TIGER-Lab/VLM2Vec-LoRA

1{2  "_name_or_path": "microsoft/Phi-3.5-vision-instruct",3  "architectures": [4    "Phi3VForCausalLM"5  ],6  "attention_dropout": 0.0,7  "auto_map": {8    "AutoConfig": "microsoft/Phi-3.5-vision-instruct--configuration_phi3_v.Phi3VConfig",9    "AutoModelForCausalLM": "microsoft/Phi-3.5-vision-instruct--modeling_phi3_v.Phi3VForCausalLM"10  },11  "bos_token_id": 1,12  "embd_layer": {13    "embedding_cls": "image",14    "hd_transform_order": "sub_glb",15    "projection_cls": "mlp",16    "use_hd_transform": true,17    "with_learnable_separator": true18  },19  "embd_pdrop": 0.0,20  "eos_token_id": 2,21  "hidden_act": "silu",22  "hidden_size": 3072,23  "img_processor": {24    "image_dim_out": 1024,25    "model_name": "openai/clip-vit-large-patch14-336",26    "name": "clip_vision_model",27    "num_img_tokens": 14428  },29  "initializer_range": 0.02,30  "intermediate_size": 8192,31  "max_position_embeddings": 131072,32  "model_type": "phi3_v",33  "num_attention_heads": 32,34  "num_hidden_layers": 32,35  "num_key_value_heads": 32,36  "original_max_position_embeddings": 4096,37  "pad_token_id": 32000,38  "padding_side": "right",39  "resid_pdrop": 0.0,40  "rms_norm_eps": 1e-05,41  "rope_scaling": {42    "long_factor": [43      1.0800000429153442,44      1.1100000143051147,45      1.1399999856948853,46      1.340000033378601,47      1.5899999141693115,48      1.600000023841858,49      1.6200000047683716,50      2.620000123977661,51      3.2300000190734863,52      3.2300000190734863,53      4.789999961853027,54      7.400000095367432,55      7.700000286102295,56      9.09000015258789,57      12.199999809265137,58      17.670000076293945,59      24.46000099182129,60      28.57000160217285,61      30.420001983642578,62      30.840002059936523,63      32.590003967285156,64      32.93000411987305,65      42.320003509521484,66      44.96000289916992,67      50.340003967285156,68      50.45000457763672,69      57.55000305175781,70      57.93000411987305,71      58.21000289916992,72      60.1400032043457,73      62.61000442504883,74      62.62000274658203,75      62.71000289916992,76      63.1400032043457,77      63.1400032043457,78      63.77000427246094,79      63.93000411987305,80      63.96000289916992,81      63.970001220703125,82      64.02999877929688,83      64.06999969482422,84      64.08000183105469,85      64.12000274658203,86      64.41000366210938,87      64.4800033569336,88      64.51000213623047,89      64.52999877929688,90      64.8399963378906291    ],92    "short_factor": [93      1.08,94      1.1,95      1.1300000000000001,96      1.2800000000000002,97      1.3100000000000003,98      1.4500000000000004,99      1.4500000000000004,100      1.9500000000000008,101      2.030000000000001,102      2.4299999999999926,103      2.5699999999999896,104      2.9499999999999815,105      3.729999999999965,106      3.869999999999962,107      4.189999999999955,108      4.43999999999995,109      4.6399999999999455,110      4.979999999999938,111      5.159999999999934,112      5.279999999999932,113      5.759999999999922,114      5.889999999999919,115      5.889999999999919,116      5.969999999999917,117      6.089999999999915,118      6.2799999999999105,119      6.7699999999999,120      6.8899999999998975,121      7.109999999999893,122      7.129999999999892,123      7.179999999999891,124      7.289999999999889,125      7.339999999999888,126      7.559999999999883,127      7.619999999999882,128      7.69999999999988,129      7.879999999999876,130      7.879999999999876,131      7.879999999999876,132      7.939999999999875,133      7.949999999999875,134      7.979999999999874,135      8.19999999999987,136      8.439999999999864,137      8.469999999999864,138      8.589999999999861,139      8.809999999999857,140      8.999999999999853141    ],142    "type": "su"143  },144  "rope_theta": 10000.0,145  "sliding_window": 262144,146  "tie_word_embeddings": false,147  "torch_dtype": "bfloat16",148  "transformers_version": "4.45.1",149  "use_cache": false,150  "vocab_size": 32064151}152