Push model using huggingface_hub.
Browse files- README.md +11 -0
- config.json +497 -0
- model.safetensors +3 -0
README.md
ADDED
|
@@ -0,0 +1,11 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
pipeline_tag: video feature extraction
|
| 4 |
+
tags:
|
| 5 |
+
- model_hub_mixin
|
| 6 |
+
- pytorch_model_hub_mixin
|
| 7 |
+
---
|
| 8 |
+
|
| 9 |
+
This model has been pushed to the Hub using the [PytorchModelHubMixin](https://huggingface.co/docs/huggingface_hub/package_reference/mixins#huggingface_hub.PyTorchModelHubMixin) integration:
|
| 10 |
+
- Library: [More Information Needed]
|
| 11 |
+
- Docs: [More Information Needed]
|
config.json
ADDED
|
@@ -0,0 +1,497 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"TextEncoders": {
|
| 3 |
+
"bert": {
|
| 4 |
+
"config": "configs/config_bert.json",
|
| 5 |
+
"d_model": 768,
|
| 6 |
+
"fusion_layer": 9,
|
| 7 |
+
"name": "bert_base",
|
| 8 |
+
"pretrained": "bert-base-uncased"
|
| 9 |
+
},
|
| 10 |
+
"bert_large": {
|
| 11 |
+
"config": "configs/config_bert_large.json",
|
| 12 |
+
"d_model": 1024,
|
| 13 |
+
"fusion_layer": 19,
|
| 14 |
+
"name": "bert_large",
|
| 15 |
+
"pretrained": "bert-large-uncased"
|
| 16 |
+
},
|
| 17 |
+
"med_bert": {
|
| 18 |
+
"config": "configs/med_config.json",
|
| 19 |
+
"d_model": 768,
|
| 20 |
+
"name": "med_bert_base",
|
| 21 |
+
"pretrained": "bert-base-uncased"
|
| 22 |
+
},
|
| 23 |
+
"med_bert_large": {
|
| 24 |
+
"config": "configs/med_large_config.json",
|
| 25 |
+
"d_model": 768,
|
| 26 |
+
"name": "med_bert_large",
|
| 27 |
+
"pretrained": "bert-base-uncased"
|
| 28 |
+
}
|
| 29 |
+
},
|
| 30 |
+
"VisionEncoders": {},
|
| 31 |
+
"auto_resume": true,
|
| 32 |
+
"available_corpus": {
|
| 33 |
+
"anet_ret_train": {
|
| 34 |
+
"anno_path": "your_path",
|
| 35 |
+
"data_root": "",
|
| 36 |
+
"is_paragraph_retrieval": true,
|
| 37 |
+
"max_txt_l": 150,
|
| 38 |
+
"media_type": "video"
|
| 39 |
+
},
|
| 40 |
+
"anet_ret_val": {
|
| 41 |
+
"anno_path": "your_path",
|
| 42 |
+
"data_root": "",
|
| 43 |
+
"is_paragraph_retrieval": true,
|
| 44 |
+
"max_txt_l": 150,
|
| 45 |
+
"media_type": "video"
|
| 46 |
+
},
|
| 47 |
+
"audiocaps_ret_test": {
|
| 48 |
+
"anno_path": "your_path",
|
| 49 |
+
"data_root": "",
|
| 50 |
+
"media_type": "audio"
|
| 51 |
+
},
|
| 52 |
+
"audiocaps_ret_train": {
|
| 53 |
+
"anno_path": "your_path",
|
| 54 |
+
"data_root": "",
|
| 55 |
+
"media_type": "audio"
|
| 56 |
+
},
|
| 57 |
+
"cc12m": {
|
| 58 |
+
"anno_path": "your_path",
|
| 59 |
+
"data_root": "",
|
| 60 |
+
"media_type": "image"
|
| 61 |
+
},
|
| 62 |
+
"cc3m": {
|
| 63 |
+
"anno_path": "your_path",
|
| 64 |
+
"data_root": "",
|
| 65 |
+
"media_type": "image"
|
| 66 |
+
},
|
| 67 |
+
"cc3m_debug": {
|
| 68 |
+
"anno_path": "your_path",
|
| 69 |
+
"data_root": "",
|
| 70 |
+
"media_type": "image"
|
| 71 |
+
},
|
| 72 |
+
"charades_mc_test": {
|
| 73 |
+
"anno_path": "your_path",
|
| 74 |
+
"data_root": "",
|
| 75 |
+
"media_type": "video"
|
| 76 |
+
},
|
| 77 |
+
"clothov1_ret_test": {
|
| 78 |
+
"anno_path": "your_path",
|
| 79 |
+
"data_root": "",
|
| 80 |
+
"media_type": "audio"
|
| 81 |
+
},
|
| 82 |
+
"clothov1_ret_train": {
|
| 83 |
+
"anno_path": "your_path",
|
| 84 |
+
"data_root": "",
|
| 85 |
+
"media_type": "audio"
|
| 86 |
+
},
|
| 87 |
+
"clothov2_ret_test": {
|
| 88 |
+
"anno_path": "your_path",
|
| 89 |
+
"data_root": "",
|
| 90 |
+
"media_type": "audio"
|
| 91 |
+
},
|
| 92 |
+
"clothov2_ret_train": {
|
| 93 |
+
"anno_path": "your_path",
|
| 94 |
+
"data_root": "",
|
| 95 |
+
"media_type": "audio"
|
| 96 |
+
},
|
| 97 |
+
"coco": {
|
| 98 |
+
"anno_path": "your_path",
|
| 99 |
+
"data_root": "",
|
| 100 |
+
"jump_filter": true,
|
| 101 |
+
"media_type": "image"
|
| 102 |
+
},
|
| 103 |
+
"data_25m": [
|
| 104 |
+
{
|
| 105 |
+
"anno_path": "your_path",
|
| 106 |
+
"data_root": "",
|
| 107 |
+
"media_type": "video"
|
| 108 |
+
},
|
| 109 |
+
{
|
| 110 |
+
"anno_path": "your_path",
|
| 111 |
+
"data_root": "",
|
| 112 |
+
"media_type": "image"
|
| 113 |
+
},
|
| 114 |
+
{
|
| 115 |
+
"anno_path": "your_path",
|
| 116 |
+
"data_root": "",
|
| 117 |
+
"jump_filter": true,
|
| 118 |
+
"media_type": "image"
|
| 119 |
+
},
|
| 120 |
+
{
|
| 121 |
+
"anno_path": "your_path",
|
| 122 |
+
"data_root": "",
|
| 123 |
+
"jump_filter": true,
|
| 124 |
+
"media_type": "image"
|
| 125 |
+
},
|
| 126 |
+
{
|
| 127 |
+
"anno_path": "your_path",
|
| 128 |
+
"data_root": "",
|
| 129 |
+
"media_type": "image"
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"anno_path": "your_path",
|
| 133 |
+
"data_root": "",
|
| 134 |
+
"media_type": "image"
|
| 135 |
+
}
|
| 136 |
+
],
|
| 137 |
+
"debug": [
|
| 138 |
+
{
|
| 139 |
+
"anno_path": "your_path",
|
| 140 |
+
"data_root": "",
|
| 141 |
+
"media_type": "image"
|
| 142 |
+
},
|
| 143 |
+
{
|
| 144 |
+
"anno_path": "your_path",
|
| 145 |
+
"data_root": "",
|
| 146 |
+
"media_type": "video"
|
| 147 |
+
}
|
| 148 |
+
],
|
| 149 |
+
"didemo_ret_test": {
|
| 150 |
+
"anno_path": "your_path",
|
| 151 |
+
"data_root": "",
|
| 152 |
+
"is_paragraph_retrieval": true,
|
| 153 |
+
"max_txt_l": 64,
|
| 154 |
+
"media_type": "video",
|
| 155 |
+
"trimmed30": true
|
| 156 |
+
},
|
| 157 |
+
"didemo_ret_train": {
|
| 158 |
+
"anno_path": "your_path",
|
| 159 |
+
"data_root": "",
|
| 160 |
+
"is_paragraph_retrieval": true,
|
| 161 |
+
"max_txt_l": 64,
|
| 162 |
+
"media_type": "video",
|
| 163 |
+
"trimmed30": true
|
| 164 |
+
},
|
| 165 |
+
"didemo_ret_val": {
|
| 166 |
+
"anno_path": "your_path",
|
| 167 |
+
"data_root": "",
|
| 168 |
+
"is_paragraph_retrieval": true,
|
| 169 |
+
"max_txt_l": 64,
|
| 170 |
+
"media_type": "video",
|
| 171 |
+
"trimmed30": true
|
| 172 |
+
},
|
| 173 |
+
"hmdb51_act_val": {
|
| 174 |
+
"anno_path": "your_path",
|
| 175 |
+
"data_root": "",
|
| 176 |
+
"is_act_rec": true,
|
| 177 |
+
"media_type": "video"
|
| 178 |
+
},
|
| 179 |
+
"internvid_v1": {
|
| 180 |
+
"anno_path": "your_path",
|
| 181 |
+
"data_root": "",
|
| 182 |
+
"jump_filter": true,
|
| 183 |
+
"media_type": "video"
|
| 184 |
+
},
|
| 185 |
+
"internvid_v2_avs_private": {
|
| 186 |
+
"anno_path": "your_path",
|
| 187 |
+
"caption_augmentation": {
|
| 188 |
+
"caption_sample_type": "avs_all"
|
| 189 |
+
},
|
| 190 |
+
"data_root": "",
|
| 191 |
+
"jump_filter": true,
|
| 192 |
+
"media_type": "audio_video",
|
| 193 |
+
"read_audio_from_video": true,
|
| 194 |
+
"read_clip_from_video": false,
|
| 195 |
+
"zero_audio_padding_for_video": true
|
| 196 |
+
},
|
| 197 |
+
"k400_act_val": {
|
| 198 |
+
"anno_path": "your_path",
|
| 199 |
+
"data_root": "",
|
| 200 |
+
"is_act_rec": true
|
| 201 |
+
},
|
| 202 |
+
"k600_act_val": {
|
| 203 |
+
"anno_path": "your_path",
|
| 204 |
+
"data_root": "",
|
| 205 |
+
"is_act_rec": true,
|
| 206 |
+
"media_type": "video"
|
| 207 |
+
},
|
| 208 |
+
"k700_act_val": {
|
| 209 |
+
"anno_path": "your_path",
|
| 210 |
+
"data_root": "",
|
| 211 |
+
"is_act_rec": true,
|
| 212 |
+
"media_type": "video"
|
| 213 |
+
},
|
| 214 |
+
"laion_2b": {
|
| 215 |
+
"anno_path": "your_path",
|
| 216 |
+
"data_root": "",
|
| 217 |
+
"jump_filter": true,
|
| 218 |
+
"media_type": "image"
|
| 219 |
+
},
|
| 220 |
+
"laion_coco": {
|
| 221 |
+
"anno_path": "your_path",
|
| 222 |
+
"data_root": "",
|
| 223 |
+
"jump_filter": true,
|
| 224 |
+
"media_type": "image"
|
| 225 |
+
},
|
| 226 |
+
"laion_pop": {
|
| 227 |
+
"anno_path": "your_path",
|
| 228 |
+
"data_root": "",
|
| 229 |
+
"jump_filter": true,
|
| 230 |
+
"media_type": "image"
|
| 231 |
+
},
|
| 232 |
+
"lsmdc_ret_test_1000": {
|
| 233 |
+
"anno_path": "your_path",
|
| 234 |
+
"data_root": "",
|
| 235 |
+
"media_type": "video"
|
| 236 |
+
},
|
| 237 |
+
"lsmdc_ret_train": {
|
| 238 |
+
"anno_path": "your_path",
|
| 239 |
+
"data_root": "",
|
| 240 |
+
"max_txt_l": 96,
|
| 241 |
+
"media_type": "video"
|
| 242 |
+
},
|
| 243 |
+
"lsmdc_ret_val": {
|
| 244 |
+
"anno_path": "your_path",
|
| 245 |
+
"data_root": "",
|
| 246 |
+
"max_txt_l": 96,
|
| 247 |
+
"media_type": "video"
|
| 248 |
+
},
|
| 249 |
+
"mit_act_val": {
|
| 250 |
+
"anno_path": "your_path",
|
| 251 |
+
"data_root": "",
|
| 252 |
+
"is_act_rec": true,
|
| 253 |
+
"media_type": "video"
|
| 254 |
+
},
|
| 255 |
+
"msrvtt_1k_test": {
|
| 256 |
+
"anno_path": "your_path",
|
| 257 |
+
"data_root": "",
|
| 258 |
+
"media_type": "video"
|
| 259 |
+
},
|
| 260 |
+
"msrvtt_ret_test1k": {
|
| 261 |
+
"anno_path": "your_path",
|
| 262 |
+
"data_root": "",
|
| 263 |
+
"media_type": "video"
|
| 264 |
+
},
|
| 265 |
+
"msrvtt_ret_train9k": {
|
| 266 |
+
"anno_path": "your_path",
|
| 267 |
+
"data_root": "",
|
| 268 |
+
"media_type": "video"
|
| 269 |
+
},
|
| 270 |
+
"msvd_ret_test": {
|
| 271 |
+
"anno_path": "your_path",
|
| 272 |
+
"data_root": "",
|
| 273 |
+
"max_txt_l": 64,
|
| 274 |
+
"media_type": "video"
|
| 275 |
+
},
|
| 276 |
+
"msvd_ret_train": {
|
| 277 |
+
"anno_path": "your_path",
|
| 278 |
+
"data_root": "",
|
| 279 |
+
"has_multi_txt_gt": true,
|
| 280 |
+
"max_txt_l": 64,
|
| 281 |
+
"media_type": "video"
|
| 282 |
+
},
|
| 283 |
+
"msvd_ret_val": {
|
| 284 |
+
"anno_path": "your_path",
|
| 285 |
+
"data_root": "",
|
| 286 |
+
"max_txt_l": 64,
|
| 287 |
+
"media_type": "video"
|
| 288 |
+
},
|
| 289 |
+
"pretrain_example_data_1B": [
|
| 290 |
+
{
|
| 291 |
+
"anno_path": "your_path",
|
| 292 |
+
"data_root": "",
|
| 293 |
+
"media_type": "image"
|
| 294 |
+
},
|
| 295 |
+
{
|
| 296 |
+
"anno_path": "your_path",
|
| 297 |
+
"data_root": "",
|
| 298 |
+
"media_type": "video"
|
| 299 |
+
}
|
| 300 |
+
],
|
| 301 |
+
"pretrain_example_data_6B": [
|
| 302 |
+
{
|
| 303 |
+
"anno_path": "your_path",
|
| 304 |
+
"data_root": "",
|
| 305 |
+
"media_type": "image"
|
| 306 |
+
},
|
| 307 |
+
{
|
| 308 |
+
"anno_path": "your_path",
|
| 309 |
+
"data_root": "",
|
| 310 |
+
"media_type": "video"
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"anno_path": "your_path",
|
| 314 |
+
"caption_augmentation": {
|
| 315 |
+
"caption_sample_type": "avs_all"
|
| 316 |
+
},
|
| 317 |
+
"data_root": "",
|
| 318 |
+
"jump_filter": true,
|
| 319 |
+
"media_type": "audio_video",
|
| 320 |
+
"read_audio_from_video": true,
|
| 321 |
+
"read_clip_from_video": false,
|
| 322 |
+
"zero_audio_padding_for_video": true
|
| 323 |
+
}
|
| 324 |
+
],
|
| 325 |
+
"sbu": {
|
| 326 |
+
"anno_path": "your_path",
|
| 327 |
+
"data_root": "",
|
| 328 |
+
"media_type": "image"
|
| 329 |
+
},
|
| 330 |
+
"ssv2_mc_val": {
|
| 331 |
+
"anno_path": "your_path",
|
| 332 |
+
"data_root": "",
|
| 333 |
+
"media_type": "video"
|
| 334 |
+
},
|
| 335 |
+
"ucf101_act_val": {
|
| 336 |
+
"anno_path": "your_path",
|
| 337 |
+
"data_root": "",
|
| 338 |
+
"is_act_rec": true,
|
| 339 |
+
"media_type": "video"
|
| 340 |
+
},
|
| 341 |
+
"vatex_ch_ret_val": {
|
| 342 |
+
"anno_path": "your_path",
|
| 343 |
+
"data_root": "",
|
| 344 |
+
"media_type": "video"
|
| 345 |
+
},
|
| 346 |
+
"vatex_en_ret_train": {
|
| 347 |
+
"anno_path": "your_path",
|
| 348 |
+
"data_root": "",
|
| 349 |
+
"has_multi_txt_gt": true,
|
| 350 |
+
"media_type": "video"
|
| 351 |
+
},
|
| 352 |
+
"vatex_en_ret_val": {
|
| 353 |
+
"anno_path": "your_path",
|
| 354 |
+
"data_root": "",
|
| 355 |
+
"media_type": "video"
|
| 356 |
+
},
|
| 357 |
+
"vg": {
|
| 358 |
+
"anno_path": "your_path",
|
| 359 |
+
"data_root": "",
|
| 360 |
+
"jump_filter": true,
|
| 361 |
+
"media_type": "image"
|
| 362 |
+
},
|
| 363 |
+
"wavcaps_400k": {
|
| 364 |
+
"anno_path": "your_path",
|
| 365 |
+
"data_root": "",
|
| 366 |
+
"media_type": "audio"
|
| 367 |
+
},
|
| 368 |
+
"webvid": {
|
| 369 |
+
"anno_path": "your_path",
|
| 370 |
+
"data_root": "",
|
| 371 |
+
"media_type": "video"
|
| 372 |
+
},
|
| 373 |
+
"webvid_10m": {
|
| 374 |
+
"anno_path": "your_path",
|
| 375 |
+
"data_root": "",
|
| 376 |
+
"media_type": "video"
|
| 377 |
+
},
|
| 378 |
+
"webvid_debug": {
|
| 379 |
+
"anno_path": "your_path",
|
| 380 |
+
"data_root": "",
|
| 381 |
+
"media_type": "video"
|
| 382 |
+
},
|
| 383 |
+
"webvid_fuse_10m": {
|
| 384 |
+
"anno_path": "your_path",
|
| 385 |
+
"data_root": "",
|
| 386 |
+
"jump_filter": true,
|
| 387 |
+
"media_type": "video"
|
| 388 |
+
}
|
| 389 |
+
},
|
| 390 |
+
"batch_size": 8,
|
| 391 |
+
"batch_size_test": 4,
|
| 392 |
+
"compile_model": false,
|
| 393 |
+
"debug": false,
|
| 394 |
+
"deep_fusion": false,
|
| 395 |
+
"deepspeed": {
|
| 396 |
+
"enable": true,
|
| 397 |
+
"stage": 1
|
| 398 |
+
},
|
| 399 |
+
"device": "cuda",
|
| 400 |
+
"dist_url": "env://",
|
| 401 |
+
"evaluate": true,
|
| 402 |
+
"evaluation": {
|
| 403 |
+
"eval_frame_ensemble": "concat",
|
| 404 |
+
"eval_offload": true,
|
| 405 |
+
"eval_x_only": false,
|
| 406 |
+
"k_test": 128
|
| 407 |
+
},
|
| 408 |
+
"gradient_checkpointing": true,
|
| 409 |
+
"inputs": {
|
| 410 |
+
"batch_size": {
|
| 411 |
+
"image": 8,
|
| 412 |
+
"video": 8
|
| 413 |
+
},
|
| 414 |
+
"batch_size_test": {
|
| 415 |
+
"image": 4,
|
| 416 |
+
"video": 4
|
| 417 |
+
},
|
| 418 |
+
"image_res": 224,
|
| 419 |
+
"max_txt_l": {
|
| 420 |
+
"image": 40,
|
| 421 |
+
"video": 40
|
| 422 |
+
},
|
| 423 |
+
"video_input": {
|
| 424 |
+
"num_frames": 4,
|
| 425 |
+
"num_frames_test": 4,
|
| 426 |
+
"random_aug": false,
|
| 427 |
+
"sample_type": "rand",
|
| 428 |
+
"sample_type_test": "middle"
|
| 429 |
+
}
|
| 430 |
+
},
|
| 431 |
+
"is_pretrain": true,
|
| 432 |
+
"jump_evaluate": false,
|
| 433 |
+
"log_freq": 100,
|
| 434 |
+
"max_txt_l": 40,
|
| 435 |
+
"mode": "pt",
|
| 436 |
+
"model": {
|
| 437 |
+
"embed_dim": 512,
|
| 438 |
+
"find_unused_parameters": false,
|
| 439 |
+
"model_cls": "InternVideo2_Stage2",
|
| 440 |
+
"multimodal": {
|
| 441 |
+
"enable": true
|
| 442 |
+
},
|
| 443 |
+
"temp": 0.07,
|
| 444 |
+
"text_encoder": {
|
| 445 |
+
"config": "configs/config_bert_large.json",
|
| 446 |
+
"d_model": 1024,
|
| 447 |
+
"fusion_layer": 19,
|
| 448 |
+
"name": "bert_large",
|
| 449 |
+
"pretrained": "bert-large-uncased"
|
| 450 |
+
},
|
| 451 |
+
"vision_encoder": {
|
| 452 |
+
"checkpoint_num": 40,
|
| 453 |
+
"clip_embed_dim": 768,
|
| 454 |
+
"clip_input_resolution": 224,
|
| 455 |
+
"clip_norm_type": "l2",
|
| 456 |
+
"clip_return_layer": 6,
|
| 457 |
+
"clip_student_return_interval": 1,
|
| 458 |
+
"clip_teacher": null,
|
| 459 |
+
"clip_teacher_embed_dim": 3200,
|
| 460 |
+
"clip_teacher_final_dim": 768,
|
| 461 |
+
"clip_teacher_return_interval": 1,
|
| 462 |
+
"d_model": 1408,
|
| 463 |
+
"image_mask_ratio": 0.5,
|
| 464 |
+
"image_mask_type": "random",
|
| 465 |
+
"img_size": 224,
|
| 466 |
+
"keep_temporal": false,
|
| 467 |
+
"name": "pretrain_internvideo2_6b_patch14_224",
|
| 468 |
+
"num_frames": 4,
|
| 469 |
+
"only_mask": true,
|
| 470 |
+
"patch_size": 14,
|
| 471 |
+
"pretrained": "/mnt/petrelfs/wangyi/internvideo2_ckpt/internvideo2-s2_6b-224p-f4.pt",
|
| 472 |
+
"sep_image_video_pos_embed": true,
|
| 473 |
+
"tubelet_size": 1,
|
| 474 |
+
"use_checkpoint": true,
|
| 475 |
+
"use_flash_attn": false,
|
| 476 |
+
"use_fused_mlp": false,
|
| 477 |
+
"use_fused_rmsnorm": false,
|
| 478 |
+
"video_mask_ratio": 0.8,
|
| 479 |
+
"video_mask_type": "random"
|
| 480 |
+
}
|
| 481 |
+
},
|
| 482 |
+
"num_frames": 4,
|
| 483 |
+
"num_frames_test": 4,
|
| 484 |
+
"num_workers": 6,
|
| 485 |
+
"origin_num_frames": 4,
|
| 486 |
+
"output_dir": null,
|
| 487 |
+
"pretrained_path": "",
|
| 488 |
+
"resume": false,
|
| 489 |
+
"save_latest": false,
|
| 490 |
+
"seed": 42,
|
| 491 |
+
"size_t": 224,
|
| 492 |
+
"text_enc": "bert_large",
|
| 493 |
+
"use_bf16": false,
|
| 494 |
+
"use_flash_sdp": false,
|
| 495 |
+
"use_half_precision": false,
|
| 496 |
+
"use_mem_efficient_sdp": false
|
| 497 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0ca222644b70507c092811e5b9e498f07b0d52afd2c313fe527b205afd998a9a
|
| 3 |
+
size 25466142320
|