ayo/ipex-llm
1
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 1

add mixed_precision argument on ppl wikitext evaluation (#11813)

Browse source 
* fix: delete ipex extension import in ppl wikitext evaluation

* feat: add mixed_precision argument on ppl wikitext evaluation

* fix: delete mix_precision command in perplex evaluation for wikitext

* fix: remove fp16 mixed-presicion argument

* fix: Add a space.

---------

Co-authored-by: Jinhe Tang <jin.tang1337@gmail.com>
This commit is contained in:
Chu,Youcheng 2024-08-15 17:58:53 +08:00 committed by GitHub
parent 828ab16537
commit 28d1c972da
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 3 additions and 2 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

5
python/llm/dev/benchmark/perplexity/run_wikitext.py
View file

@ -20,7 +20,7 @@
import argparse import argparse
import torch import torch
from tqdm import tqdm from tqdm import tqdm
from datasets import concatenate_datasets, load_dataset from datasets import load_dataset
from ipex_llm.utils.common import invalidInputError from ipex_llm.utils.common import invalidInputError
@ -34,6 +34,7 @@ parser.add_argument("--device", type=str, default="xpu")
parser.add_argument("--precision", type=str, default="sym_int4") parser.add_argument("--precision", type=str, default="sym_int4")
parser.add_argument("--use-cache", action="store_true") parser.add_argument("--use-cache", action="store_true")
parser.add_argument("--max_length", type=int, default=None) parser.add_argument("--max_length", type=int, default=None)
parser.add_argument("--mixed_precision", action="store_true")
args = parser.parse_args() args = parser.parse_args()
if args.precision == "fp16": # ipex fp16 if args.precision == "fp16": # ipex fp16
@ -43,7 +44,7 @@ if args.precision == "fp16": # ipex fp16
else: # ipex-llm else: # ipex-llm
from ipex_llm.transformers import AutoModelForCausalLM from ipex_llm.transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained(args.model_path, load_in_low_bit=args.precision, model = AutoModelForCausalLM.from_pretrained(args.model_path, load_in_low_bit=args.precision,
use_cache=args.use_cache, trust_remote_code=True) use_cache=args.use_cache, trust_remote_code=True, mixed_precision= args.mixed_precision)
model = model.half() model = model.half()
model = model.to(args.device) model = model.to(args.device)
model = model.eval() model = model.eval()