args.json

{
  "use_ray": false,
  "ray_exp_name": null,
  "device_groups": null,
  "model": "/mnt/models/basemodel/DeepSeek-V4-Flash",
  "model_type": "deepseek_v4",
  "model_revision": null,
  "task_type": "causal_lm",
  "torch_dtype": "bfloat16",
  "attn_impl": null,
  "experts_impl": "eager",
  "new_special_tokens": [],
  "num_labels": null,
  "problem_type": null,
  "rope_scaling": null,
  "device_map": null,
  "max_memory": {},
  "max_model_len": null,
  "local_repo_path": null,
  "init_strategy": null,
  "template": "deepseek_v4",
  "system": null,
  "max_length": 16384,
  "truncation_strategy": "delete",
  "max_pixels": null,
  "agent_template": "deepseek_v4",
  "norm_bbox": null,
  "use_chat_template": true,
  "padding_side": "right",
  "padding_free": false,
  "loss_scale": "default+ignore_empty_think",
  "sequence_parallel_size": 1,
  "is_binary_loss_scale": null,
  "template_backend": "swift",
  "response_prefix": null,
  "enable_thinking": null,
  "preserve_thinking": null,
  "add_non_thinking_prefix": true,
  "disable_ignore_empty_think": false,
  "dataset": [],
  "val_dataset": [],
  "cached_dataset": [],
  "cached_val_dataset": [],
  "split_dataset_ratio": 0.0,
  "data_seed": 42,
  "dataset_num_proc": 1,
  "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,
  "disable_auto_column_mapping": false,
  "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": 0.9,
  "top_k": -1,
  "top_p": 1.0,
  "repetition_penalty": 1.0,
  "num_beams": 1,
  "stream": false,
  "stop_words": [],
  "logprobs": false,
  "top_logprobs": null,
  "structured_outputs_regex": null,
  "tuner_backend": "peft",
  "tuner_type": "full",
  "adapters": [],
  "external_plugins": [],
  "custom_register_path": [],
  "seed": 42,
  "model_kwargs": {},
  "enable_npu_model_patch": true,
  "load_args": true,
  "load_data_args": false,
  "packing": false,
  "packing_length": null,
  "packing_num_proc": 1,
  "packing_strategy": "binpack",
  "lazy_tokenize": false,
  "use_hf": false,
  "hub_token": null,
  "ddp_timeout": 18000000,
  "ddp_backend": "nccl",
  "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,
  "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,
  "loss_type": null,
  "mcore_ref_model": null,
  "mcore_ref_adapter": null,
  "beta": null,
  "rpo_alpha": null,
  "reference_free": false,
  "label_smoothing": 0.0,
  "f_divergence_type": "reverse_kl",
  "desirable_weight": 1.0,
  "undesirable_weight": 1.0,
  "calculate_KL": null,
  "center_rewards_coefficient": null,
  "teacher_model": null,
  "teacher_model_type": null,
  "teacher_model_revision": null,
  "_teacher_use_disable_adapter": false,
  "teacher_model_server": null,
  "teacher_tag_key": "dataset",
  "gkd_logits_topk": null,
  "lmbda": 0.5,
  "seq_kd": false,
  "offload_teacher_model": false,
  "sft_alpha": 0.0,
  "teacher_kl_coef": 1.0,
  "generation_batch_size": null,
  "steps_per_generation": null,
  "num_generations": 8,
  "num_generations_eval": null,
  "max_completion_length": 512,
  "importance_sampling_level": "token",
  "tau_pos": 1.0,
  "tau_neg": 1.05,
  "real_tau": 0.5,
  "fipo_decay_rate": 32.0,
  "fipo_clip_range": 0.2,
  "fipo_clip_high_only": true,
  "fipo_safety_threshold": 4.0,
  "epsilon": 0.2,
  "epsilon_high": null,
  "delta": null,
  "use_vllm": true,
  "vllm_mode": null,
  "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,
  "vllm_max_num_seqs": null,
  "vllm_mm_processor_cache_gb": null,
  "vllm_engine_kwargs": null,
  "vllm_enable_lora": 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": null,
  "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,
  "log_rollout_offpolicy_metrics": false,
  "off_policy_sequence_mask_delta": null,
  "log_entropy": false,
  "top_entropy_quantile": 1.0,
  "router_replay_mode": "disabled",
  "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,
  "use_gym_env": null,
  "gym_env": null,
  "num_iterations": 1,
  "micro_batch_size": 1,
  "global_batch_size": 16,
  "recompute_granularity": null,
  "recompute_method": null,
  "recompute_num_layers": null,
  "recompute_modules": [
    "core_attn"
  ],
  "train_iters": null,
  "num_train_epochs": null,
  "masked_softmax_fusion": true,
  "bias_dropout_fusion": true,
  "bias_activation_fusion": true,
  "apply_rope_fusion": false,
  "gradient_accumulation_fusion": true,
  "cross_entropy_loss_fusion": true,
  "cross_entropy_fusion_impl": "native",
  "calculate_per_token_loss": true,
  "attention_backend": "<AttnBackend.flash: 1>",
  "optimizer": "adam",
  "optimizer_cpu_offload": false,
  "optimizer_offload_fraction": 1.0,
  "optimizer_cuda_graph": false,
  "use_precision_aware_optimizer": false,
  "main_grads_dtype": "float32",
  "main_params_dtype": "float32",
  "exp_avg_dtype": "float32",
  "exp_avg_sq_dtype": "float32",
  "manual_gc": false,
  "manual_gc_steps": 0,
  "manual_gc_eval": true,
  "train_dataloader_shuffle": true,
  "dataloader_num_workers": 4,
  "dataloader_pin_memory": true,
  "dataloader_persistent_workers": false,
  "dataloader_prefetch_factor": 2,
  "data_sharding": false,
  "group_by_length": false,
  "te_rng_tracker": false,
  "data_parallel_random_init": false,
  "mlp_padding_free": false,
  "lr_warmup_init": 0.0,
  "lr": 1e-05,
  "lr_decay_style": "cosine",
  "lr_decay_iters": null,
  "lr_warmup_iters": 0,
  "lr_warmup_fraction": null,
  "min_lr": 0,
  "lr_wsd_decay_style": "exponential",
  "lr_wsd_decay_iters": null,
  "weight_decay": 0.1,
  "weight_decay_incr_style": "constant",
  "start_weight_decay": 0.1,
  "end_weight_decay": 0.1,
  "clip_grad": 1.0,
  "adam_beta1": 0.9,
  "adam_beta2": 0.95,
  "adam_eps": 1e-08,
  "sgd_momentum": 0.9,
  "muon_momentum": 0.9,
  "muon_split_qkv": true,
  "muon_use_nesterov": false,
  "muon_scale_mode": "spectral",
  "muon_fp32_matmul_prec": "medium",
  "muon_coefficient_type": "quintic",
  "muon_num_ns_steps": 5,
  "muon_tp_mode": "blockwise",
  "muon_extra_scale_factor": 1.0,
  "muon_scalar_optimizer": "adam",
  "output_dir": "/mnt/data/output/megatron_8node_deepseek_v4_flash_v10_tau_style_v4_flashmla_16k_ppbal_mcoreckpt/v0-20260711-123643/checkpoint-5934-hf-safetensors",
  "save_steps": 500,
  "no_save_optim": true,
  "no_save_rng": true,
  "mcore_model": "/mnt/data/output/megatron_8node_deepseek_v4_flash_v10_tau_style_v4_flashmla_16k_ppbal_mcoreckpt/v0-20260711-123643/checkpoint-5934",
  "mcore_adapter": null,
  "no_load_optim": true,
  "no_load_rng": true,
  "finetune": true,
  "perform_initialization": false,
  "use_cpu_initialization": false,
  "async_save": false,
  "save_total_limit": null,
  "metric_for_best_model": "loss",
  "greater_is_better": false,
  "use_persistent_ckpt_worker": false,
  "dist_ckpt_save_pre_mcore_014": false,
  "dist_ckpt_optim_fully_reshardable": false,
  "distrib_optim_fully_reshardable_mem_efficient": false,
  "local_rank": 0,
  "use_distributed_optimizer": true,
  "tensor_model_parallel_size": 1,
  "pipeline_model_parallel_size": 8,
  "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,
  "overlap_p2p_comm": false,
  "batch_p2p_comm": true,
  "align_param_gather": false,
  "sequence_parallel": false,
  "context_parallel_size": 1,
  "cp_partition_mode": "zigzag",
  "sequence_packing_scheduler": null,
  "tp_comm_overlap": false,
  "overlap_grad_reduce": false,
  "overlap_param_gather": false,
  "overlap_param_gather_with_optimizer_step": false,
  "align_grad_reduce": true,
  "nccl_comm_warmup": false,
  "virtual_pipeline_model_parallel_size": null,
  "microbatch_group_size_per_vp_stage": null,
  "pipeline_model_parallel_layout": "Et*6|t*6|t*6|t*6|t*6|t*5|t*5|t*3,m,L",
  "expert_model_parallel_size": 8,
  "expert_tensor_parallel_size": 1,
  "report_to": [
    "tensorboard"
  ],
  "logging_steps": 5,
  "tensorboard_dir": null,
  "tensorboard_queue_size": 50,
  "wandb_project": "megatron-swift",
  "wandb_exp_name": null,
  "swanlab_project": "megatron-swift",
  "swanlab_exp_name": null,
  "eval_iters": -1,
  "eval_steps": 500,
  "fp8_format": "e4m3",
  "fp8_recipe": "blockwise",
  "fp8_param_gather": true,
  "fp8_amax_history_len": 1024,
  "fp8_amax_compute_algo": "max",
  "fp4_format": null,
  "fp4_recipe": "nvfp4",
  "fp4_param_gather": false,
  "fp16": false,
  "bf16": true,
  "apply_query_key_layer_scaling": false,
  "attention_softmax_in_fp32": true,
  "accumulate_allreduce_grads_in_fp32": true,
  "moe_router_load_balancing_type": null,
  "moe_router_dtype": "fp32",
  "moe_token_dispatcher_type": null,
  "moe_enable_deepep": false,
  "moe_grouped_gemm": true,
  "moe_permute_fusion": true,
  "moe_aux_loss_coeff": null,
  "moe_z_loss_coeff": null,
  "moe_shared_expert_overlap": false,
  "moe_layer_recompute": false,
  "moe_expert_capacity_factor": null,
  "moe_pad_expert_input_to_capacity": false,
  "moe_token_drop_policy": null,
  "mtp_num_layers": 1,
  "mtp_loss_scaling_factor": 0.1,
  "mtp_decoder_input_detach": false,
  "mtp_shared_weights": false,
  "save_safetensors": true,
  "ref_model": null,
  "ref_adapters": [],
  "merge_lora": true,
  "max_shard_size": "5GB",
  "vit_gradient_checkpointing": false,
  "vit_gradient_checkpointing_kwargs": null,
  "vit_attn_impl": null,
  "vit_lr": null,
  "aligner_lr": null,
  "dsa_indexer_loss_coeff": 0.0,
  "dsa_indexer_use_sparse_loss": false,
  "apply_dsa_kernel_fusion": true,
  "csa_dense_mode": false,
  "use_fused_mhc": false,
  "mhc_recompute_layer_num": null,
  "megatron_extra_kwargs": null,
  "language_model_only": false,
  "check_model": true,
  "apply_wd_to_qk_layernorm": false,
  "linear_decoupled_in_proj": false,
  "enable_dft_loss": false,
  "enable_channel_loss": false,
  "mrl_dims": null,
  "save_strategy": "steps",
  "callbacks": [
    "print",
    "default_flow",
    "tensorboard"
  ],
  "skip_megatron_init": false,
  "to_mcore": false,
  "to_hf": true,
  "test_convert_precision": false,
  "test_convert_dtype": "float32",
  "exist_ok": true,
  "swift_version": "4.5.0.dev0",
  "ckpt_dir": null,
  "rank": 0,
  "global_world_size": 64,
  "local_world_size": 8,
  "model_suffix": "DeepSeek-V4-Flash",
  "model_info": "ModelInfo(model_type='deepseek_v4', model_dir='/mnt/models/basemodel/DeepSeek-V4-Flash', torch_dtype=torch.bfloat16, max_model_len=1048576, quant_method='fp8', quant_bits=None, rope_scaling={'main': {'rope_type': 'default', 'rope_theta': 10000, 'partial_rotary_factor': 0.125}, 'compress': {'beta_fast': 32, 'beta_slow': 1, 'factor': 16, 'original_max_position_embeddings': 65536, 'type': 'yarn', 'rope_theta': 160000, 'rope_type': 'yarn', 'partial_rotary_factor': 0.125, 'attention_factor': 1.0}}, is_moe_model=True, is_multimodal=False, config=None, task_type='causal_lm', num_labels=None)",
  "model_meta": "ModelMeta(model_type='deepseek_v4', model_groups=[ModelGroup(models=[Model(ms_model_id='deepseek-ai/DeepSeek-V4-Flash', hf_model_id='deepseek-ai/DeepSeek-V4-Flash', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='deepseek-ai/DeepSeek-V4-Flash-Base', hf_model_id='deepseek-ai/DeepSeek-V4-Flash-Base', model_path=None, ms_revision=None, hf_revision=None)], template=None, ignore_patterns=None, requires=None, tags=[]), ModelGroup(models=[Model(ms_model_id='deepseek-ai/DeepSeek-V4-Pro', hf_model_id='deepseek-ai/DeepSeek-V4-Pro', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='deepseek-ai/DeepSeek-V4-Pro-Base', hf_model_id='deepseek-ai/DeepSeek-V4-Pro-Base', model_path=None, ms_revision=None, hf_revision=None)], template=None, ignore_patterns=None, requires=None, tags=[])], loader=<class 'swift.model.register.ModelLoader'>, template='deepseek_v4', model_arch=None, mcore_model_type=None, architectures=['DeepseekV4ForCausalLM'], additional_saved_files=[], torch_dtype=None, is_multimodal=False, is_reward=False, task_type=None, ignore_patterns=None, requires=[], tags=[])",
  "model_dir": "/mnt/models/basemodel/DeepSeek-V4-Flash",
  "template_meta": "DeepseekV2_5TemplateMeta(template_type='deepseek_v4', prefix=['<｜begin▁of▁sentence｜>'], prompt=['<｜User｜>{{QUERY}}<｜Assistant｜>'], chat_sep=['<｜end▁of▁sentence｜>'], suffix=['<｜end▁of▁sentence｜>'], template_cls=<class 'swift.template.templates.deepseek.DeepseekV4Template'>, system_prefix=['<｜begin▁of▁sentence｜>{{SYSTEM}}'], default_system=None, auto_add_bos=False, stop_words=[], agent_template='deepseek_v4', is_thinking=True, thinking_prefix='<think>', non_thinking_prefix='</think>', history_thinking_prefix='</think>')",
  "_val_dataset_exists": false,
  "hub": "<class 'swift.hub.hub.MSHub'>",
  "seq_length": 16384,
  "is_multimodal": false,
  "megatron_model_meta": "ModelMeta(model_type='deepseek_v4', model_types=['deepseek_v4'], bridge_cls=<class 'mcore_bridge.model.gpts.deepseek_v4.DeepseekV4Bridge'>, visual_cls=None, is_multimodal=False, loader=<class 'mcore_bridge.model.gpts.deepseek_v4.DeepseekV4Loader'>)",
  "fp8": "e4m3",
  "fp4": null,
  "world_size": 64,
  "data_parallel_size": 8,
  "num_microbatches": 2
}