Solve encoding issue of repocard.py (#3235)
I discover a problem (even lots of problems...) when I wanted to use push_to_hub on an agent of smolagents, and the problem is due to huggingface_hub, in the file :
Bug #1: UnicodeEncodeError when creating README.md
Error:
UnicodeEncodeError: 'charmap' codec can't encode character '\U0001f440' (for exemple) in position 32: character maps to <undefined>
Root cause:
Location: Path.write_text() in huggingface_hub/repocard.py:279
Problem: Windows uses CP1252 encoding by default instead of UTF-8
Trigger: A Unicode emoji (\U0001f440 = 👀 for exemple) in the README.md metadata
Context:
smolagents automatically generates a README.md with emojis, but Windows cannot encode them using CP1252
Bug mechanism:
agent.push_to_hub() calls metadata_update()
metadata_update() creates a RepoCard with an emoji
RepoCard.push_to_hub() uses Path.write_text() without specifying UTF-8
Windows defaults to CP1252 → crash on emoji
Solution:
So I juste use encoding="utf-8" in write_text()
