args.json

{
  "use_ray": false,
  "ray_exp_name": null,
  "device_groups": null,
  "model": "Qwen/Qwen3-VL-30B-A3B-Instruct",
  "model_type": "qwen3_moe_vl",
  "model_revision": null,
  "task_type": "causal_lm",
  "torch_dtype": "bfloat16",
  "attn_impl": null,
  "new_special_tokens": [],
  "num_labels": null,
  "problem_type": null,
  "rope_scaling": {
    "mrope_interleaved": true,
    "mrope_section": [
      24,
      20,
      20
    ],
    "rope_type": "default"
  },
  "device_map": null,
  "max_memory": {},
  "max_model_len": null,
  "local_repo_path": null,
  "init_strategy": null,
  "template": "qwen3_vl",
  "system": null,
  "max_length": 32768,
  "truncation_strategy": "delete",
  "max_pixels": null,
  "agent_template": null,
  "norm_bbox": null,
  "use_chat_template": true,
  "padding_free": true,
  "padding_side": "right",
  "loss_scale": "default",
  "sequence_parallel_size": 1,
  "response_prefix": null,
  "template_backend": "swift",
  "dataset": [
    "browser-use/sft-merged-all-modes-101225"
  ],
  "val_dataset": [],
  "cached_dataset": [],
  "cached_val_dataset": [],
  "split_dataset_ratio": 0.01,
  "data_seed": 42,
  "dataset_num_proc": 160,
  "load_from_cache_file": false,
  "dataset_shuffle": true,
  "val_dataset_shuffle": false,
  "streaming": false,
  "interleave_prob": null,
  "stopping_strategy": "first_exhausted",
  "shuffle_buffer_size": 1000,
  "download_mode": "reuse_dataset_if_exists",
  "columns": {},
  "strict": false,
  "remove_unused_columns": true,
  "model_name": null,
  "model_author": null,
  "custom_dataset_info": [],
  "quant_method": null,
  "quant_bits": null,
  "hqq_axis": null,
  "bnb_4bit_compute_dtype": "bfloat16",
  "bnb_4bit_quant_type": "nf4",
  "bnb_4bit_use_double_quant": true,
  "bnb_4bit_quant_storage": null,
  "max_new_tokens": null,
  "temperature": null,
  "top_k": 50,
  "top_p": 0.9,
  "repetition_penalty": 1.0,
  "num_beams": 1,
  "stream": false,
  "stop_words": [],
  "logprobs": false,
  "top_logprobs": null,
  "ckpt_dir": null,
  "lora_modules": [],
  "tuner_backend": "peft",
  "train_type": "full",
  "adapters": [],
  "external_plugins": [],
  "seed": 42,
  "model_kwargs": {},
  "load_args": false,
  "load_data_args": false,
  "packing": true,
  "packing_length": 32768,
  "packing_num_proc": 1,
  "lazy_tokenize": false,
  "custom_register_path": [],
  "use_hf": true,
  "hub_token": null,
  "ddp_timeout": 18000000,
  "ddp_backend": null,
  "ignore_args_error": false,
  "use_swift_lora": false,
  "freeze_llm": false,
  "freeze_vit": true,
  "freeze_aligner": true,
  "freeze_parameters": [],
  "freeze_parameters_regex": null,
  "freeze_parameters_ratio": 0.0,
  "trainable_parameters": [],
  "trainable_parameters_regex": null,
  "adapter_load": null,
  "target_modules": [
    "all-linear"
  ],
  "target_regex": null,
  "modules_to_save": [],
  "lora_rank": 8,
  "lora_alpha": 32,
  "lora_dropout": 0.05,
  "lora_bias": "none",
  "lora_dtype": null,
  "use_rslora": false,
  "rlhf_type": null,
  "ref_load": null,
  "ref_adapter_load": null,
  "beta": 0.1,
  "rpo_alpha": null,
  "reference_free": false,
  "label_smoothing": 0.0,
  "f_divergence_type": "reverse_kl",
  "loss_type": null,
  "desirable_weight": 1.0,
  "undesirable_weight": 1.0,
  "calculate_KL": null,
  "center_rewards_coefficient": null,
  "generation_batch_size": null,
  "steps_per_generation": null,
  "num_generations": 8,
  "max_completion_length": 512,
  "importance_sampling_level": "token",
  "tau_pos": 1.0,
  "tau_neg": 1.05,
  "epsilon": 0.2,
  "epsilon_high": null,
  "delta": null,
  "use_vllm": true,
  "vllm_mode": "colocate",
  "vllm_enable_prefix_caching": true,
  "vllm_gpu_memory_utilization": 0.9,
  "vllm_tensor_parallel_size": 1,
  "vllm_max_model_len": null,
  "vllm_enforce_eager": false,
  "vllm_limit_mm_per_prompt": null,
  "vllm_disable_cascade_attn": false,
  "sleep_level": 0,
  "offload_optimizer": false,
  "offload_model": false,
  "offload_bridge": false,
  "vllm_server_base_url": null,
  "vllm_server_host": null,
  "vllm_server_port": [
    8000
  ],
  "vllm_server_timeout": 240.0,
  "vllm_server_group_port": [
    51216
  ],
  "reward_funcs": [],
  "reward_weights": null,
  "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": null,
  "repetition_n_grams": 3,
  "repetition_max_penalty": -1.0,
  "soft_max_length": null,
  "soft_cache_length": null,
  "dynamic_sample": false,
  "max_resample_times": 3,
  "overlong_filter": false,
  "scale_rewards": "group",
  "advantage_estimator": "grpo",
  "kl_in_reward": false,
  "wandb_log_unique_prompts": null,
  "log_completions": false,
  "rollout_importance_sampling_mode": null,
  "rollout_importance_sampling_threshold": 2.0,
  "reward_model": null,
  "reward_model_plugin": null,
  "sync_ref_model": false,
  "ref_model_sync_steps": 512,
  "ref_model_mixup_alpha": 0.6,
  "async_generate": false,
  "move_model_batches": null,
  "multi_turn_scheduler": null,
  "max_turns": null,
  "completion_length_limit_scope": "per_round",
  "vllm_server_pass_dataset": false,
  "log_entropy": false,
  "top_entropy_quantile": 1.0,
  "num_iterations": 1,
  "padded_vocab_size": 151936,
  "initialize_embedding": false,
  "mlp_padding_free": false,
  "load_safetensors": true,
  "save_safetensors": true,
  "ref_model": null,
  "ref_adapters": [],
  "merge_lora": true,
  "max_shard_size": "5GB",
  "dataloader_persistent_workers": true,
  "dataloader_prefetch_factor": 10,
  "architectures": "Qwen3VLMoeForConditionalGeneration",
  "llm_architectures": "Qwen3VLMoeForConditionalGeneration",
  "max_epochs": 3,
  "enable_dft_loss": false,
  "enable_channel_loss": false,
  "save_strategy": "steps",
  "original_max_position_embeddings": null,
  "partial_rotary_factor": null,
  "use_shared_expert_gate": false,
  "vit_gradient_checkpointing": true,
  "vit_lr": null,
  "aligner_lr": null,
  "gradient_checkpointing_kwargs": null,
  "linear_num_value_heads": null,
  "linear_num_key_heads": null,
  "linear_key_head_dim": null,
  "linear_value_head_dim": null,
  "linear_conv_kernel_dim": null,
  "layer_types": null,
  "mrope_interleaved": true,
  "micro_batch_size": 1,
  "global_batch_size": 32,
  "recompute_granularity": "full",
  "recompute_method": "uniform",
  "recompute_num_layers": 1,
  "recompute_modules": [
    "core_attn"
  ],
  "use_cpu_initialization": false,
  "deterministic_mode": false,
  "train_iters": null,
  "log_interval": 5,
  "tensorboard_dir": "/home/ubuntu/mert/models/Qwen3-VL-30B-A3B-Instruct-allmodes-101225/v3-20251210-155100/runs",
  "no_masked_softmax_fusion": false,
  "no_bias_dropout_fusion": false,
  "no_bias_swiglu_fusion": false,
  "no_rope_fusion": false,
  "no_gradient_accumulation_fusion": false,
  "cross_entropy_loss_fusion": false,
  "cross_entropy_fusion_impl": "native",
  "calculate_per_token_loss": true,
  "use_flash_attn": false,
  "attention_backend": "flash",
  "optimizer": "adam",
  "optimizer_cpu_offload": true,
  "optimizer_offload_fraction": 0.0,
  "use_precision_aware_optimizer": true,
  "main_grads_dtype": "fp32",
  "main_params_dtype": "fp32",
  "exp_avg_dtype": "fp32",
  "exp_avg_sq_dtype": "fp32",
  "dataloader_type": "cyclic",
  "manual_gc": false,
  "manual_gc_interval": 0,
  "lr": 2e-05,
  "lr_decay_style": "cosine",
  "lr_decay_iters": null,
  "lr_warmup_iters": 0,
  "lr_warmup_fraction": 0.05,
  "min_lr": 2e-06,
  "weight_decay": 0.1,
  "clip_grad": 1.0,
  "adam_beta1": 0.9,
  "adam_beta2": 0.95,
  "adam_eps": 1e-08,
  "sgd_momentum": 0.9,
  "save": "/home/ubuntu/mert/models/Qwen3-VL-30B-A3B-Instruct-allmodes-101225/v3-20251210-155100",
  "save_interval": 200,
  "save_retain_interval": null,
  "no_save_optim": true,
  "no_save_rng": true,
  "load": null,
  "no_load_optim": false,
  "no_load_rng": false,
  "finetune": true,
  "ckpt_format": "torch_dist",
  "no_initialization": true,
  "auto_detect_ckpt_format": true,
  "exit_on_missing_checkpoint": true,
  "async_save": false,
  "use_persistent_ckpt_worker": false,
  "ckpt_fully_parallel_load": false,
  "ckpt_assume_constant_structure": false,
  "distributed_backend": "nccl",
  "local_rank": 0,
  "use_distributed_optimizer": true,
  "tensor_model_parallel_size": 1,
  "pipeline_model_parallel_size": 1,
  "decoder_first_pipeline_num_layers": null,
  "decoder_last_pipeline_num_layers": null,
  "account_for_embedding_in_pipeline_split": false,
  "account_for_loss_in_pipeline_split": false,
  "sequence_parallel": true,
  "context_parallel_size": 1,
  "tp_comm_overlap": false,
  "overlap_grad_reduce": false,
  "overlap_param_gather": false,
  "distributed_timeout_minutes": 300000,
  "num_layers_per_virtual_pipeline_stage": null,
  "num_virtual_stages_per_pipeline_rank": null,
  "microbatch_group_size_per_virtual_pipeline_stage": null,
  "pipeline_model_parallel_layout": null,
  "num_layers": 48,
  "hidden_size": 2048,
  "ffn_hidden_size": null,
  "num_attention_heads": 32,
  "group_query_attention": true,
  "num_query_groups": 4,
  "softmax_type": null,
  "window_size": null,
  "window_attn_skip_freq": null,
  "max_position_embeddings": 262144,
  "position_embedding_type": "mrope",
  "mrope_section": [
    24,
    20,
    20
  ],
  "rotary_base": 5000000,
  "rotary_percent": 1.0,
  "rotary_interleaved": false,
  "normalization": "RMSNorm",
  "norm_epsilon": 1e-06,
  "swiglu": true,
  "quick_geglu": false,
  "activation_func_clamp_value": null,
  "glu_linear_offset": null,
  "untie_embeddings_and_output_weights": true,
  "disable_bias_linear": true,
  "add_qkv_bias": false,
  "attention_dropout": 0.0,
  "hidden_dropout": 0.0,
  "kv_channels": 128,
  "qk_layernorm": true,
  "qk_l2_norm": null,
  "no_rope_freq": null,
  "moe_apply_probs_on_input": null,
  "transformer_impl": "transformer_engine",
  "num_experts": 128,
  "moe_layer_freq": "1",
  "moe_ffn_hidden_size": 768,
  "moe_shared_expert_intermediate_size": null,
  "moe_router_topk": 8,
  "moe_router_num_groups": null,
  "moe_router_group_topk": null,
  "moe_router_pre_softmax": false,
  "moe_router_dtype": "fp32",
  "moe_router_score_function": "softmax",
  "moe_router_bias_update_rate": null,
  "moe_router_enable_expert_bias": false,
  "moe_router_topk_scaling_factor": null,
  "moe_router_load_balancing_type": "aux_loss",
  "expert_model_parallel_size": 4,
  "expert_tensor_parallel_size": 1,
  "moe_token_dispatcher_type": null,
  "moe_enable_deepep": false,
  "moe_grouped_gemm": true,
  "moe_permute_fusion": true,
  "moe_aux_loss_coeff": 1e-06,
  "moe_z_loss_coeff": null,
  "moe_shared_expert_overlap": true,
  "moe_layer_recompute": false,
  "moe_expert_capacity_factor": 2.0,
  "moe_pad_expert_input_to_capacity": false,
  "moe_token_drop_policy": null,
  "multi_latent_attention": false,
  "q_lora_rank": null,
  "kv_lora_rank": 32,
  "qk_head_dim": 128,
  "qk_pos_emb_head_dim": 64,
  "mtp_num_layers": null,
  "mtp_loss_scaling_factor": 0.1,
  "fp8_format": null,
  "fp8_recipe": "delayed",
  "fp8_amax_history_len": 1024,
  "fp8_amax_compute_algo": "max",
  "fp8_param_gather": false,
  "fp16": false,
  "bf16": true,
  "apply_query_key_layer_scaling": false,
  "attention_softmax_in_fp32": true,
  "log_params_norm": false,
  "log_throughput": false,
  "tensorboard_log_interval": 1,
  "tensorboard_queue_size": 50,
  "log_timers_to_tensorboard": true,
  "no_log_learning_rate_to_tensorboard": false,
  "log_validation_ppl_to_tensorboard": true,
  "log_memory_to_tensorboard": true,
  "logging_level": null,
  "wandb_project": "browseruse",
  "wandb_exp_name": "qwenvl_30ba3b_megatron_allmodes_101225",
  "wandb_save_dir": "/home/ubuntu/custom-llm/train/megatron-swift/wandb/qwenvl_30ba3b_megatron_allmodes_101225",
  "eval_iters": -1,
  "eval_interval": 200,
  "seq_length": 32768,
  "num_workers": 160,
  "megatron_extra_kwargs": {},
  "add_version": true,
  "rank": 0,
  "global_world_size": 8,
  "local_world_size": 8,
  "model_suffix": "Qwen3-VL-30B-A3B-Instruct",
  "model_info": "ModelInfo(model_type='qwen3_moe_vl', model_dir='/root/.cache/huggingface/hub/models--Qwen--Qwen3-VL-30B-A3B-Instruct/snapshots/9c4b90e1e4ba969fd3b5378b57d966d725f1b86c', 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=True, config=None, task_type='causal_lm', num_labels=None)",
  "model_meta": "ModelMeta(model_type='qwen3_moe_vl', model_groups=[ModelGroup(models=[Model(ms_model_id='Qwen/Qwen3-VL-30B-A3B-Instruct', hf_model_id='Qwen/Qwen3-VL-30B-A3B-Instruct', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-VL-30B-A3B-Thinking', hf_model_id='Qwen/Qwen3-VL-30B-A3B-Thinking', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-VL-30B-A3B-Instruct-FP8', hf_model_id='Qwen/Qwen3-VL-30B-A3B-Instruct-FP8', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-VL-30B-A3B-Thinking-FP8', hf_model_id='Qwen/Qwen3-VL-30B-A3B-Thinking-FP8', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-VL-235B-A22B-Instruct', hf_model_id='Qwen/Qwen3-VL-235B-A22B-Instruct', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-VL-235B-A22B-Thinking', hf_model_id='Qwen/Qwen3-VL-235B-A22B-Thinking', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-VL-235B-A22B-Instruct-FP8', hf_model_id='Qwen/Qwen3-VL-235B-A22B-Instruct-FP8', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='Qwen/Qwen3-VL-235B-A22B-Thinking-FP8', hf_model_id='Qwen/Qwen3-VL-235B-A22B-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_moe_vl at 0x7367fbe0e3e0>, 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=['Qwen3VLMoeForConditionalGeneration'], 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": "/root/.cache/huggingface/hub/models--Qwen--Qwen3-VL-30B-A3B-Instruct/snapshots/9c4b90e1e4ba969fd3b5378b57d966d725f1b86c",
  "_val_dataset_exists": true,
  "hub": "<class 'swift.hub.hub.HFHub'>",
  "megatron_model_meta": "MegatronModelMeta(megatron_model_type='qwen3_vl', model_types=['qwen3_vl', 'qwen3_moe_vl'], is_multimodal=True, bridge_cls=<class 'swift.megatron.model.gpt_bridge.MultimodalGPTBridge'>, model_cls=<class 'swift.megatron.model.mm_gpt.qwen3_vl.Qwen3VLGPTModel'>, get_transformer_layer_spec=None, model_provider=<function model_provider at 0x73648e5ca2a0>, visual_cls=<class 'swift.megatron.model.mm_gpt.qwen3_vl.Qwen3VL_Vit'>, extra_args_provider=None)",
  "extra_args": {
    "model_dir": "/root/.cache/huggingface/hub/models--Qwen--Qwen3-VL-30B-A3B-Instruct/snapshots/9c4b90e1e4ba969fd3b5378b57d966d725f1b86c",
    "is_multimodal": true,
    "hf_model_type": "qwen3_moe_vl",
    "use_ray": false,
    "ray_exp_name": null,
    "device_groups": null,
    "model": "Qwen/Qwen3-VL-30B-A3B-Instruct",
    "model_type": "qwen3_moe_vl",
    "model_revision": null,
    "task_type": "causal_lm",
    "torch_dtype": "bfloat16",
    "attn_impl": null,
    "new_special_tokens": [],
    "num_labels": null,
    "problem_type": null,
    "rope_scaling": {
      "mrope_interleaved": true,
      "mrope_section": [
        24,
        20,
        20
      ],
      "rope_type": "default"
    },
    "device_map": null,
    "max_memory": {},
    "max_model_len": null,
    "local_repo_path": null,
    "init_strategy": null,
    "template": "qwen3_vl",
    "system": null,
    "max_length": 32768,
    "truncation_strategy": "delete",
    "max_pixels": null,
    "agent_template": null,
    "norm_bbox": null,
    "use_chat_template": true,
    "padding_free": true,
    "padding_side": "right",
    "sequence_parallel_size": 1,
    "response_prefix": null,
    "template_backend": "swift",
    "dataset": [
      "browser-use/sft-merged-all-modes-101225"
    ],
    "val_dataset": [],
    "cached_dataset": [],
    "cached_val_dataset": [],
    "split_dataset_ratio": 0.01,
    "data_seed": 42,
    "dataset_num_proc": 160,
    "load_from_cache_file": false,
    "dataset_shuffle": true,
    "val_dataset_shuffle": false,
    "streaming": false,
    "interleave_prob": null,
    "stopping_strategy": "first_exhausted",
    "shuffle_buffer_size": 1000,
    "download_mode": "reuse_dataset_if_exists",
    "columns": {},
    "strict": false,
    "remove_unused_columns": true,
    "model_name": null,
    "model_author": null,
    "custom_dataset_info": [],
    "quant_method": null,
    "quant_bits": null,
    "hqq_axis": null,
    "bnb_4bit_compute_dtype": "bfloat16",
    "bnb_4bit_quant_type": "nf4",
    "bnb_4bit_use_double_quant": true,
    "bnb_4bit_quant_storage": null,
    "max_new_tokens": null,
    "temperature": null,
    "top_k": 50,
    "top_p": 0.9,
    "repetition_penalty": 1.0,
    "num_beams": 1,
    "stream": false,
    "stop_words": [],
    "logprobs": false,
    "top_logprobs": null,
    "ckpt_dir": null,
    "lora_modules": [],
    "tuner_backend": "peft",
    "train_type": "full",
    "adapters": [],
    "external_plugins": [],
    "model_kwargs": {},
    "load_args": false,
    "load_data_args": false,
    "packing": true,
    "packing_length": 32768,
    "packing_num_proc": 1,
    "lazy_tokenize": false,
    "custom_register_path": [],
    "use_hf": true,
    "hub_token": null,
    "ddp_timeout": 18000000,
    "ddp_backend": null,
    "ignore_args_error": false,
    "use_swift_lora": false,
    "freeze_llm": false,
    "freeze_vit": true,
    "freeze_aligner": true,
    "freeze_parameters": [],
    "freeze_parameters_regex": null,
    "freeze_parameters_ratio": 0.0,
    "trainable_parameters": [],
    "trainable_parameters_regex": null,
    "adapter_load": null,
    "target_modules": [
      "all-linear"
    ],
    "target_regex": null,
    "modules_to_save": [],
    "lora_rank": 8,
    "lora_alpha": 32,
    "lora_dropout": 0.05,
    "lora_bias": "none",
    "lora_dtype": null,
    "use_rslora": false,
    "rlhf_type": null,
    "ref_load": null,
    "ref_adapter_load": null,
    "beta": 0.1,
    "rpo_alpha": null,
    "reference_free": false,
    "label_smoothing": 0.0,
    "f_divergence_type": "reverse_kl",
    "loss_type": null,
    "desirable_weight": 1.0,
    "undesirable_weight": 1.0,
    "calculate_KL": null,
    "center_rewards_coefficient": null,
    "generation_batch_size": null,
    "steps_per_generation": null,
    "num_generations": 8,
    "max_completion_length": 512,
    "importance_sampling_level": "token",
    "tau_pos": 1.0,
    "tau_neg": 1.05,
    "epsilon": 0.2,
    "epsilon_high": null,
    "delta": null,
    "use_vllm": true,
    "vllm_mode": "colocate",
    "vllm_enable_prefix_caching": true,
    "vllm_gpu_memory_utilization": 0.9,
    "vllm_tensor_parallel_size": 1,
    "vllm_max_model_len": null,
    "vllm_enforce_eager": false,
    "vllm_limit_mm_per_prompt": null,
    "vllm_disable_cascade_attn": false,
    "sleep_level": 0,
    "offload_optimizer": false,
    "offload_model": false,
    "offload_bridge": false,
    "vllm_server_base_url": null,
    "vllm_server_host": null,
    "vllm_server_port": [
      8000
    ],
    "vllm_server_timeout": 240.0,
    "vllm_server_group_port": [
      51216
    ],
    "reward_funcs": [],
    "reward_weights": null,
    "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": null,
    "repetition_n_grams": 3,
    "repetition_max_penalty": -1.0,
    "soft_max_length": null,
    "soft_cache_length": null,
    "dynamic_sample": false,
    "max_resample_times": 3,
    "overlong_filter": false,
    "scale_rewards": "group",
    "advantage_estimator": "grpo",
    "kl_in_reward": false,
    "wandb_log_unique_prompts": null,
    "log_completions": false,
    "rollout_importance_sampling_mode": null,
    "rollout_importance_sampling_threshold": 2.0,
    "reward_model": null,
    "reward_model_plugin": null,
    "sync_ref_model": false,
    "ref_model_sync_steps": 512,
    "ref_model_mixup_alpha": 0.6,
    "async_generate": false,
    "move_model_batches": null,
    "multi_turn_scheduler": null,
    "max_turns": null,
    "completion_length_limit_scope": "per_round",
    "vllm_server_pass_dataset": false,
    "log_entropy": false,
    "top_entropy_quantile": 1.0,
    "num_iterations": 1,
    "padded_vocab_size": 151936,
    "initialize_embedding": false,
    "mlp_padding_free": false,
    "load_safetensors": true,
    "save_safetensors": true,
    "ref_model": null,
    "ref_adapters": [],
    "merge_lora": true,
    "max_shard_size": "5GB",
    "dataloader_persistent_workers": true,
    "dataloader_prefetch_factor": 10,
    "architectures": "Qwen3VLMoeForConditionalGeneration",
    "llm_architectures": "Qwen3VLMoeForConditionalGeneration",
    "max_epochs": 3,
    "enable_dft_loss": false,
    "enable_channel_loss": false,
    "save_strategy": "steps",
    "original_max_position_embeddings": null,
    "partial_rotary_factor": null,
    "use_shared_expert_gate": false,
    "vit_gradient_checkpointing": true,
    "vit_lr": null,
    "aligner_lr": null,
    "gradient_checkpointing_kwargs": null,
    "linear_num_value_heads": null,
    "linear_num_key_heads": null,
    "linear_key_head_dim": null,
    "linear_value_head_dim": null,
    "linear_conv_kernel_dim": null,
    "layer_types": null,
    "mrope_interleaved": true,
    "add_version": true
  }
}