[GLM4.1V and GLM4.5V] Add vision transformer num_dummy_head support: max tp=4 -> max tp=8 (#9059)

This commit is contained in:
Binyao Jiang
2025-08-18 14:40:13 -07:00
committed by GitHub
parent 98b44e9e56
commit c2fbf60f39
9 changed files with 150 additions and 102 deletions

View File

@@ -187,9 +187,13 @@ async def eval_mmmu(args) -> None:
print("Profiler stopped")
print(f"Benchmark time: {time.perf_counter() - start}")
args.output_path = f"./val_sglang.json"
args.output_path = "./answer_sglang.json"
save_json(args.output_path, out_samples)
eval_result(model_answer_path=args.output_path, answer_dict=answer_dict)
eval_result(
model_answer_path=args.output_path,
answer_dict=answer_dict,
eval_output_path="./val_sglang.json",
)
def parse_args():