dev(narugo): add tag lazy mode, ci skip

.github/workflows/wd14.yml

@@ -3,6 +3,11 @@ name: Sync WD14 Models
 on:
   #  push:
   workflow_dispatch:
+    inputs:
+      tag_lazy_mode:
+        description: 'Enable Tag Lazy Mode'
+        type: boolean
+        default: false
   schedule:
     - cron: '30 18 * * 0'
 
@@ -36,6 +41,16 @@ jobs:
           if [ -f requirements-test.txt ]; then pip install -r requirements-test.txt; fi
           if [ -f requirements-test.txt ]; then pip install -r requirements-zoo.txt; fi
           pip install --upgrade build
+      - name: Enable Tag Lazy Mode
+        if: ${{ (github.event.inputs.drop_multi || 'false') == 'true' }}
+        shell: bash
+        run: |
+          echo 'TAG_LAZY_MODE=1' >> $GITHUB_ENV
+      - name: Disable Tag Lazy Mode
+        if: ${{ (github.event.inputs.drop_multi || 'false') == 'false' }}
+        shell: bash
+        run: |
+          echo 'TAG_LAZY_MODE=' >> $GITHUB_ENV
       - name: Sync Models
         env:
           HF_TOKEN: ${{ secrets.HF_TOKEN }}

zoo/wd14/sync.py

@@ -51,7 +51,7 @@ def _seg_split(text):
 }
 
 
-def sync():
+def sync(tag_lazy_mode: bool = False):
     hf_fs = get_hf_fs()
 
     import onnxruntime
@@ -134,7 +134,7 @@ def _is_fc(name):
             else:
                 invertible = False
 
-            df = _make_tag_info(model_name)
+            df = _make_tag_info(model_name, lazy_mode=tag_lazy_mode)
             assert len(df) == _get_model_tags_length(model_name)
             df.to_csv(os.path.join(td, MODEL_NAMES[model_name], 'tags_info.csv'), index=False)
 
@@ -176,4 +176,6 @@ def _is_fc(name):
 
 
 if __name__ == '__main__':
-    sync()
+    sync(
+        tag_lazy_mode=bool(os.environ.get('TAG_LAZY_MODE')),
+    )

zoo/wd14/tags.py

@@ -1,10 +1,13 @@
 import json
 import logging
+import os
 from functools import lru_cache
-from typing import List, Set
+from typing import List, Set, Dict
 
+import numpy as np
 import pandas as pd
 from ditk import logging
+from hfutils.operate import get_hf_fs
 from huggingface_hub import hf_hub_download
 from tqdm import tqdm
 from waifuc.source import DanbooruSource
@@ -79,7 +82,24 @@ def _tags_name_set(model_name) -> Set[str]:
     return set(_tags_list(model_name)['name'])
 
 
-def _make_tag_info(model_name='ConvNext') -> pd.DataFrame:
+@lru_cache()
+def _load_all_previous(repository: str = 'deepghs/wd14_tagger_with_embeddings') -> Dict[int, dict]:
+    hf_fs = get_hf_fs()
+    d = {}
+    for path in hf_fs.glob(f'{repository}/**/tags_info.csv'):
+        relpath = os.path.relpath(path, f'{repository}')
+        df = pd.read_csv(hf_hub_download(
+            repo_id=repository,
+            repo_type='model',
+            filename=relpath,
+        )).replace(np.nan, None)
+        for item in df.to_dict('records'):
+            if item['tag_id'] not in d:
+                d[item['tag_id']] = item
+    return d
+
+
+def _make_tag_info(model_name='ConvNext', lazy_mode: bool = False) -> pd.DataFrame:
     with open(hf_hub_download(
             repo_id='deepghs/tags_meta',
             repo_type='dataset',
@@ -88,18 +108,22 @@ def _make_tag_info(model_name='ConvNext') -> pd.DataFrame:
         attire_tags = json.load(f)
 
     df = _tags_list(model_name)
+    d = _load_all_previous()
     records = []
     for item in tqdm(df.to_dict('records')):
-        if item['category'] != 9:
-            tag_info = _get_tag_by_id(item['tag_id'])
-            item['count'] = tag_info['post_count']
-            aliases = _search_related_tags(item['name'], model_name)
-            logging.info(f'Aliases {aliases!r} --> {item["name"]!r}')
-            item['aliases'] = json.dumps(aliases)
+        if lazy_mode and item['tag_id'] in d:
+            item = d[item['tag_id']]
         else:
-            item['aliases'] = json.dumps([item['name']])
-        item['is_core'] = (item['category'] == 0) and (is_basic_character_tag(item['name']))
-        item['is_attire'] = (item['category'] == 0) and (item['name'] in attire_tags)
+            if item['category'] != 9:
+                tag_info = _get_tag_by_id(item['tag_id'])
+                item['count'] = tag_info['post_count']
+                aliases = _search_related_tags(item['name'], model_name)
+                logging.info(f'Aliases {aliases!r} --> {item["name"]!r}')
+                item['aliases'] = json.dumps(aliases)
+            else:
+                item['aliases'] = json.dumps([item['name']])
+            item['is_core'] = (item['category'] == 0) and (is_basic_character_tag(item['name']))
+            item['is_attire'] = (item['category'] == 0) and (item['name'] in attire_tags)
         records.append(item)
 
     df_records = pd.DataFrame(records)