{ "architectures": [ "Lfm3VlForConditionalGeneration" ], "auto_map": { "AutoConfig": "configuration_lfm3_vl.Lfm3VlConfig", "AutoModelForCausalLM": "modeling_lfm3_vl.Lfm3VlForConditionalGeneration", "AutoModelForImageTextToText": "modeling_lfm3_vl.Lfm3VlForConditionalGeneration" }, "bos_token_id": 124894, "conv_L_cache": 7, "conv_bias": false, "downsample_factor": 2, "dtype": "float32", "eos_token_id": 124900, "full_attention_rope_theta": 1000000.0, "head_dim": 8, "hidden_size": 16, "image_token_id": 125030, "intermediate_size": 32, "layer_types": [ "conv", "conv", "full_attention", "conv", "sliding_attention", "conv", "sliding_attention" ], "max_position_embeddings": 128000, "model_type": "lfm3_vl", "num_attention_heads": 4, "num_hidden_layers": 7, "num_key_value_heads": 2, "pad_token_id": 124893, "partial_rotary_factor": 0.5, "projector_hidden_size": 32, "projector_layer_norm_eps": 1e-06, "rms_norm_eps": 1e-05, "sliding_attention_rope_theta": 10000, "sliding_window": 4, "tie_word_embeddings": true, "transformers_version": "5.20.0", "use_cache": true, "use_sandwich_norm": true, "video_token_id": 125029, "vision_end_token_id": 125010, "vision_hidden_size": 16, "vision_intermediate_size": 32, "vision_layer_types": [ "conv", "full_attention", "conv", "full_attention" ], "vision_max_position_embeddings": 2304, "vision_num_attention_heads": 2, "vision_num_channels": 3, "vision_num_hidden_layers": 4, "vision_num_key_value_heads": 1, "vision_patch_dim": 1536, "vision_patch_size": 16, "vision_rope_theta": 1000000.0, "vision_spatial_merge_size": 2, "vision_start_token_id": 125009, "vision_temporal_patch_size": 2, "vocab_size": 128000 }