pullrequests Search Results · bucket language:JavaScript language:Java language:Python language:JavaScript language:JavaScript
Filter by
1M results
Summary
- gmm_v2 computes wrong rows when a tiling splits the contraction (tile_k K) and a group ends inside a row tile (a row
bucket below tile_m, which happens whenever tile_m bucket ...
Changed allgather_bucket_size from 5e8 to 1e8 in zero_optimization settings.
Sourcery 摘要
增强功能:
- 将 DeepSpeed ZeRO 的 all-gather bucket 大小从 5e8 减小至 1e8。
details summary Original summary in ...
... aside a strip whose pass carries less than half a bucket to the carrier. In every soil, set-aside strips are
only used when no other strip has a cut worth taking.
- RESULTS.md: new section ...
... of serialized dtype buckets. A rank with no
tensors could either construct FlattenedTensorBucket([]) before participating in the Gloo gather, or leave the gather
source indexing a missing column when ...
First demo PR. Its signed decision receipt was produced locally with the same pipeline the Action runs (see the README
section A receipt you can verify right now ). The PR stays a draft because this repository ...
🚨 Severity: HIGH
💡 Vulnerability: get_user_id used get_unverified_claims to extract the user ID for rate limiting, allowing an attacker
to spoof tokens and poison other users rate limit buckets ...
dependencies
python:uv
Description
Models that set use_raw_control_images = True (Flux.2, Qwen Image Edit Plus, Flux3, Krea2, Fibo2, Mageflow, ...) skip
the bucket resize, crop and flip for control images in ...