SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

Fix some typos about benchmark scripts. (#37027)

Browse Source 
Signed-off-by: zhanluxianshen <zhanluxianshen@163.com>
This commit is contained in:
湛露先生
2025-03-28 22:10:20 +08:00
committed by GitHub
parent 1fcaad6df9
commit c90e6e9625
3 changed files with 5 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
benchmark/llama.py
View File

@@ -215,7 +215,7 @@ def run_benchmark(logger: Logger, branch: str, commit_id: str, commit_msg: str,
torch.cuda.synchronize()
end = perf_counter()
time_to_second_token = end - start
logger.info(f"completed second compile generation in: {time_to_first_token}s")
logger.info(f"completed second compile generation in: {time_to_second_token}s")
cache_position += 1
all_generated_tokens += next_token.clone().detach().cpu().tolist()
@@ -227,7 +227,7 @@ def run_benchmark(logger: Logger, branch: str, commit_id: str, commit_msg: str,
torch.cuda.synchronize()
end = perf_counter()
time_to_third_token = end - start
logger.info(f"completed third compile forward in: {time_to_first_token}s")
logger.info(f"completed third compile forward in: {time_to_third_token}s")
cache_position += 1
all_generated_tokens += next_token.clone().detach().cpu().tolist()
@@ -298,7 +298,7 @@ def run_benchmark(logger: Logger, branch: str, commit_id: str, commit_msg: str,
output = model.generate(**inputs, past_key_values=past_key_values)
end = perf_counter()
third_compile_generate_time = end - start
logger.info(f"completed second compile generation in: {third_compile_generate_time}s")
logger.info(f"completed third compile generation in: {third_compile_generate_time}s")
logger.info(f"generated: {tokenizer.batch_decode(output.cpu().tolist())}")
past_key_values = StaticCache(
@@ -313,7 +313,7 @@ def run_benchmark(logger: Logger, branch: str, commit_id: str, commit_msg: str,
output = model.generate(**inputs, past_key_values=past_key_values)
end = perf_counter()
fourth_compile_generate_time = end - start
logger.info(f"completed second compile generation in: {fourth_compile_generate_time}s")
logger.info(f"completed fourth compile generation in: {fourth_compile_generate_time}s")
logger.info(f"generated: {tokenizer.batch_decode(output.cpu().tolist())}")
metrics_recorder.collect_model_measurements(