From ac5061fa5d3c3e55e886313cac8f5362f88cbcbc Mon Sep 17 00:00:00 2001 From: Gahow Wang Date: Sun, 19 Jul 2026 06:19:08 +0800 Subject: [PATCH] Align Qwen235 attention profile sequence limit --- runs/frontier-fidelity-envelope-v1/run_qwen235_v020_profiles.sh | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/runs/frontier-fidelity-envelope-v1/run_qwen235_v020_profiles.sh b/runs/frontier-fidelity-envelope-v1/run_qwen235_v020_profiles.sh index ed775b3..1dc0e42 100644 --- a/runs/frontier-fidelity-envelope-v1/run_qwen235_v020_profiles.sh +++ b/runs/frontier-fidelity-envelope-v1/run_qwen235_v020_profiles.sh @@ -35,7 +35,7 @@ common_profile() { env CUDA_VISIBLE_DEVICES="${gpu}" PYTHONPATH="${FRONTIER_SOURCE}" \ "${VENV_ROOT}/bin/python" -m frontier.profiling.attention.main \ --disable_ray --num_gpus 1 --device h20 --output_dir "${root}" --models "${MODEL}" \ - --num_tensor_parallel_workers 4 8 --max_model_len 40960 --max_seq_len 8192 \ + --num_tensor_parallel_workers 4 8 --max_model_len 40960 --max_seq_len 40960 \ --batch_size_list "${BATCHES[@]}" --decode_kv_cache_size_list "${KV[@]}" \ --fixed_chunked_prefill_size 8192 --attention_backend FLASHINFER \ --profile_method "${method}" --yes