Upload config.json with huggingface_hub
Browse files- config.json +40 -0
config.json
ADDED
|
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "/home/dmgcsilva/project/PROJECT/dmgcsilva/followup/experiments/vista_stage46/llama8binst_siglip400m_mlp_ln_patches_stage46/checkpoint_3252",
|
| 3 |
+
"architectures": [
|
| 4 |
+
"ReVista"
|
| 5 |
+
],
|
| 6 |
+
"cap_loss_scale": 1.0,
|
| 7 |
+
"connector_type": "mlp",
|
| 8 |
+
"end_ret_token_id": 128258,
|
| 9 |
+
"eos_token_id": 128009,
|
| 10 |
+
"freeze_cap": false,
|
| 11 |
+
"freeze_emb": false,
|
| 12 |
+
"freeze_lm": false,
|
| 13 |
+
"freeze_ret": false,
|
| 14 |
+
"freeze_vm": true,
|
| 15 |
+
"image_embed_dropout_prob": 0.0,
|
| 16 |
+
"img_token_id": 128256,
|
| 17 |
+
"lm_lora_alpha": 128,
|
| 18 |
+
"lm_lora_dropout": 0.03,
|
| 19 |
+
"lm_lora_rank": 32,
|
| 20 |
+
"model_type": "revistamodel",
|
| 21 |
+
"n_visual_tokens": 1,
|
| 22 |
+
"negative_count": 512,
|
| 23 |
+
"pad_token_id": 128009,
|
| 24 |
+
"projector_type": "linear",
|
| 25 |
+
"ret_loss_scale": 1.0,
|
| 26 |
+
"ret_token_id": 128257,
|
| 27 |
+
"shared_emb_dim": 512,
|
| 28 |
+
"start_ret_token_id": 128257,
|
| 29 |
+
"task": "captioning",
|
| 30 |
+
"text_decoder": "meta-llama/Llama-3.1-8B-Instruct",
|
| 31 |
+
"text_embed_dropout_prob": 0.1,
|
| 32 |
+
"torch_dtype": "float32",
|
| 33 |
+
"transformers_version": "4.49.0",
|
| 34 |
+
"use_cls_token": false,
|
| 35 |
+
"use_lora_on_lm": false,
|
| 36 |
+
"use_negatives": false,
|
| 37 |
+
"use_pos_emb": true,
|
| 38 |
+
"visual_encoder": "google/siglip-so400m-patch14-224",
|
| 39 |
+
"vocab_size": 128259
|
| 40 |
+
}
|