!!python/object:gr00t.configs.base_config.Config data: !!python/object:gr00t.configs.data.data_config.DataConfig allow_padding: false datasets: - !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig dataset_paths: - /workspace/Isaac-GR00T/examples/so101/merged_v2_out/thinhpt/merged_data_29042026 dataset_type: physical_embodiment embodiment_tag: new_embodiment mix_ratio: 1.0 val_dataset_path: null download_cache: false episode_sampling_rate: 0.1 image_crop_size: - 244 - 244 image_target_size: - 224 - 224 mock_dataset_mode: false modality_configs: new_embodiment: action: !!python/object:gr00t.data.types.ModalityConfig action_configs: - !!python/object:gr00t.data.types.ActionConfig format: &id001 !!python/object/apply:gr00t.data.types.ActionFormat - default rep: !!python/object/apply:gr00t.data.types.ActionRepresentation - relative state_key: null type: &id002 !!python/object/apply:gr00t.data.types.ActionType - non_eef - !!python/object:gr00t.data.types.ActionConfig format: *id001 rep: !!python/object/apply:gr00t.data.types.ActionRepresentation - absolute state_key: null type: *id002 delta_indices: - 0 - 1 - 2 - 3 - 4 - 5 - 6 - 7 - 8 - 9 - 10 - 11 - 12 - 13 - 14 - 15 mean_std_embedding_keys: null modality_keys: - single_arm - gripper sin_cos_embedding_keys: null language: !!python/object:gr00t.data.types.ModalityConfig action_configs: null delta_indices: - 0 mean_std_embedding_keys: null modality_keys: - annotation.human.task_description sin_cos_embedding_keys: null state: !!python/object:gr00t.data.types.ModalityConfig action_configs: null delta_indices: - 0 mean_std_embedding_keys: null modality_keys: - single_arm - gripper sin_cos_embedding_keys: null video: !!python/object:gr00t.data.types.ModalityConfig action_configs: null delta_indices: - 0 mean_std_embedding_keys: null modality_keys: - top - side - wrist sin_cos_embedding_keys: null mode: single_turn multiprocessing_context: fork num_shards_per_epoch: 100000 override_pretraining_statistics: true random_chop: 0.0 seed: 42 shard_size: 1024 shuffle: true subsample_ratio: 1.0 video_backend: torchcodec load_config_path: null model: !!python/object:gr00t.configs.model.gr00t_n1d7.Gr00tN1d7Config _attn_implementation_internal: null _commit_hash: null _name_or_path: '' _output_attentions: false add_cross_attention: false architectures: null backbone_trainable_params_fp32: true bad_words_ids: null begin_suppress_tokens: null bos_token_id: null chunk_size_feed_forward: 0 color_jitter_params: brightness: 0.3 contrast: 0.4 hue: 0.08 saturation: 0.5 cross_attention_hidden_size: null decoder_start_token_id: null diffusion_model_cfg: attention_head_dim: 48 dropout: 0.2 final_dropout: true interleave_self_attention: true norm_type: ada_norm num_attention_heads: 32 num_layers: 16 output_dim: 1024 positional_embeddings: null diversity_penalty: 0.0 do_sample: false dtype: null early_stopping: false encoder_no_repeat_ngram_size: 0 eos_token_id: null exponential_decay_length_penalty: null extra_augmentation_config: null finetuning_task: null forced_bos_token_id: null forced_eos_token_id: null id2label: 0: LABEL_0 1: LABEL_1 is_decoder: false is_encoder_decoder: false label2id: LABEL_0: 0 LABEL_1: 1 length_penalty: 1.0 load_bf16: false max_length: 20 min_length: 0 model_name: nvidia/Cosmos-Reason2-2B no_repeat_ngram_size: 0 num_beam_groups: 1 num_beams: 1 num_return_sequences: 1 output_hidden_states: false output_scores: false pad_token_id: null prefix: null problem_type: null pruned_heads: {} random_rotation_angle: null remove_invalid_values: false repetition_penalty: 1.0 reproject_vision: false return_dict: true return_dict_in_generate: false sep_token_id: null state_dropout_prob: 0.2 suppress_tokens: null task_specific_params: null temperature: 1.0 tf_legacy_loss: false tie_encoder_decoder: false tie_word_embeddings: true tokenizer_class: null top_k: 50 top_p: 1.0 torchscript: false transformers_version: null tune_diffusion_model: true tune_llm: false tune_projector: true tune_visual: false typical_p: 1.0 use_bfloat16: false use_relative_action: true training: !!python/object:gr00t.configs.training.training_config.TrainingConfig add_rl_callback: false assert_loss_less_than: null batch_size: null bf16: true dataloader_num_workers: 4 ddp_bucket_cap_mb: 100 deepspeed_stage: 2 enable_open_loop_eval: false enable_profiling: false eval_batch_size: 2 eval_bf16: true eval_set_split_ratio: 0.1 eval_steps: 500 eval_strategy: 'no' experiment_name: null fp16: false global_batch_size: 1 gradient_accumulation_steps: 1 gradient_checkpointing: false learning_rate: 0.0001 logging_steps: 10 lr_scheduler_type: cosine max_concurrent_uploads: 2 max_grad_norm: 1.0 max_retries: 3 max_steps: 1 num_gpus: 1 open_loop_eval_plot_indices: null open_loop_eval_steps_per_traj: 100 open_loop_eval_traj_ids: - 0 optim: adamw_torch output_dir: /workspace/Isaac-GR00T/outputs/so101_from_a100 remove_unused_columns: false save_best_eval_metric_greater_is_better: true save_best_eval_metric_name: '' save_only_model: false save_steps: 1 save_total_limit: 5 save_vl_model: false skip_weight_loading: false start_from_checkpoint: nvidia/GR00T-N1.7-3B tf32: true transformers_access_token: null transformers_cache_dir: null transformers_local_files_only: false transformers_trust_remote_code: true upload_checkpoints: false upload_every: 1000 upload_last_n_checkpoints: 5 use_ddp: false use_wandb: false wandb_project: finetune-gr00t-n1d7 warmup_ratio: 0.05 warmup_steps: 0 weight_decay: 1.0e-05