{ "arch_specifier": "linear", "architectures": [ "PrismaticForConditionalGeneration" ], "auto_map": { "AutoConfig": "configuration_prismatic.PrismaticConfig", "AutoModelForVision2Seq": "modelling_pi.PrismaticForConditionalGeneration" }, "hf_llm_id": "agopalkr/gemma-2b", "image_resize_strategy": "resize-naive", "image_sizes": [ 224 ], "llm_backbone_id": "gemma-2b", "llm_max_length": 2048, "model_type": "prismatic", "output_projector_states": false, "pad_to_multiple_of": 64, "pad_token_id": 257153, "text_config": { "bos_token_id": 2, "eos_token_id": 1, "head_dim": 256, "hidden_act": "gelu_pytorch_tanh", "hidden_activation": null, "hidden_size": 2048, "intermediate_size": 24576, "max_position_embeddings": 8192, "model_type": "gemma", "num_attention_heads": 16, "num_hidden_layers": 28, "num_key_value_heads": 16, "pad_token_id": 257153, "tie_word_embeddings": true, "torch_dtype": "bfloat16", "vocab_size": 257216 }, "timm_model_ids": [ "vit_so400m_patch14_siglip_224" ], "timm_override_act_layers": [ null ], "torch_dtype": "bfloat16", "transformers_version": "4.44.2", "use_fused_vision_backbone": false, "vision_backbone_id": "siglip-vit-so400m" }