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 e939009..ed775b3 100644 --- a/runs/frontier-fidelity-envelope-v1/run_qwen235_v020_profiles.sh +++ b/runs/frontier-fidelity-envelope-v1/run_qwen235_v020_profiles.sh @@ -30,7 +30,8 @@ common_profile() { "${VENV_ROOT}/bin/python" -m frontier.profiling.linear_op.main \ --disable_ray --num_gpus 1 --device h20 --output_dir "${root}" --models "${MODEL}" \ --num_tensor_parallel_workers 1 4 8 --attn_tp 4 8 --ffn_tp 1 4 \ - --num_tokens_list "${TOKENS[@]}" --profile_method "${method}" --yes + --max_tokens 8192 --num_tokens_list "${TOKENS[@]}" \ + --profile_method "${method}" --yes env CUDA_VISIBLE_DEVICES="${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}" \ @@ -46,7 +47,7 @@ moe_profile() { "${VENV_ROOT}/bin/python" -m frontier.profiling.moe.main \ --disable_ray --num_gpus 1 --device h20 --output_dir "${root}" --models "${MODEL}" \ --num_tensor_parallel_workers "${tp}" --expert_parallel_sizes "${ep}" \ - --num_tokens_list "${TOKENS[@]}" --load_distributions uniform \ + --max_tokens 8192 --num_tokens_list "${TOKENS[@]}" --load_distributions uniform \ --num_samples_per_distribution 1 --profile_method "${method}" --yes } @@ -59,7 +60,7 @@ moe_profile 4 record_function 4 1 "${OUTPUT_ROOT}/kernel-moe-tp4" > "${OUTPUT_RO moe_profile 5 record_function 1 8 "${OUTPUT_ROOT}/kernel-moe-ep8" > "${OUTPUT_ROOT}/logs/kernel-moe-ep8.log" 2>&1 & pids+=("$!") failed=0 for pid in "${pids[@]}"; do wait "${pid}" || failed=1; done -[[ "${failed}" -eq 0 ]] || { tail -80 "${OUTPUT_ROOT}"/logs/*.log; exit 1; } +[[ "${failed}" -eq 0 ]] || { tail -n 80 "${OUTPUT_ROOT}"/logs/*.log; exit 1; } "${VENV_ROOT}/bin/python" "${RUNNER_DIR}/assemble_qwen235_v020_profiles.py" \ --cuda-common "${OUTPUT_ROOT}/cuda-common" \