PR #2661 fix: prefer inplace softmax to avoid copy

fix: prefer inplace softmax to avoid copy #2661

drbh merged 2 commits into main from prefer-inplace-softmax-for-prefill-logprobs

fix: prefer inplace softmax to avoid copy

8d7448de

Narsil dismissed these changes on 2024-10-17

Narsil commented on 2024-10-17

drbh dismissed their stale review via 3e0a82d5 1 year ago

Update server/text_generation_server/models/flash_causal_lm.py

3e0a82d5

drbh merged 5f32dea1 into main 1 year ago

drbh deleted the prefer-inplace-softmax-for-prefill-logprobs branch 1 year ago

Reviewers

Narsil

Assignees

No one assigned

Labels

None yet

Milestone

No milestone