Upload config (2).json
#4
by
thanhnhan29
- opened
- config (2).json +72 -0
config (2).json
ADDED
|
@@ -0,0 +1,72 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "/tmp/iopath_cache/manifold_cache/tree/users/shenx/finetune/10162317-cambrian_llama3_2-1b_t576_ov",
|
| 3 |
+
"architectures": ["CambrianLlamaForCausalLM"],
|
| 4 |
+
"attention_bias": false,
|
| 5 |
+
"attention_dropout": 0.0,
|
| 6 |
+
"bos_token_id": 128000,
|
| 7 |
+
"connect_layer": 2,
|
| 8 |
+
"connector_depth": 3,
|
| 9 |
+
"connector_only": true,
|
| 10 |
+
"dino_threshold": 0.83,
|
| 11 |
+
"drop_threshold": 0.8,
|
| 12 |
+
"eos_token_id": [128001, 128008, 128009],
|
| 13 |
+
"frame_pos": false,
|
| 14 |
+
"freeze_mm_mlp_adapter": false,
|
| 15 |
+
"head_dim": 64,
|
| 16 |
+
"hidden_act": "silu",
|
| 17 |
+
"hidden_size": 2048,
|
| 18 |
+
"highres": true,
|
| 19 |
+
"highres_connect": false,
|
| 20 |
+
"image_aspect_ratio": "pad",
|
| 21 |
+
"image_position": 91,
|
| 22 |
+
"image_token_len": 144,
|
| 23 |
+
"initializer_range": 0.02,
|
| 24 |
+
"intermediate_size": 8192,
|
| 25 |
+
"is_image_newline": true,
|
| 26 |
+
"is_st_sampler": false,
|
| 27 |
+
"lowres_token": 8,
|
| 28 |
+
"max_position_embeddings": 131072,
|
| 29 |
+
"mlp_bias": false,
|
| 30 |
+
"mm_patch_merge_type": "flat",
|
| 31 |
+
"mm_projector_lr": null,
|
| 32 |
+
"mm_projector_type": "sva",
|
| 33 |
+
"mm_use_im_patch_token": false,
|
| 34 |
+
"mm_use_im_start_end": false,
|
| 35 |
+
"mm_vision_sampler_lr": null,
|
| 36 |
+
"mm_vision_select_feature": "patch",
|
| 37 |
+
"mm_vision_select_layer": -2,
|
| 38 |
+
"mm_vision_tower_aux_list": [
|
| 39 |
+
"siglip/CLIP-ViT-SO400M-14-384",
|
| 40 |
+
"facebook/dinov2-giant-res378"
|
| 41 |
+
],
|
| 42 |
+
"mm_vision_tower_aux_token_len_list": [576, 576],
|
| 43 |
+
"mm_vision_tower_lr": null,
|
| 44 |
+
"model_type": "cambrian_llama",
|
| 45 |
+
"num_attention_heads": 32,
|
| 46 |
+
"num_hidden_layers": 16,
|
| 47 |
+
"num_key_value_heads": 8,
|
| 48 |
+
"num_of_vision_sampler_layers": 10,
|
| 49 |
+
"num_query_group": 1,
|
| 50 |
+
"pretraining_tp": 1,
|
| 51 |
+
"query_num_list": [144],
|
| 52 |
+
"rms_norm_eps": 1e-5,
|
| 53 |
+
"rope_scaling": null,
|
| 54 |
+
"rope_theta": 500000.0,
|
| 55 |
+
"spmd_debug": null,
|
| 56 |
+
"spmd_fsdp_sharding": null,
|
| 57 |
+
"spmd_mesh": null,
|
| 58 |
+
"start_of_vision_sampler_layers": 0,
|
| 59 |
+
"stride_of_vision_sampler_layers": 3,
|
| 60 |
+
"tie_word_embeddings": false,
|
| 61 |
+
"tokenizer_model_max_length": 16000,
|
| 62 |
+
"tokenizer_padding_side": "right",
|
| 63 |
+
"torch_dtype": "float32",
|
| 64 |
+
"transformers_version": "4.44.2",
|
| 65 |
+
"tune_mm_mlp_adapter": false,
|
| 66 |
+
"unfreeze_mm_vision_tower": false,
|
| 67 |
+
"use_cache": false,
|
| 68 |
+
"use_mm_proj": true,
|
| 69 |
+
"vision_hidden_size": 1024,
|
| 70 |
+
"vision_tower_aux_token_len_list": [576, 576],
|
| 71 |
+
"vocab_size": 128256
|
| 72 |
+
}
|