[Important] Previous Upload Was Corrupted — Latest Version Verified & Complete
Hi everyone,
I want to flag an issue with a previous version of this model that was uploaded to this repository.
The earlier upload was incomplete/corrupted due to an error during the upload process. Some model shards may have been missing or improperly transferred, which could have caused loading failures or unexpected behavior.
The repository has now been updated. The latest version has been fully re-uploaded and verified — all shards are intact and the model loads correctly end-to-end. I've tested it locally with both transformers and a standard inference pipeline to confirm everything is working as expected.
What you should do:
- If you previously downloaded this model, please re-download it to get the corrected files.
- If you cached the old version locally, clear your cache before loading (
~/.cache/huggingface/hub).
Sorry for any inconvenience this may have caused. Feel free to reply here if you run into any issues with the new version — I'll be monitoring this thread.
— stevewu
Great job! I've tried other "heretic" abliterated models but they are not truly uncensored in thinking mode, also when thinking is disabled they refuse. This one really feels like the original but without guardrails. Thanks!