Upload config.json with huggingface_hub
Browse files- config.json +64 -0
config.json
ADDED
|
@@ -0,0 +1,64 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"X": [
|
| 3 |
+
"VIDEO",
|
| 4 |
+
"IMAGE"
|
| 5 |
+
],
|
| 6 |
+
"_name_or_path": "./checkpoints/llava-v1.5-7b",
|
| 7 |
+
"architectures": [
|
| 8 |
+
"LlavaLlamaForCausalLMRMT"
|
| 9 |
+
],
|
| 10 |
+
"attention_bias": false,
|
| 11 |
+
"attention_dropout": 0.0,
|
| 12 |
+
"bos_token_id": 1,
|
| 13 |
+
"eos_token_id": 2,
|
| 14 |
+
"freeze_mm_mlp_adapter": false,
|
| 15 |
+
"freeze_mm_vision_resampler": false,
|
| 16 |
+
"hidden_act": "silu",
|
| 17 |
+
"hidden_size": 4096,
|
| 18 |
+
"image_aspect_ratio": "pad",
|
| 19 |
+
"initializer_range": 0.02,
|
| 20 |
+
"intermediate_size": 11008,
|
| 21 |
+
"max_length": 4096,
|
| 22 |
+
"max_position_embeddings": 4096,
|
| 23 |
+
"mm_attention_probs_dropout_prob": 0.1,
|
| 24 |
+
"mm_hidden_act": "gelu",
|
| 25 |
+
"mm_hidden_dropout_prob": 0.1,
|
| 26 |
+
"mm_hidden_size": 1024,
|
| 27 |
+
"mm_image_tower": "./checkpoints/LanguageBind_Image",
|
| 28 |
+
"mm_intermediate_size": 4096,
|
| 29 |
+
"mm_layer_norm_eps": 1e-12,
|
| 30 |
+
"mm_num_attention_heads": 8,
|
| 31 |
+
"mm_patch_merge_type": "flat",
|
| 32 |
+
"mm_projector_lr": null,
|
| 33 |
+
"mm_projector_type": "rmt_r_transformer1x",
|
| 34 |
+
"mm_resampler_type": null,
|
| 35 |
+
"mm_use_im_patch_token": false,
|
| 36 |
+
"mm_use_im_start_end": false,
|
| 37 |
+
"mm_use_x_patch_token": false,
|
| 38 |
+
"mm_use_x_start_end": false,
|
| 39 |
+
"mm_video_tower": "./checkpoints/LanguageBind_Video_merge",
|
| 40 |
+
"mm_vision_select_feature": "patch",
|
| 41 |
+
"mm_vision_select_layer": -2,
|
| 42 |
+
"mm_vision_tower": "openai/clip-vit-large-patch14-336",
|
| 43 |
+
"model_type": "llava_llama",
|
| 44 |
+
"num_attention_heads": 32,
|
| 45 |
+
"num_frames": 16,
|
| 46 |
+
"num_hidden_layers": 32,
|
| 47 |
+
"num_key_value_heads": 32,
|
| 48 |
+
"pad_token_id": 0,
|
| 49 |
+
"pretraining_tp": 1,
|
| 50 |
+
"rms_norm_eps": 1e-05,
|
| 51 |
+
"rope_scaling": null,
|
| 52 |
+
"rope_theta": 10000.0,
|
| 53 |
+
"tie_word_embeddings": false,
|
| 54 |
+
"tokenizer_model_max_length": 2048,
|
| 55 |
+
"tokenizer_padding_side": "right",
|
| 56 |
+
"torch_dtype": "bfloat16",
|
| 57 |
+
"transformers_version": "4.39.1",
|
| 58 |
+
"tune_mm_mlp_adapter": false,
|
| 59 |
+
"tune_mm_vision_resampler": false,
|
| 60 |
+
"unfreeze_mm_vision_tower": false,
|
| 61 |
+
"use_cache": true,
|
| 62 |
+
"use_mm_proj": true,
|
| 63 |
+
"vocab_size": 32000
|
| 64 |
+
}
|