Skip to content

PTQ-Level Efficiency, QAT-Level Accuracy | 训练后量化的效率,量化感知训练的精度

revert the change

wenhuach218e058fa82090
diff --git a/examples/main.py b/examples/main.py
index f5dcff857..937f0341f 100644
--- a/examples/main.py
+++ b/examples/main.py
@@ -88,18 +88,18 @@ if __name__ == '__main__':
# parser.add_argument("--tasks", default=["lambada_openai", "hellaswag", "winogrande", "piqa"],
# help="lm-eval tasks")
- parser.add_argument("--tasks", default=["lambada_openai","wikitext2",'hendrycksTest-*'],
- help="lm-eval tasks")
+ # parser.add_argument("--tasks", default=["lambada_openai","wikitext2",'hendrycksTest-*'],
+ # help="lm-eval tasks")
# parser.add_argument("--tasks",
# default=['wikitext2', 'ptb-new', 'c4-new', 'lambada_openai', 'hellaswag', 'winogrande', 'piqa',
# 'coqa', 'truthfulqa_mc', 'openbookqa', 'boolq', 'rte', 'arc_easy', 'arc_challenge',
# 'hendrycksTest-*', 'wikitext', 'drop', 'gsm8k'],##all
- # parser.add_argument("--tasks",
- # default=['wikitext2', 'ptb-new', 'c4-new', 'lambada_openai', 'hellaswag', 'winogrande', 'piqa',
- # "hendrycksTest-*", "wikitext", "truthfulqa_mc", "openbookqa", "boolq", "rte",
- # "arc_easy", "arc_challenge"],
- # help="lm-eval tasks") # "truthfulqa_gen"
+ parser.add_argument("--tasks",
+ default=['wikitext2', 'ptb-new', 'c4-new', 'lambada_openai', 'hellaswag', 'winogrande', 'piqa',
+ "hendrycksTest-*", "wikitext", "truthfulqa_mc", "openbookqa", "boolq", "rte",
+ "arc_easy", "arc_challenge"],
+ help="lm-eval tasks")
# parser.add_argument("--tasks", default=["lambada_openai"],
# help="lm-eval tasks")