EmbodiedReasoningAgent/EPL-RL-Model_EB-Manipulation

1{2  "_name_or_path": "/u/ry21/scratch/rl_train_third_folder/ebman-grounding_3/grounding+2epoch-traj-with-reason-rl_mini48-process2outcome_complete_randomized42_data_with_shuffled_sampler-20250814_034739/global_step_48/actor/huggingface",3  "architectures": [4    "Qwen2_5_VLForConditionalGeneration"5  ],6  "attention_dropout": 0.0,7  "bos_token_id": 151643,8  "eos_token_id": [9    15166510  ],11  "hidden_act": "silu",12  "hidden_size": 2048,13  "image_token_id": 151655,14  "initializer_range": 0.02,15  "intermediate_size": 11008,16  "max_position_embeddings": 128000,17  "max_window_layers": 70,18  "model_type": "qwen2_5_vl",19  "num_attention_heads": 16,20  "num_hidden_layers": 36,21  "num_key_value_heads": 2,22  "rms_norm_eps": 1e-06,23  "rope_scaling": {24    "mrope_section": [25      16,26      24,27      2428    ],29    "rope_type": "default",30    "type": "default"31  },32  "rope_theta": 1000000.0,33  "sliding_window": 32768,34  "tie_word_embeddings": true,35  "torch_dtype": "bfloat16",36  "transformers_version": "4.50.3",37  "use_cache": true,38  "use_sliding_window": false,39  "video_token_id": 151656,40  "vision_config": {41    "depth": 32,42    "fullatt_block_indexes": [43      7,44      15,45      23,46      3147    ],48    "hidden_act": "silu",49    "hidden_size": 1280,50    "in_channels": 3,51    "in_chans": 3,52    "intermediate_size": 3420,53    "model_type": "qwen2_5_vl",54    "num_heads": 16,55    "out_hidden_size": 2048,56    "patch_size": 14,57    "spatial_merge_size": 2,58    "spatial_patch_size": 14,59    "temporal_patch_size": 2,60    "tokens_per_second": 2,61    "torch_dtype": "bfloat16",62    "window_size": 11263  },64  "vision_end_token_id": 151653,65  "vision_start_token_id": 151652,66  "vision_token_id": 151654,67  "vocab_size": 15167068}69