ByteDance-Seed/M3-Agent-Memorization
16127
1{2 "architectures": [3 "Qwen2_5OmniThinkerForConditionalGeneration"4 ],5 "audio_config": {6 "_attn_implementation_autoset": true,7 "activation_dropout": 0.0,8 "activation_function": "gelu",9 "attention_dropout": 0.0,10 "d_model": 1280,11 "dropout": 0.0,12 "encoder_attention_heads": 20,13 "encoder_ffn_dim": 5120,14 "encoder_layerdrop": 0.0,15 "encoder_layers": 32,16 "init_std": 0.02,17 "max_source_positions": 1500,18 "model_type": "qwen2_5_omni_audio_encoder",19 "n_window": 100,20 "num_hidden_layers": 32,21 "num_mel_bins": 128,22 "output_dim": 3584,23 "scale_embedding": false24 },25 "audio_end_token_id": 151648,26 "audio_start_token_id": 151647,27 "audio_token_index": 151646,28 "bos_token_id": 151644,29 "enable_audio_output": true,30 "enable_talker": true,31 "eos_token_id": 151645,32 "ignore_index": -100,33 "image_token_index": 151655,34 "init_std": 0.02,35 "model_type": "qwen2_5_omni_thinker",36 "pad_token_id": 151643,37 "position_id_per_seconds": 25,38 "seconds_per_chunk": 2,39 "text_config": {40 "attention_dropout": 0.0,41 "hidden_act": "silu",42 "hidden_size": 3584,43 "init_std": 0.02,44 "intermediate_size": 18944,45 "max_position_embeddings": 32768,46 "max_window_layers": 28,47 "model_type": "qwen2_5_omni_text",48 "num_attention_heads": 28,49 "num_hidden_layers": 28,50 "num_key_value_heads": 4,51 "rms_norm_eps": 1e-06,52 "rope_scaling": {53 "mrope_section": [54 16,55 24,56 2457 ],58 "rope_type": "default",59 "type": "default"60 },61 "rope_theta": 1000000.0,62 "sliding_window": 32768,63 "use_cache": true,64 "use_sliding_window": false,65 "vocab_size": 15206466 },67 "tie_word_embeddings": false,68 "torch_dtype": "bfloat16",69 "transformers_version": "4.50.0.dev0",70 "use_mrope": false,71 "use_reentrant": false,72 "user_token_id": 872,73 "video_token_index": 151656,74 "vision_config": {75 "_attn_implementation_autoset": true,76 "depth": 32,77 "embed_dim": 1280,78 "fullatt_block_indexes": [79 7,80 15,81 23,82 3183 ],84 "hidden_act": "silu",85 "hidden_size": 1280,86 "in_channels": 3,87 "in_chans": 3,88 "init_std": 0.02,89 "intermediate_size": 3420,90 "model_type": "qwen2_5_omni_vision_encoder",91 "num_heads": 16,92 "out_hidden_size": 3584,93 "patch_size": 14,94 "spatial_merge_size": 2,95 "spatial_patch_size": 14,96 "temporal_patch_size": 2,97 "tokens_per_second": 25,98 "window_size": 11299 },100 "vision_end_token_id": 151653,101 "vision_start_token_id": 151652,102 "vision_token_id": 151654103}104 