huggingface / huggingface/candle
Why is llama_multiprocess's num_shards 4 slower than 2?
Open
- Dominant language
- Rust
- Stars
- 21.1k
- Forks
- 1.8k
- Avg merge
- 16h 42m
- Merged PRs (30d)
- 25
Description
Left：num_shards=2; Right: num_shards=4
Contributor guide
No contributing guide indexed for this repository
Research direction
Start by locating the llama_multiprocess entry point and the num_shards handling, then compare the 2-shard and 4-shard execution paths. Reproduce the reported timing difference and inspect how work is divided across shards. Done means the slowdown is explained and, if it is a defect, a focused fix is identified.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- rust
- Domain
- machine-learning, performance
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100