Skip to content

add mixed_precision argument on ppl wikitext evaluation #11813

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 6 commits into from
Aug 15, 2024
Merged

add mixed_precision argument on ppl wikitext evaluation #11813

Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
eda17e9
fix: delete ipex extension import in ppl wikitext evaluation
ada-jt1725 Aug 15, 2024
06b1a31
feat: add mixed_precision argument on ppl wikitext evaluation
ada-jt1725 Aug 15, 2024
8aa16c2
Merge branch 'main' into main
cranechu0131 Aug 15, 2024
997bb2e
fix: delete mix_precision command in perplex evaluation for wikitext
ada-jt1725 Aug 15, 2024
e00d95b
fix: remove fp16 mixed-presicion argument
ada-jt1725 Aug 15, 2024
bb907eb
fix: Add a space.
ada-jt1725 Aug 15, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 3 additions & 2 deletions python/llm/dev/benchmark/perplexity/run_wikitext.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -20,7 +20,7 @@
import argparse
import torch
from tqdm import tqdm
from datasets import concatenate_datasets, load_dataset
from datasets import load_dataset
from ipex_llm.utils.common import invalidInputError


Expand All @@ -34,6 +34,7 @@
parser.add_argument("--precision", type=str, default="sym_int4")
parser.add_argument("--use-cache", action="store_true")
parser.add_argument("--max_length", type=int, default=None)
parser.add_argument("--mixed_precision", action="store_true")
args = parser.parse_args()

if args.precision == "fp16": # ipex fp16
Expand All @@ -43,7 +44,7 @@
else: # ipex-llm
from ipex_llm.transformers import AutoModelForCausalLM
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.to(args.device)
model = model.eval()
Expand Down