Files
vLLM-Kunlunxin_p-800-tokeni…/fix_tokenizer.py

96 lines
3.1 KiB
Python
Raw Permalink Normal View History

2026-06-29 17:23:40 +08:00
#!/usr/bin/env python3
"""
2026-06-30 13:57:38 +08:00
检测并修复 tokenizer_config.json 中的两类问题
1. tokenizer_class transformers 中不存在 TokenizersBackend
2. extra_special_tokens list 格式transformers 要求 dict
若存在问题 tokenizer 文件复制到 /tmp/fixed_tokenizer/ 并修复
最后将修复目录路径输出到 stdout若无需修复输出为空
2026-06-29 17:23:40 +08:00
"""
2026-06-29 17:04:41 +08:00
import os
2026-06-29 17:23:40 +08:00
import sys
2026-06-29 17:04:41 +08:00
import json
2026-06-29 17:23:40 +08:00
import shutil
2026-06-29 17:04:41 +08:00
2026-06-29 17:23:40 +08:00
MODEL_DIR = sys.argv[1] if len(sys.argv) > 1 else os.environ.get("MODEL_DIR", "/model")
OUT_DIR = "/tmp/fixed_tokenizer"
2026-06-29 17:04:41 +08:00
2026-06-29 17:23:40 +08:00
def main():
cfg_path = os.path.join(MODEL_DIR, "tokenizer_config.json")
if not os.path.exists(cfg_path):
return
2026-06-29 17:04:41 +08:00
with open(cfg_path) as f:
cfg = json.load(f)
2026-06-30 13:57:38 +08:00
fixes = []
# --- 检测 1tokenizer_class 是否在 transformers 中存在 ---
2026-06-29 17:23:40 +08:00
tokenizer_class = cfg.get("tokenizer_class", "")
2026-06-30 13:57:38 +08:00
bad_tokenizer_class = False
if tokenizer_class:
import transformers
if getattr(transformers, tokenizer_class, None) is None:
bad_tokenizer_class = True
fixes.append(f"tokenizer_class '{tokenizer_class}' not found in transformers")
2026-06-29 17:23:40 +08:00
2026-06-30 13:57:38 +08:00
# --- 检测 2extra_special_tokens 是否为 list 格式 ---
bad_extra_special_tokens = (
"extra_special_tokens" in cfg
and isinstance(cfg["extra_special_tokens"], list)
2026-06-29 17:23:40 +08:00
)
2026-06-30 13:57:38 +08:00
if bad_extra_special_tokens:
fixes.append("extra_special_tokens is a list, expected dict")
2026-06-29 17:23:40 +08:00
2026-06-30 13:57:38 +08:00
if not fixes:
return # 无需修复
# 复制 tokenizer 文件到临时目录
2026-06-29 17:23:40 +08:00
os.makedirs(OUT_DIR, exist_ok=True)
for fname in [
"tokenizer.json",
"tokenizer_config.json",
"special_tokens_map.json",
"vocab.json",
"merges.txt",
"tokenizer.model",
]:
src = os.path.join(MODEL_DIR, fname)
if os.path.exists(src):
shutil.copy(src, OUT_DIR)
2026-06-30 13:57:38 +08:00
# --- 修复 1替换 tokenizer_class ---
if bad_tokenizer_class:
files = os.listdir(MODEL_DIR)
if "tokenizer.json" in files:
fixed_class = "PreTrainedTokenizerFast"
elif "tokenizer.model" in files:
fixed_class = "LlamaTokenizer"
elif "vocab.json" in files and "merges.txt" in files:
fixed_class = "GPT2TokenizerFast"
else:
fixed_class = "PreTrainedTokenizerFast"
cfg["tokenizer_class"] = fixed_class
print(
f"[fix_tokenizer] tokenizer_class: '{tokenizer_class}''{fixed_class}'",
file=sys.stderr,
)
# --- 修复 2extra_special_tokens list → dict ---
if bad_extra_special_tokens:
orig_list = cfg["extra_special_tokens"]
cfg["extra_special_tokens"] = {token: token for token in orig_list}
print(
f"[fix_tokenizer] extra_special_tokens: list({len(orig_list)}) → dict",
file=sys.stderr,
)
2026-06-29 17:23:40 +08:00
with open(os.path.join(OUT_DIR, "tokenizer_config.json"), "w") as f:
json.dump(cfg, f, indent=2)
print(OUT_DIR) # 输出修复目录,供 entrypoint.sh 捕获
main()