vLLM issue 4232: Total number of attention heads (40) must be divisible by tensor parallel size (6)
repo
Fields
id
sources/mgpu-vllm-4232-head-divisibilityPublisher
vllm-project/vllm
Kind
repo
Published
2024-04-21
Accessed
2026-10-03
notes
Original report of the attention-head divisibility error string, on Qwen-1.5-32B with TP=6.
Notes
Original report of the attention-head divisibility error string, on Qwen-1.5-32B with TP=6.
Cross-references
This record references no other records.