vyalamar comments

Repositories
Issues
Comments

Results 5 comments of


                                            vyalamar

[Feature] Request to 8-bit Quantization of Attention with SageAttention

@zhaochenyang20 I can take a first pass if that’s open. I’m planning to integrate SageAttention’s 8-bit path behind a runtime flag so we can benchmark and verify correctness before defaulting...

Review all potential panics in the entire codebase.

can I try this as this is a good first issue?

Review all potential panics in the entire codebase.

/assign

[Bug] Decode OOM with spec

@hnyls2002 can I also try this?

[MODEL] Context degradation and working memory failure during problem-solving session

I'd like to help investigate this issue from the client side. Scope: Reproducible analysis of context degradation in long Claude Code sessions — instrumentation and data collection only, not model...