diff --git a/benchmarks/inference/collect_results.py b/benchmarks/inference/collect_results.py index 74f60e442470..ba58d4d29b86 100644 --- a/benchmarks/inference/collect_results.py +++ b/benchmarks/inference/collect_results.py @@ -75,6 +75,14 @@ def get_generated_text(file_content, gen_text_n): return {f"generated-text-{key}": val for key, val in matches} +def get_error(file_content): + matches = re.findall(r"TypeError:\s+(.+?)\n", file_content) + if matches is []: + return False + else: + return {f"error": val for val in matches} + + if __name__ == "__main__": # List to collect data from all benchmarks benchmarks_data = [] @@ -112,15 +120,17 @@ def get_generated_text(file_content, gen_text_n): perf_data = get_perf_data(file_content) if not perf_data: print( - f"WARNING: Could not detect benchmark performance data for file {file_path}, skipping" + f"WARNING: Could not detect benchmark performance data for file {file_path}" ) - continue generated_text = get_generated_text(file_content, args.gen_text_n) if not generated_text: - print( - f"WARNING: Could not detect generated text for file {file_path}, skipping" - ) + print(f"WARNING: Could not detect generated text for file {file_path}") + + error = get_error(file_content) + if error: + print(f"Error found in {file_path}, collecting error info...") + benchmarks_data.append({"branch": branch, **params, **error}) continue benchmarks_data.append({ diff --git a/benchmarks/inference/run_model.sh b/benchmarks/inference/run_model.sh index 7b7c4b6988e1..8e5fe3ac0133 100644 --- a/benchmarks/inference/run_model.sh +++ b/benchmarks/inference/run_model.sh @@ -1,4 +1,4 @@ -set -ex +set -x model=$1 branch1=$2 diff --git a/benchmarks/inference/sweep.sh b/benchmarks/inference/sweep.sh index 46d90435a8ce..aabcb0bfdbd8 100644 --- a/benchmarks/inference/sweep.sh +++ b/benchmarks/inference/sweep.sh @@ -1,4 +1,4 @@ -set -ex +set -x export TRANSFORMERS_CACHE=/tmp/hf-cache