| { |
| "schema_version": 2, |
| "phases": [ |
| { |
| "per_device_train_batch_size": 2, |
| "num_train_epochs": 1, |
| "max_steps": 480, |
| "learning_rate": 5e-06, |
| "lr_scheduler_type": "cosine", |
| "lr_scheduler_kwargs": null, |
| "warmup_steps": 0.05, |
| "optim": "paged_adamw_8bit", |
| "optim_args": null, |
| "weight_decay": 0.0, |
| "adam_beta1": 0.9, |
| "adam_beta2": 0.999, |
| "adam_epsilon": 1e-08, |
| "optim_target_modules": null, |
| "gradient_accumulation_steps": 4, |
| "average_tokens_across_devices": true, |
| "max_grad_norm": 1.0, |
| "label_smoothing_factor": 0.0, |
| "bf16": true, |
| "fp16": false, |
| "bf16_full_eval": false, |
| "fp16_full_eval": false, |
| "tf32": null, |
| "gradient_checkpointing": true, |
| "gradient_checkpointing_kwargs": null, |
| "torch_compile": false, |
| "torch_compile_backend": null, |
| "torch_compile_mode": null, |
| "use_liger_kernel": false, |
| "liger_kernel_config": null, |
| "use_cache": false, |
| "neftune_noise_alpha": null, |
| "torch_empty_cache_steps": null, |
| "auto_find_batch_size": false, |
| "logging_strategy": "steps", |
| "logging_steps": 5, |
| "logging_first_step": false, |
| "log_on_each_node": true, |
| "logging_nan_inf_filter": true, |
| "include_num_input_tokens_seen": "no", |
| "log_level": "passive", |
| "log_level_replica": "warning", |
| "disable_tqdm": false, |
| "report_to": [ |
| "tensorboard" |
| ], |
| "run_name": null, |
| "project": "huggingface", |
| "trackio_space_id": null, |
| "trackio_bucket_id": null, |
| "trackio_static_space_id": null, |
| "eval_strategy": "no", |
| "eval_steps": null, |
| "eval_delay": 0, |
| "per_device_eval_batch_size": 8, |
| "prediction_loss_only": false, |
| "eval_on_start": false, |
| "eval_do_concat_batches": true, |
| "eval_use_gather_object": false, |
| "eval_accumulation_steps": null, |
| "include_for_metrics": [], |
| "batch_eval_metrics": false, |
| "save_only_model": false, |
| "save_strategy": "steps", |
| "save_steps": 120, |
| "save_on_each_node": false, |
| "save_total_limit": 4, |
| "enable_jit_checkpoint": false, |
| "push_to_hub": false, |
| "hub_token": "<HUB_TOKEN>", |
| "hub_private_repo": null, |
| "hub_model_id": null, |
| "hub_strategy": "every_save", |
| "hub_always_push": false, |
| "hub_revision": null, |
| "load_best_model_at_end": false, |
| "metric_for_best_model": null, |
| "greater_is_better": null, |
| "ignore_data_skip": false, |
| "restore_callback_states_from_checkpoint": false, |
| "full_determinism": false, |
| "seed": 42, |
| "data_seed": null, |
| "use_cpu": false, |
| "accelerator_config": { |
| "split_batches": false, |
| "dispatch_batches": null, |
| "even_batches": true, |
| "use_seedable_sampler": true, |
| "non_blocking": false, |
| "gradient_accumulation_kwargs": null |
| }, |
| "parallelism_config": null, |
| "dataloader_drop_last": false, |
| "dataloader_num_workers": 0, |
| "dataloader_pin_memory": true, |
| "dataloader_persistent_workers": false, |
| "dataloader_prefetch_factor": null, |
| "remove_unused_columns": false, |
| "label_names": null, |
| "train_sampling_strategy": "random", |
| "length_column_name": "length", |
| "ddp_find_unused_parameters": null, |
| "ddp_bucket_cap_mb": null, |
| "ddp_broadcast_buffers": null, |
| "ddp_static_graph": null, |
| "ddp_backend": null, |
| "ddp_timeout": 1800, |
| "fsdp": null, |
| "fsdp_config": null, |
| "deepspeed": null, |
| "debug": [], |
| "skip_memory_metrics": true, |
| "do_train": false, |
| "do_eval": false, |
| "do_predict": false, |
| "resume_from_checkpoint": null, |
| "warmup_ratio": null, |
| "logging_dir": null, |
| "local_rank": -1, |
| "model_init_kwargs": { |
| "dtype": "bfloat16", |
| "revision": "1cfa9a7208912126459214e8b04321603b3df60c" |
| }, |
| "trust_remote_code": false, |
| "router_aux_loss_coef": 0.001, |
| "disable_dropout": false, |
| "cast_lm_head_to_fp32": false, |
| "num_generations": 8, |
| "num_generations_eval": null, |
| "max_completion_length": 3072, |
| "ds3_gather_for_generation": true, |
| "shuffle_dataset": true, |
| "pad_to_multiple_of": null, |
| "generation_batch_size": 8, |
| "steps_per_generation": 4, |
| "temperature": 0.6, |
| "top_p": 0.95, |
| "top_k": 20, |
| "min_p": null, |
| "generation_kwargs": null, |
| "chat_template_kwargs": { |
| "enable_thinking": true |
| }, |
| "repetition_penalty": 1.0, |
| "cache_implementation": null, |
| "use_vllm": true, |
| "vllm_mode": "colocate", |
| "vllm_model_impl": "vllm", |
| "vllm_enable_sleep_mode": false, |
| "vllm_structured_outputs_regex": null, |
| "vllm_server_base_url": null, |
| "vllm_server_host": "0.0.0.0", |
| "vllm_server_port": 8000, |
| "vllm_server_timeout": 240.0, |
| "vllm_group_port": 51216, |
| "vllm_gpu_memory_utilization": 0.25, |
| "vllm_max_model_length": null, |
| "vllm_tensor_parallel_size": 1, |
| "beta": 0.001, |
| "num_iterations": 1, |
| "epsilon": 0.2, |
| "delta": null, |
| "epsilon_high": null, |
| "sapo_temperature_neg": 1.05, |
| "sapo_temperature_pos": 1.0, |
| "vespo_k_pos": 2.0, |
| "vespo_lambda_pos": 3.0, |
| "vespo_k_neg": 3.0, |
| "vespo_lambda_neg": 2.0, |
| "importance_sampling_level": "token", |
| "reward_weights": [ |
| 0.3, |
| 0.2, |
| 0.5 |
| ], |
| "multi_objective_aggregation": "sum_then_normalize", |
| "scale_rewards": "group", |
| "loss_type": "grpo", |
| "mask_truncated_completions": true, |
| "sync_ref_model": false, |
| "ref_model_mixup_alpha": 0.6, |
| "ref_model_sync_steps": 512, |
| "top_entropy_quantile": 1.0, |
| "entropy_coef": 0.0, |
| "use_adaptive_entropy": false, |
| "entropy_coef_min": 0.0, |
| "entropy_coef_max": 1.0, |
| "entropy_coef_delta": 0.005, |
| "entropy_target": 0.2, |
| "max_tool_calling_iterations": null, |
| "vllm_importance_sampling_correction": true, |
| "vllm_importance_sampling_mode": "sequence_mask", |
| "vllm_importance_sampling_clip_max": 3.0, |
| "vllm_importance_sampling_clip_min": null, |
| "off_policy_mask_threshold": null, |
| "use_bias_correction_kl": false, |
| "log_completions": false, |
| "log_multimodal": true, |
| "num_completions_to_print": null, |
| "log_unique_prompts": false, |
| "log_completions_hub_repo": null, |
| "use_transformers_continuous_batching": false, |
| "transformers_continuous_batching_config": null, |
| "use_transformers_paged": false, |
| "vllm_importance_sampling_cap": null |
| }, |
| { |
| "per_device_train_batch_size": 2, |
| "num_train_epochs": 1, |
| "max_steps": 480, |
| "learning_rate": 5e-06, |
| "lr_scheduler_type": "cosine", |
| "lr_scheduler_kwargs": null, |
| "warmup_steps": 0.05, |
| "optim": "paged_adamw_8bit", |
| "optim_args": null, |
| "weight_decay": 0.0, |
| "adam_beta1": 0.9, |
| "adam_beta2": 0.999, |
| "adam_epsilon": 1e-08, |
| "optim_target_modules": null, |
| "gradient_accumulation_steps": 4, |
| "average_tokens_across_devices": true, |
| "max_grad_norm": 1.0, |
| "label_smoothing_factor": 0.0, |
| "bf16": true, |
| "fp16": false, |
| "bf16_full_eval": false, |
| "fp16_full_eval": false, |
| "tf32": null, |
| "gradient_checkpointing": true, |
| "gradient_checkpointing_kwargs": null, |
| "torch_compile": false, |
| "torch_compile_backend": null, |
| "torch_compile_mode": null, |
| "use_liger_kernel": false, |
| "liger_kernel_config": null, |
| "use_cache": false, |
| "neftune_noise_alpha": null, |
| "torch_empty_cache_steps": null, |
| "auto_find_batch_size": false, |
| "logging_strategy": "steps", |
| "logging_steps": 5, |
| "logging_first_step": false, |
| "log_on_each_node": true, |
| "logging_nan_inf_filter": true, |
| "include_num_input_tokens_seen": "no", |
| "log_level": "passive", |
| "log_level_replica": "warning", |
| "disable_tqdm": false, |
| "report_to": [ |
| "tensorboard" |
| ], |
| "run_name": null, |
| "project": "huggingface", |
| "trackio_space_id": null, |
| "trackio_bucket_id": null, |
| "trackio_static_space_id": null, |
| "eval_strategy": "no", |
| "eval_steps": null, |
| "eval_delay": 0, |
| "per_device_eval_batch_size": 8, |
| "prediction_loss_only": false, |
| "eval_on_start": false, |
| "eval_do_concat_batches": true, |
| "eval_use_gather_object": false, |
| "eval_accumulation_steps": null, |
| "include_for_metrics": [], |
| "batch_eval_metrics": false, |
| "save_only_model": false, |
| "save_strategy": "steps", |
| "save_steps": 120, |
| "save_on_each_node": false, |
| "save_total_limit": 4, |
| "enable_jit_checkpoint": false, |
| "push_to_hub": false, |
| "hub_token": "<HUB_TOKEN>", |
| "hub_private_repo": null, |
| "hub_model_id": null, |
| "hub_strategy": "every_save", |
| "hub_always_push": false, |
| "hub_revision": null, |
| "load_best_model_at_end": false, |
| "metric_for_best_model": null, |
| "greater_is_better": null, |
| "ignore_data_skip": false, |
| "restore_callback_states_from_checkpoint": false, |
| "full_determinism": false, |
| "seed": 42, |
| "data_seed": null, |
| "use_cpu": false, |
| "accelerator_config": { |
| "split_batches": false, |
| "dispatch_batches": null, |
| "even_batches": true, |
| "use_seedable_sampler": true, |
| "non_blocking": false, |
| "gradient_accumulation_kwargs": null |
| }, |
| "parallelism_config": null, |
| "dataloader_drop_last": false, |
| "dataloader_num_workers": 0, |
| "dataloader_pin_memory": true, |
| "dataloader_persistent_workers": false, |
| "dataloader_prefetch_factor": null, |
| "remove_unused_columns": false, |
| "label_names": null, |
| "train_sampling_strategy": "random", |
| "length_column_name": "length", |
| "ddp_find_unused_parameters": null, |
| "ddp_bucket_cap_mb": null, |
| "ddp_broadcast_buffers": null, |
| "ddp_static_graph": null, |
| "ddp_backend": null, |
| "ddp_timeout": 1800, |
| "fsdp": null, |
| "fsdp_config": null, |
| "deepspeed": null, |
| "debug": [], |
| "skip_memory_metrics": true, |
| "do_train": false, |
| "do_eval": false, |
| "do_predict": false, |
| "resume_from_checkpoint": null, |
| "warmup_ratio": null, |
| "logging_dir": null, |
| "local_rank": -1, |
| "model_init_kwargs": { |
| "dtype": "bfloat16", |
| "revision": "1cfa9a7208912126459214e8b04321603b3df60c" |
| }, |
| "trust_remote_code": false, |
| "router_aux_loss_coef": 0.001, |
| "disable_dropout": false, |
| "cast_lm_head_to_fp32": false, |
| "num_generations": 8, |
| "num_generations_eval": null, |
| "max_completion_length": 3072, |
| "ds3_gather_for_generation": true, |
| "shuffle_dataset": true, |
| "pad_to_multiple_of": null, |
| "generation_batch_size": 8, |
| "steps_per_generation": 4, |
| "temperature": 0.6, |
| "top_p": 0.95, |
| "top_k": 20, |
| "min_p": null, |
| "generation_kwargs": null, |
| "chat_template_kwargs": { |
| "enable_thinking": true |
| }, |
| "repetition_penalty": 1.0, |
| "cache_implementation": null, |
| "use_vllm": true, |
| "vllm_mode": "colocate", |
| "vllm_model_impl": "vllm", |
| "vllm_enable_sleep_mode": false, |
| "vllm_structured_outputs_regex": null, |
| "vllm_server_base_url": null, |
| "vllm_server_host": "0.0.0.0", |
| "vllm_server_port": 8000, |
| "vllm_server_timeout": 240.0, |
| "vllm_group_port": 51216, |
| "vllm_gpu_memory_utilization": 0.25, |
| "vllm_max_model_length": null, |
| "vllm_tensor_parallel_size": 1, |
| "beta": 0.001, |
| "num_iterations": 1, |
| "epsilon": 0.2, |
| "delta": null, |
| "epsilon_high": null, |
| "sapo_temperature_neg": 1.05, |
| "sapo_temperature_pos": 1.0, |
| "vespo_k_pos": 2.0, |
| "vespo_lambda_pos": 3.0, |
| "vespo_k_neg": 3.0, |
| "vespo_lambda_neg": 2.0, |
| "importance_sampling_level": "token", |
| "reward_weights": [ |
| 0.3, |
| 0.2, |
| 0.5 |
| ], |
| "multi_objective_aggregation": "sum_then_normalize", |
| "scale_rewards": "group", |
| "loss_type": "grpo", |
| "mask_truncated_completions": true, |
| "sync_ref_model": false, |
| "ref_model_mixup_alpha": 0.6, |
| "ref_model_sync_steps": 512, |
| "top_entropy_quantile": 1.0, |
| "entropy_coef": 0.0, |
| "use_adaptive_entropy": false, |
| "entropy_coef_min": 0.0, |
| "entropy_coef_max": 1.0, |
| "entropy_coef_delta": 0.005, |
| "entropy_target": 0.2, |
| "max_tool_calling_iterations": null, |
| "vllm_importance_sampling_correction": true, |
| "vllm_importance_sampling_mode": "sequence_mask", |
| "vllm_importance_sampling_clip_max": 3.0, |
| "vllm_importance_sampling_clip_min": null, |
| "off_policy_mask_threshold": null, |
| "use_bias_correction_kl": false, |
| "log_completions": false, |
| "log_multimodal": true, |
| "num_completions_to_print": null, |
| "log_unique_prompts": false, |
| "log_completions_hub_repo": null, |
| "use_transformers_continuous_batching": false, |
| "transformers_continuous_batching_config": null, |
| "use_transformers_paged": false, |
| "vllm_importance_sampling_cap": null |
| } |
| ] |
| } |
|
|
