ManishThota/CustomModel
1160
1{2 "_name_or_path": "ManishThota/Sparrow",3 "activation_function": "gelu_new",4 "architectures": [5 "ImpForCausalLM"6 ],7 "attn_pdrop": 0.0,8 "auto_map": {9 "AutoConfig": "configuration_imp.ImpConfig",10 "AutoModelForCausalLM": "modeling_imp.ImpForCausalLM"11 },12 "embd_pdrop": 0.0,13 "eos_token_id": 50295,14 "flash_attn": false,15 "flash_rotary": false,16 "freeze_mm_mlp_adapter": false,17 "fused_dense": false,18 "image_aspect_ratio": "square",19 "image_token": "<image>",20 "image_token_index": 50296,21 "img_processor": null,22 "initializer_range": 0.02,23 "layer_norm_epsilon": 1e-05,24 "mm_hidden_size": 1152,25 "mm_projector_lr": 2e-05,26 "mm_projector_type": "mlp2x_gelu",27 "mm_use_im_patch_token": false,28 "mm_use_im_start_end": false,29 "mm_vision_select_feature": "patch",30 "mm_vision_select_layer": -2,31 "mm_vision_tower": "google/siglip-so400m-patch14-384",32 "model_type": "Sparrow",33 "n_embd": 2560,34 "n_head": 32,35 "n_head_kv": null,36 "n_inner": null,37 "n_layer": 32,38 "n_positions": 3072,39 "pad_token_id": 50256,40 "resid_pdrop": 0.1,41 "rotary_dim": 32,42 "tie_word_embeddings": false,43 "tokenizer_model_max_length": 3072,44 "tokenizer_padding_side": "right",45 "torch_dtype": "float16",46 "transformers_version": "4.31.0",47 "tune_mm_mlp_adapter": false,48 "use_cache": true,49 "use_mm_proj": true,50 "vision_tower_config": {51 "attention_dropout": 0.0,52 "hidden_act": "gelu_pytorch_tanh",53 "hidden_size": 1152,54 "image_size": 384,55 "intermediate_size": 4304,56 "layer_norm_eps": 1e-06,57 "model_type": "siglip_vision_model",58 "num_attention_heads": 16,59 "num_channels": 3,60 "num_hidden_layers": 27,61 "patch_size": 1462 },63 "vocab_size": 5120064}65 