trl-internal-testing/tiny-Qwen2_5_VLForConditionalGeneration
0319k
1{2 "architectures": [3 "Qwen2_5_VLForConditionalGeneration"4 ],5 "attention_dropout": 0.0,6 "bos_token_id": 151643,7 "dtype": "bfloat16",8 "eos_token_id": 151645,9 "hidden_act": "silu",10 "hidden_size": 16,11 "image_token_id": 151655,12 "initializer_range": 0.02,13 "intermediate_size": 11008,14 "max_position_embeddings": 128000,15 "max_window_layers": 70,16 "model_type": "qwen2_5_vl",17 "num_attention_heads": 4,18 "num_hidden_layers": 2,19 "num_key_value_heads": 2,20 "rms_norm_eps": 1e-06,21 "rope_scaling": {22 "mrope_section": [23 1,24 1,25 026 ],27 "rope_type": "default",28 "type": "default"29 },30 "rope_theta": 1000000.0,31 "sliding_window": 32768,32 "text_config": {33 "architectures": [34 "Qwen2_5_VLForConditionalGeneration"35 ],36 "attention_dropout": 0.0,37 "bos_token_id": 151643,38 "dtype": "bfloat16",39 "eos_token_id": 151645,40 "hidden_act": "silu",41 "hidden_size": 16,42 "image_token_id": null,43 "initializer_range": 0.02,44 "intermediate_size": 11008,45 "layer_types": [46 "full_attention",47 "full_attention"48 ],49 "max_position_embeddings": 128000,50 "max_window_layers": 70,51 "model_type": "qwen2_5_vl_text",52 "num_attention_heads": 4,53 "num_hidden_layers": 2,54 "num_key_value_heads": 2,55 "rms_norm_eps": 1e-06,56 "rope_scaling": {57 "mrope_section": [58 1,59 1,60 061 ],62 "rope_type": "default",63 "type": "default"64 },65 "rope_theta": 1000000.0,66 "sliding_window": null,67 "tie_word_embeddings": true,68 "use_cache": true,69 "use_sliding_window": false,70 "video_token_id": null,71 "vision_end_token_id": 151653,72 "vision_start_token_id": 151652,73 "vision_token_id": 151654,74 "vocab_size": 15193675 },76 "transformers_version": "4.56.2",77 "use_cache": true,78 "use_sliding_window": false,79 "video_token_id": 151656,80 "vision_config": {81 "depth": 2,82 "fullatt_block_indexes": [83 184 ],85 "hidden_act": "silu",86 "hidden_size": 16,87 "in_channels": 3,88 "in_chans": 3,89 "initializer_range": 0.02,90 "intermediate_size": 3420,91 "model_type": "qwen2_5_vl",92 "num_heads": 4,93 "out_hidden_size": 16,94 "patch_size": 14,95 "spatial_merge_size": 2,96 "spatial_patch_size": 14,97 "temporal_patch_size": 2,98 "tokens_per_second": 2,99 "window_size": 112100 },101 "vision_end_token_id": 151653,102 "vision_start_token_id": 151652,103 "vision_token_id": 151654,104 "vocab_size": 151936105}106 