PTQ-Level Efficiency, QAT-Level Accuracy | 训练后量化的效率,量化感知训练的精度
stop try except and add back missing arguments (#884)
Xin He233a6c8f2ab8
diff --git a/auto_round/__main__.py b/auto_round/__main__.pyindex b806b583b..fb5acbd9e 100644--- a/auto_round/__main__.py+++ b/auto_round/__main__.py@@ -18,7 +18,7 @@ import reimport sysfrom auto_round.compressors import BaseCompressor-from auto_round.eval.eval_cli import EvalArgumentParser, _eval_init, eval, eval_task_by_task, eval_with_vllm+from auto_round.eval.eval_cli import EvalArgumentParser, _eval_init, eval, eval_task_by_taskfrom auto_round.schemes import PRESET_SCHEMESfrom auto_round.utils import (clear_memory,diff --git a/auto_round/eval/eval_cli.py b/auto_round/eval/eval_cli.pyindex f66d93316..8a4b61ad9 100644--- a/auto_round/eval/eval_cli.py+++ b/auto_round/eval/eval_cli.py@@ -54,6 +54,7 @@ class EvalArgumentParser(argparse.ArgumentParser):self.add_argument("--disable_trust_remote_code", action="store_true", help="whether to disable trust_remote_code")+ self.add_argument("--seed", default=42, type=int, help="random seed")self.add_argument("--eval_bs", "--bs", "--batch_size", default=None, type=int, help="batch size in evaluation")self.add_argument("--eval_task_by_task", action="store_true", help="whether to eval task by task.")self.add_argument(@@ -113,15 +114,9 @@ def _eval_init(tasks, model_path, device, disable_trust_remote_code=False, dtypedef eval(args):if args.eval_backend == "vllm":- try:- assert isinstance(args.model, str), "vllm evaluation only supports model name or path."- eval_with_vllm(args)- return- except Exception as e: # pragma: no cover- print(f"vllm evaluation failed: {e}, fallback to default hf backend evaluation.")- args.eval_backend = "hf"- clear_memory()-+ assert isinstance(args.model, str), "vllm evaluation only supports model name or path."+ eval_with_vllm(args)+ returntasks, model_args, device_str = _eval_init(args.tasks, args.model, args.device_map, args.disable_trust_remote_code, args.eval_model_dtype)@@ -308,7 +303,7 @@ def eval_with_vllm(args):st = time.time()os.environ["TOKENIZERS_PARALLELISM"] = "false"- device_str, _ = get_device_and_parallelism(args.device)+ device_str, _ = get_device_and_parallelism(args.device_map)eval_model_dtype = get_model_dtype(args.eval_model_dtype, "auto")if (batch_size := args.eval_bs) is None:batch_size = "auto:8"