Skip to content
Merged
Show file tree
Hide file tree
Changes from 4 commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Original file line number Diff line number Diff line change
Expand Up @@ -3,8 +3,10 @@ accuracy_threshold: 0.90
num_questions: 1319
num_fewshot: 5
max_concurrency: 100
# Preserve KV-cache headroom after CUDA graph memory profiling.
server_args: >-
--compilation-config '{"cudagraph_mode":"PIECEWISE"}'
--gpu-memory-utilization 0.96
--max-model-len 4096
--max-num-batched-tokens 32768
--safetensors-load-strategy prefetch
Expand Down
2 changes: 2 additions & 0 deletions tests/evals/gsm8k/configs/GLM-5.2-NVFP4-TP2-PCP2-EP.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -3,8 +3,10 @@ accuracy_threshold: 0.90
num_questions: 1319
num_fewshot: 5
max_concurrency: 100
# Leave headroom for FlashInfer CUTLASS autotuning after KV-cache allocation.
server_args: >-
--compilation-config '{"cudagraph_mode":"NONE"}'
--gpu-memory-utilization 0.85
--max-model-len 4096
--max-num-batched-tokens 32768
--safetensors-load-strategy prefetch
Expand Down
Loading