Spaces:

---

Jellyfish042

/

LLM-Compressor

like 1

Running

Fixed, issue caused by RWKV model update

Thanks! But I think I accidentally DoSed the decompressor function by changing the first character of the base64 to decompress, which makes it try to generate a few billion tokens.

You should be able to fix that by adding a max_tokens param to decompress_bytes (conservatively passing in MAX_INPUT_CHARS from the demo app), and making decompress_bytes raise ValueError if total_tokens exceeds this.

Thanks! This is a good catch. Security checks have been added.