcr_grpo_threshold_e6_ckpt2550_gbs32/v2-20260726-094546/runs",
+  "logging_dir": "/home/tione/notebook/users/youqiwang/FakeAgent/Agent/Zoom-In-Forensics-Agent/GRPO/qwen3-vl/4B_noCoT/train/IMD_FR/outputs/nocot_fullsft_5e-5_128_256_cls_loc_tooluse_20251229_034758/v0-20251229-034816/runs",
   "logging_strategy": "steps",
   "logging_first_step": true,
   "logging_steps": 1,
@@ -64,7 +64,7 @@
   "dataloader_num_workers": 4,
   "dataloader_prefetch_factor": null,
   "past_index": -1,
-  "run_name": "forgeryvcr_grpo_threshold_e6_ckpt2550_gbs32",
+  "run_name": "/home/tione/notebook/users/youqiwang/FakeAgent/Agent/Zoom-In-Forensics-Agent/GRPO/qwen3-vl/4B_noCoT/train/IMD_FR/outputs/nocot_fullsft_5e-5_128_256_cls_loc_tooluse_20251229_034758/v0-20251229-034816",
   "disable_tqdm": null,
   "remove_unused_columns": false,
   "label_names": null,
@@ -72,7 +72,7 @@
   "metric_for_best_model": "loss",
   "greater_is_better": false,
   "ignore_data_skip": false,
-  "fsdp": null,
+  "fsdp": [],
   "fsdp_min_num_params": 0,
   "fsdp_config": null,
   "fsdp_transformer_layer_cls_to_wrap": null,
@@ -119,7 +119,7 @@
   "group_by_length": false,
   "length_column_name": "length",
   "report_to": [
-    "wandb"
+    "tensorboard"
   ],
   "project": "huggingface",
   "trackio_space_id": "trackio",
@@ -198,7 +198,7 @@
   "use_ray": false,
   "ray_exp_name": null,
   "device_groups": null,
-  "model": "/home/tione/notebook/users/youqiwang/ForgeryVCR-Test/train/sft/output_forgeryvcr_top4_top2_balanced_threshold_e6/v3-20260725-232416/checkpoint-2550-merged",
+  "model": "/home/tione/notebook/users/youqiwang/FakeAgent/Agent/Zoom-In-Forensics-Agent/SFT/qwen3-vl/4B/train/output_no_cot_full_2700steps_agent_zoomin_forensics_4B_832xgt_bbox/v4-20251228-102320/checkpoint-2736-merged",
   "model_type": "qwen3_vl",
   "model_revision": null,
   "task_type": "causal_lm",
@@ -228,8 +228,8 @@
   "response_prefix": null,
   "template_backend": "swift",
   "dataset": [
-    "/home/tione/notebook/users/youqiwang/ForgeryVCR-Test/data/rl/datasets/imd20_rl_data.json",
-    "/home/tione/notebook/users/youqiwang/ForgeryVCR-Test/data/rl/datasets/fr_rl_data.json"
+    "/home/tione/notebook/users/youqiwang/Agent_All/ForensicsTools/Agent/GRPO/Datasets/832x/imd20_rl_data_zoomin_forensics.json",
+    "/home/tione/notebook/users/youqiwang/Agent_All/ForensicsTools/Agent/GRPO/Datasets/832x/fr_rl_data_zoomin_forensics.json"
   ],
   "val_dataset": [],
   "cached_dataset": [],
@@ -266,12 +266,12 @@
   "stop_words": [],
   "logprobs": false,
   "top_logprobs": null,
-  "ckpt_dir": "/home/tione/notebook/users/youqiwang/ForgeryVCR-Test/train/sft/output_forgeryvcr_top4_top2_balanced_threshold_e6/v3-20260725-232416/checkpoint-2550-merged",
+  "ckpt_dir": "/home/tione/notebook/users/youqiwang/FakeAgent/Agent/Zoom-In-Forensics-Agent/SFT/qwen3-vl/4B/train/output_no_cot_full_2700steps_agent_zoomin_forensics_4B_832xgt_bbox/v4-20251228-102320/checkpoint-2736-merged",
   "lora_modules": [],
   "train_type": "lora",
   "adapters": [],
   "external_plugins": [
-    "/home/tione/notebook/users/youqiwang/ForgeryVCR-Test/train/grpo/plugin/plugin.py"
+    "/home/tione/notebook/users/youqiwang/FakeAgent/Agent/Zoom-In-Forensics-Agent/GRPO/qwen3-vl/4B_noCoT/train/plugin/plugin_nozoom.py"
   ],
   "model_kwargs": {},
   "load_args": false,
@@ -398,8 +398,8 @@
   "vllm_tensor_parallel_size": 1,
   "vllm_pipeline_parallel_size": 1,
   "vllm_enable_expert_parallel": false,
-  "vllm_max_num_seqs": 32,
-  "vllm_max_model_len": 16384,
+  "vllm_max_num_seqs": null,
+  "vllm_max_model_len": 10240,
   "vllm_disable_custom_all_reduce": true,
   "vllm_enforce_eager": false,
   "vllm_limit_mm_per_prompt": null,
@@ -409,7 +409,7 @@
   "vllm_quantization": null,
   "vllm_reasoning_parser": null,
   "vllm_disable_cascade_attn": false,
-  "vllm_mm_processor_cache_gb": 0.0,
+  "vllm_mm_processor_cache_gb": null,
   "vllm_speculative_config": null,
   "vllm_engine_kwargs": {},
   "vllm_data_parallel_size": 1,
@@ -461,7 +461,7 @@
   "tau_neg": 1.05,
   "advantage_estimator": "grpo",
   "kl_in_reward": false,
-  "generation_batch_size": 32,
+  "generation_batch_size": 256,
   "steps_per_generation": null,
   "rollout_importance_sampling_mode": null,
   "rollout_importance_sampling_threshold": 2.0,
@@ -501,12 +501,12 @@
   "rank": 0,
   "global_world_size": 8,
   "local_world_size": 8,
-  "model_suffix": "checkpoint-2550-merged",
-  "model_info": "ModelInfo(model_type='qwen3_vl', model_dir='/home/tione/notebook/users/youqiwang/ForgeryVCR-Test/train/sft/output_forgeryvcr_top4_top2_balanced_threshold_e6/v3-20260725-232416/checkpoint-2550-merged', torch_dtype=torch.bfloat16, max_model_len=262144, quant_method=None, quant_bits=None, rope_scaling={'mrope_interleaved': True, 'mrope_section': [24, 20, 20], 'rope_type': 'default'}, is_moe_model=False, is_multimodal=True, config=None, task_type='causal_lm', num_labels=None)",
-  "model_meta": "ModelMeta(model_type='qwen3_vl', model_groups=[ModelGroup(models=[Model(ms_model_id='Qwen/Qwen3-VL-2B-Instruct', hf_model_id='Qwen/Qwen3-VL-2B-Instruct', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-VL-2B-Thinking', hf_model_id='Qwen/Qwen3-VL-2B-Thinking', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-VL-2B-Instruct-FP8', hf_model_id='Qwen/Qwen3-VL-2B-Instruct-FP8', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-VL-2B-Thinking-FP8', hf_model_id='Qwen/Qwen3-VL-2B-Thinking-FP8', model_path=None, ms_revision=None, hf_revision=None), 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-4B-Thinking', hf_model_id='Qwen/Qwen3-VL-4B-Thinking', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-VL-4B-Instruct-FP8', hf_model_id='Qwen/Qwen3-VL-4B-Instruct-FP8', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-VL-4B-Thinking-FP8', hf_model_id='Qwen/Qwen3-VL-4B-Thinking-FP8', 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), Model(ms_model_id='Qwen/Qwen3-VL-8B-Thinking', hf_model_id='Qwen/Qwen3-VL-8B-Thinking', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-VL-8B-Instruct-FP8', hf_model_id='Qwen/Qwen3-VL-8B-Instruct-FP8', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-VL-8B-Thinking-FP8', hf_model_id='Qwen/Qwen3-VL-8B-Thinking-FP8', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-VL-32B-Instruct', hf_model_id='Qwen/Qwen3-VL-32B-Instruct', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-VL-32B-Thinking', hf_model_id='Qwen/Qwen3-VL-32B-Thinking', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-VL-32B-Instruct-FP8', hf_model_id='Qwen/Qwen3-VL-32B-Instruct-FP8', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-VL-32B-Thinking-FP8', hf_model_id='Qwen/Qwen3-VL-32B-Thinking-FP8', model_path=None, ms_revision=None, hf_revision=None)], ignore_patterns=None, requires=None, tags=[])], template='qwen3_vl', get_function=<function get_model_tokenizer_qwen3_vl at 0x7fb3602cc720>, model_arch=MultiModelKeys(arch_name='qwen3_vl', embedding=None, module_list=None, lm_head=None, q_proj=None, k_proj=None, v_proj=None, o_proj=None, attention=None, mlp=None, down_proj=None, qkv_proj=None, qk_proj=None, qa_proj=None, qb_proj=None, kv_proj=None, kva_proj=None, kvb_proj=None, language_model=['model.language_model', 'lm_head'], aligner=['model.visual.merger', 'model.visual.deepstack_merger_list'], vision_tower=['model.visual'], generator=[]), architectures=['Qwen3VLForConditionalGeneration'], additional_saved_files=[], torch_dtype=None, is_multimodal=True, is_reward=False, is_reranker=False, task_type=None, ignore_patterns=None, requires=['transformers>=4.57', 'qwen_vl_utils>=0.0.14', 'decord'], tags=['vision', 'video'])",
-  "model_dir": "/home/tione/notebook/users/youqiwang/ForgeryVCR-Test/train/sft/output_forgeryvcr_top4_top2_balanced_threshold_e6/v3-20260725-232416/checkpoint-2550-merged",
+  "model_suffix": "checkpoint-2736-merged",
+  "model_info": "ModelInfo(model_type='qwen3_vl', model_dir='/home/tione/notebook/users/youqiwang/FakeAgent/Agent/Zoom-In-Forensics-Agent/SFT/qwen3-vl/4B/train/output_no_cot_full_2700steps_agent_zoomin_forensics_4B_832xgt_bbox/v4-20251228-102320/checkpoint-2736-merged', torch_dtype=torch.bfloat16, max_model_len=262144, quant_method=None, quant_bits=None, rope_scaling={'mrope_interleaved': True, 'mrope_section': [24, 20, 20], 'rope_type': 'default'}, is_moe_model=False, is_multimodal=True, config=None, task_type='causal_lm', num_labels=None)",
+  "model_meta": "ModelMeta(model_type='qwen3_vl', model_groups=[ModelGroup(models=[Model(ms_model_id='Qwen/Qwen3-VL-2B-Instruct', hf_model_id='Qwen/Qwen3-VL-2B-Instruct', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-VL-2B-Thinking', hf_model_id='Qwen/Qwen3-VL-2B-Thinking', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-VL-2B-Instruct-FP8', hf_model_id='Qwen/Qwen3-VL-2B-Instruct-FP8', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-VL-2B-Thinking-FP8', hf_model_id='Qwen/Qwen3-VL-2B-Thinking-FP8', model_path=None, ms_revision=None, hf_revision=None), 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-4B-Thinking', hf_model_id='Qwen/Qwen3-VL-4B-Thinking', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-VL-4B-Instruct-FP8', hf_model_id='Qwen/Qwen3-VL-4B-Instruct-FP8', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-VL-4B-Thinking-FP8', hf_model_id='Qwen/Qwen3-VL-4B-Thinking-FP8', 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), Model(ms_model_id='Qwen/Qwen3-VL-8B-Thinking', hf_model_id='Qwen/Qwen3-VL-8B-Thinking', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-VL-8B-Instruct-FP8', hf_model_id='Qwen/Qwen3-VL-8B-Instruct-FP8', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-VL-8B-Thinking-FP8', hf_model_id='Qwen/Qwen3-VL-8B-Thinking-FP8', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-VL-32B-Instruct', hf_model_id='Qwen/Qwen3-VL-32B-Instruct', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-VL-32B-Thinking', hf_model_id='Qwen/Qwen3-VL-32B-Thinking', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-VL-32B-Instruct-FP8', hf_model_id='Qwen/Qwen3-VL-32B-Instruct-FP8', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-VL-32B-Thinking-FP8', hf_model_id='Qwen/Qwen3-VL-32B-Thinking-FP8', model_path=None, ms_revision=None, hf_revision=None)], ignore_patterns=None, requires=None, tags=[])], template='qwen3_vl', get_function=<function get_model_tokenizer_qwen3_vl at 0x7f67451ee7a0>, model_arch=MultiModelKeys(arch_name='qwen3_vl', embedding=None, module_list=None, lm_head=None, q_proj=None, k_proj=None, v_proj=None, o_proj=None, attention=None, mlp=None, down_proj=None, qkv_proj=None, qk_proj=None, qa_proj=None, qb_proj=None, kv_proj=None, kva_proj=None, kvb_proj=None, language_model=['model.language_model', 'lm_head'], aligner=['model.visual.merger', 'model.visual.deepstack_merger_list'], vision_tower=['model.visual'], generator=[]), architectures=['Qwen3VLForConditionalGeneration'], additional_saved_files=[], torch_dtype=None, is_multimodal=True, is_reward=False, is_reranker=False, task_type=None, ignore_patterns=None, requires=['transformers>=4.57', 'qwen_vl_utils>=0.0.14', 'decord'], tags=['vision', 'video'])",
+  "model_dir": "/home/tione/notebook/users/youqiwang/FakeAgent/Agent/Zoom-In-Forensics-Agent/SFT/qwen3-vl/4B/train/output_no_cot_full_2700steps_agent_zoomin_forensics_4B_832xgt_bbox/v4-20251228-102320/checkpoint-2736-merged",
   "_val_dataset_exists": [],
   "hub": "<class 'swift.hub.hub.MSHub'>",
   "evaluation_strategy": "steps",
-  "training_args": "GRPOConfig(output_dir='/home/tione/notebook/users/youqiwang/ForgeryVCR-Test/train/grpo/outputs/forgeryvcr_grpo_threshold_e6_ckpt2550_gbs32/v2-20260726-094546', overwrite_output_dir=False, do_train=False, do_eval=False, do_predict=False, eval_strategy=<IntervalStrategy.NO: 'no'>, prediction_loss_only=False, per_device_train_batch_size=4, per_device_eval_batch_size=1, per_gpu_train_batch_size=None, per_gpu_eval_batch_size=None, gradient_accumulation_steps=8, eval_accumulation_steps=None, eval_delay=0, torch_empty_cache_steps=None, learning_rate=5e-05, weight_decay=0.1, adam_beta1=0.9, adam_beta2=0.95, adam_epsilon=1e-08, max_grad_norm=1.0, num_train_epochs=2.0, max_steps=-1, lr_scheduler_type=<SchedulerType.COSINE: 'cosine'>, lr_scheduler_kwargs=None, warmup_ratio=0.05, warmup_steps=0, log_level='passive', log_level_replica='warning', log_on_each_node=True, logging_dir='/home/tione/notebook/users/youqiwang/ForgeryVCR-Test/train/grpo/outputs/forgeryvcr_grpo_threshold_e6_ckpt2550_gbs32/v2-20260726-094546/runs', logging_strategy=<IntervalStrategy.STEPS: 'steps'>, logging_first_step=True, logging_steps=1, logging_nan_inf_filter=True, save_strategy=<SaveStrategy.STEPS: 'steps'>, save_steps=10, save_total_limit=100, save_safetensors=True, save_on_each_node=False, save_only_model=False, restore_callback_states_from_checkpoint=False, no_cuda=False, use_cpu=False, use_mps_device=False, seed=42, data_seed=42, jit_mode_eval=False, bf16=True, fp16=False, fp16_opt_level='O1', half_precision_backend='auto', bf16_full_eval=False, fp16_full_eval=False, tf32=None, local_rank=0, ddp_backend=None, tpu_num_cores=None, tpu_metrics_debug=False, debug=[], dataloader_drop_last=True, eval_steps=10.0, dataloader_num_workers=4, dataloader_prefetch_factor=10, past_index=-1, run_name='forgeryvcr_grpo_threshold_e6_ckpt2550_gbs32', disable_tqdm=True, remove_unused_columns=False, label_names=None, load_best_model_at_end=False, metric_for_best_model='loss', greater_is_better=False, ignore_data_skip=False, fsdp=[], fsdp_min_num_params=0, fsdp_config={'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}, fsdp_transformer_layer_cls_to_wrap=None, 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, 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}, label_smoothing_factor=0.0, optim=<OptimizerNames.ADAMW_TORCH_FUSED: 'adamw_torch_fused'>, optim_args=None, adafactor=False, group_by_length=False, length_column_name='length', report_to=['wandb'], project='huggingface', trackio_space_id='trackio', ddp_find_unused_parameters=None, ddp_bucket_cap_mb=None, ddp_broadcast_buffers=None, dataloader_pin_memory=True, dataloader_persistent_workers=False, skip_memory_metrics=True, use_legacy_prediction_loop=False, push_to_hub=False, resume_from_checkpoint=None, hub_model_id=None, hub_strategy=<HubStrategy.EVERY_SAVE: 'every_save'>, hub_token=None, hub_private_repo=None, hub_always_push=False, hub_revision=None, gradient_checkpointing=True, gradient_checkpointing_kwargs={'use_reentrant': False}, include_inputs_for_metrics=False, include_for_metrics=[], eval_do_concat_batches=True, fp16_backend='auto', push_to_hub_model_id=None, push_to_hub_organization=None, push_to_hub_token=None, mp_parameters='', auto_find_batch_size=False, full_determinism=False, torchdynamo=None, ray_scope='last', ddp_timeout=18000000, torch_compile=False, torch_compile_backend=None, torch_compile_mode=None, include_tokens_per_second=None, include_num_input_tokens_seen=None, neftune_noise_alpha=None, optim_target_modules=None, batch_eval_metrics=False, eval_on_start=False, use_liger_kernel=False, liger_kernel_config=None, eval_use_gather_object=False, average_tokens_across_devices=None, model_init_kwargs=None, disable_dropout=False, max_prompt_length=512, num_generations=8, max_completion_length=2048, ds3_gather_for_generation=True, shuffle_dataset=True, generation_batch_size=32, steps_per_generation=1, temperature=1.0, top_p=1.0, top_k=0, min_p=None, generation_kwargs=None, repetition_penalty=1.0, use_transformers_paged=False, cache_implementation=None, use_vllm=True, vllm_mode='colocate', vllm_model_impl='vllm', vllm_enable_sleep_mode=False, vllm_guided_decoding_regex=None, vllm_server_base_url=None, vllm_server_host=None, vllm_server_port=[8000], vllm_server_timeout=240.0, vllm_gpu_memory_utilization=0.4, vllm_tensor_parallel_size=1, beta=0.1, num_iterations=1, epsilon=0.2, delta=None, epsilon_high=None, importance_sampling_level='sequence', reward_weights=None, scale_rewards='group', loss_type='grpo', mask_truncated_completions=False, sync_ref_model=False, ref_model_mixup_alpha=0.6, ref_model_sync_steps=512, top_entropy_quantile=1.0, use_liger_loss=False, vllm_importance_sampling_correction=True, vllm_importance_sampling_cap=2.0, log_completions=False, num_completions_to_print=None, wandb_log_unique_prompts=None, tuner_backend='peft', vit_gradient_checkpointing=True, router_aux_loss_coef=0.0, enable_dft_loss=False, enable_channel_loss=False, check_model=True, acc_strategy='token', train_dataloader_shuffle=True, max_epochs=None, aligner_lr=None, vit_lr=None, use_logits_to_keep=None, resume_only_model=False, optimizer=None, metric=None, eval_use_evalscope=False, eval_dataset=[], eval_dataset_args=None, eval_limit=None, eval_generation_config=None, extra_eval_args=None, use_flash_ckpt=False, sft_alpha=0, chord_sft_dataset=[], chord_sft_per_device_train_batch_size=None, chord_enable_phi_function=False, chord_mu_warmup_steps=None, chord_mu_decay_steps=None, chord_mu_peak=None, chord_mu_valley=None, train_type='lora', local_repo_path=None, galore_config=None, padding_side='right', padding_free=False, task_type='causal_lm', problem_type=None, vllm_pipeline_parallel_size=1, vllm_enable_expert_parallel=False, vllm_max_num_seqs=32, vllm_max_model_len=16384, vllm_disable_custom_all_reduce=True, vllm_enforce_eager=False, vllm_limit_mm_per_prompt=None, vllm_max_lora_rank=16, vllm_enable_prefix_caching=True, vllm_use_async_engine=True, vllm_quantization=None, vllm_reasoning_parser=None, vllm_disable_cascade_attn=False, vllm_mm_processor_cache_gb=0.0, vllm_speculative_config=None, vllm_engine_kwargs={}, vllm_data_parallel_size=1, stop_words=[], vllm_enable_lora=False, lora_rank=128, vllm_server_group_port=None, enable_flattened_weight_sync=True, async_generate=False, sleep_level=0, move_model_batches=None, offload_optimizer=False, offload_model=False, cosine_min_len_value_wrong=-0.5, cosine_max_len_value_wrong=0.0, cosine_min_len_value_correct=1.0, cosine_max_len_value_correct=0.5, cosine_max_len=2048, repetition_n_grams=3, repetition_max_penalty=-1.0, reward_model=None, reward_model_plugin=None, multi_turn_scheduler='forgery_detection_scheduler', max_turns=8, completion_length_limit_scope='per_round', vllm_server_pass_dataset=False, dynamic_sample=False, max_resample_times=3, overlong_filter=False, soft_max_length=None, soft_cache_length=None, log_entropy=True, tau_pos=1.0, tau_neg=1.05, advantage_estimator='grpo', kl_in_reward=False, dataset_shuffle=True, rollout_importance_sampling_mode=None, rollout_importance_sampling_threshold=2.0, log_rollout_offpolicy_metrics=False)"
+  "training_args": "GRPOConfig(output_dir='/home/tione/notebook/users/youqiwang/FakeAgent/Agent/Zoom-In-Forensics-Agent/GRPO/qwen3-vl/4B_noCoT/train/IMD_FR/outputs/nocot_fullsft_5e-5_128_256_cls_loc_tooluse_20251229_034758/v0-20251229-034816', overwrite_output_dir=False, do_train=False, do_eval=False, do_predict=False, eval_strategy=<IntervalStrategy.NO: 'no'>, prediction_loss_only=False, per_device_train_batch_size=4, per_device_eval_batch_size=1, per_gpu_train_batch_size=None, per_gpu_eval_batch_size=None, gradient_accumulation_steps=8, eval_accumulation_steps=None, eval_delay=0, torch_empty_cache_steps=None, learning_rate=5e-05, weight_decay=0.1, adam_beta1=0.9, adam_beta2=0.95, adam_epsilon=1e-08, max_grad_norm=1.0, num_train_epochs=3.0, max_steps=776, lr_scheduler_type=<SchedulerType.COSINE: 'cosine'>, lr_scheduler_kwargs=None, warmup_ratio=0.05, warmup_steps=0, log_level='passive', log_level_replica='warning', log_on_each_node=True, logging_dir='/home/tione/notebook/users/youqiwang/FakeAgent/Agent/Zoom-In-Forensics-Agent/GRPO/qwen3-vl/4B_noCoT/train/IMD_FR/outputs/nocot_fullsft_5e-5_128_256_cls_loc_tooluse_20251229_034758/v0-20251229-034816/runs', logging_strategy=<IntervalStrategy.STEPS: 'steps'>, logging_first_step=True, logging_steps=1, logging_nan_inf_filter=True, save_strategy=<SaveStrategy.STEPS: 'steps'>, save_steps=10, save_total_limit=100, save_safetensors=True, save_on_each_node=False, save_only_model=False, restore_callback_states_from_checkpoint=False, no_cuda=False, use_cpu=False, use_mps_device=False, seed=42, data_seed=42, jit_mode_eval=False, bf16=True, fp16=False, fp16_opt_level='O1', half_precision_backend='auto', bf16_full_eval=False, fp16_full_eval=False, tf32=None, local_rank=0, ddp_backend=None, tpu_num_cores=None, tpu_metrics_debug=False, debug=[], dataloader_drop_last=True, eval_steps=10.0, dataloader_num_workers=4, dataloader_prefetch_factor=10, past_index=-1, run_name='/home/tione/notebook/users/youqiwang/FakeAgent/Agent/Zoom-In-Forensics-Agent/GRPO/qwen3-vl/4B_noCoT/train/IMD_FR/outputs/nocot_fullsft_5e-5_128_256_cls_loc_tooluse_20251229_034758/v0-20251229-034816', disable_tqdm=True, remove_unused_columns=False, label_names=None, load_best_model_at_end=False, metric_for_best_model='loss', greater_is_better=False, ignore_data_skip=False, fsdp=[], fsdp_min_num_params=0, fsdp_config={'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}, fsdp_transformer_layer_cls_to_wrap=None, 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, 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}, label_smoothing_factor=0.0, optim=<OptimizerNames.ADAMW_TORCH_FUSED: 'adamw_torch_fused'>, optim_args=None, adafactor=False, group_by_length=False, length_column_name='length', report_to=['tensorboard'], project='huggingface', trackio_space_id='trackio', ddp_find_unused_parameters=None, ddp_bucket_cap_mb=None, ddp_broadcast_buffers=None, dataloader_pin_memory=True, dataloader_persistent_workers=False, skip_memory_metrics=True, use_legacy_prediction_loop=False, push_to_hub=False, resume_from_checkpoint=None, hub_model_id=None, hub_strategy=<HubStrategy.EVERY_SAVE: 'every_save'>, hub_token=None, hub_private_repo=None, hub_always_push=False, hub_revision=None, gradient_checkpointing=True, gradient_checkpointing_kwargs={'use_reentrant': False}, include_inputs_for_metrics=False, include_for_metrics=[], eval_do_concat_batches=True, fp16_backend='auto', push_to_hub_model_id=None, push_to_hub_organization=None, push_to_hub_token=None, mp_parameters='', auto_find_batch_size=False, full_determinism=False, torchdynamo=None, ray_scope='last', ddp_timeout=18000000, torch_compile=False, torch_compile_backend=None, torch_compile_mode=None, include_tokens_per_second=None, include_num_input_tokens_seen=None, neftune_noise_alpha=None, optim_target_modules=None, batch_eval_metrics=False, eval_on_start=False, use_liger_kernel=False, liger_kernel_config=None, eval_use_gather_object=False, average_tokens_across_devices=None, model_init_kwargs=None, disable_dropout=False, max_prompt_length=512, num_generations=8, max_completion_length=2048, ds3_gather_for_generation=True, shuffle_dataset=True, generation_batch_size=256, steps_per_generation=8, temperature=1.0, top_p=1.0, top_k=0, min_p=None, generation_kwargs=None, repetition_penalty=1.0, use_transformers_paged=False, cache_implementation=None, use_vllm=True, vllm_mode='colocate', vllm_model_impl='vllm', vllm_enable_sleep_mode=False, vllm_guided_decoding_regex=None, vllm_server_base_url=None, vllm_server_host=None, vllm_server_port=[8000], vllm_server_timeout=240.0, vllm_gpu_memory_utilization=0.4, vllm_tensor_parallel_size=1, beta=0.1, num_iterations=1, epsilon=0.2, delta=None, epsilon_high=None, importance_sampling_level='sequence', reward_weights=None, scale_rewards='group', loss_type='grpo', mask_truncated_completions=False, sync_ref_model=False, ref_model_mixup_alpha=0.6, ref_model_sync_steps=512, top_entropy_quantile=1.0, use_liger_loss=False, vllm_importance_sampling_correction=True, vllm_importance_sampling_cap=2.0, log_completions=False, num_completions_to_print=None, wandb_log_unique_prompts=None, tuner_backend='peft', vit_gradient_checkpointing=True, router_aux_loss_coef=0.0, enable_dft_loss=False, enable_channel_loss=False, check_model=True, acc_strategy='token', train_dataloader_shuffle=True, max_epochs=None, aligner_lr=None, vit_lr=None, use_logits_to_keep=None, resume_only_model=False, optimizer=None, metric=None, eval_use_evalscope=False, eval_dataset=[], eval_dataset_args=None, eval_limit=None, eval_generation_config=None, extra_eval_args=None, use_flash_ckpt=False, sft_alpha=0, chord_sft_dataset=[], chord_sft_per_device_train_batch_size=None, chord_enable_phi_function=False, chord_mu_warmup_steps=None, chord_mu_decay_steps=None, chord_mu_peak=None, chord_mu_valley=None, train_type='lora', local_repo_path=None, galore_config=None, padding_side='right', padding_free=False, task_type='causal_lm', problem_type=None, vllm_pipeline_parallel_size=1, vllm_enable_expert_parallel=False, vllm_max_num_seqs=None, vllm_max_model_len=10240, vllm_disable_custom_all_reduce=True, vllm_enforce_eager=False, vllm_limit_mm_per_prompt=None, vllm_max_lora_rank=16, vllm_enable_prefix_caching=True, vllm_use_async_engine=True, vllm_quantization=None, vllm_reasoning_parser=None, vllm_disable_cascade_attn=False, vllm_mm_processor_cache_gb=None, vllm_speculative_config=None, vllm_engine_kwargs={}, vllm_data_parallel_size=1, stop_words=[], vllm_enable_lora=False, lora_rank=128, vllm_server_group_port=None, enable_flattened_weight_sync=True, async_generate=False, sleep_level=0, move_model_batches=None, offload_optimizer=False, offload_model=False, cosine_min_len_value_wrong=-0.5, cosine_max_len_value_wrong=0.0, cosine_min_len_value_correct=1.0, cosine_max_len_value_correct=0.5, cosine_max_len=2048, repetition_n_grams=3, repetition_max_penalty=-1.0, reward_model=None, reward_model_plugin=None, multi_turn_scheduler='forgery_detection_scheduler', max_turns=8, completion_length_limit_scope='per_round', vllm_server_pass_dataset=False, dynamic_sample=False, max_resample_times=3, overlong_filter=False, soft_max_length=None, soft_cache_length=None, log_entropy=True, tau_pos=1.0, tau_neg=1.05, advantage_estimator='grpo', kl_in_reward=False, dataset_shuffle=True, rollout_importance_sampling_mode=None, rollout_importance_sampling_threshold=2.0, log_rollout_offpolicy_metrics=False)"
 }

GRPO/model-00001-of-00002.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:99537f539418632ecedc26fb6046efc1d63d4551ab83a39bf8c1e95e65e4511a
+oid sha256:e59e475ad9b46eebce6cef566bd568d5f97da9a3af441957e835fa1414a8fa84
 size 4990497880

GRPO/model-00002-of-00002.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8e963a82eb7334ee30dbb14266f6ae13fb3e13854bda0b3f4be6aedae19745ab
+oid sha256:fd04deabb603f07779ae732379a46bea7f01b9481992321afc7f4c784d66bc4d
 size 3885221448