46a1cbfa64
* feat: add mixed_precision argument on ppl longbench evaluation * fix: delete two spaces --------- Co-authored-by: Jinhe Tang <jin.tang1337@gmail.com>
114 lines
No EOL
4.5 KiB
Python
114 lines
No EOL
4.5 KiB
Python
#
|
|
# Copyright 2016 The BigDL Authors.
|
|
#
|
|
# Licensed under the Apache License, Version 2.0 (the "License");
|
|
# you may not use this file except in compliance with the License.
|
|
# You may obtain a copy of the License at
|
|
#
|
|
# http://www.apache.org/licenses/LICENSE-2.0
|
|
#
|
|
# Unless required by applicable law or agreed to in writing, software
|
|
# distributed under the License is distributed on an "AS IS" BASIS,
|
|
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
# See the License for the specific language governing permissions and
|
|
# limitations under the License.
|
|
#
|
|
# This file is adapted from
|
|
# https://github.com/insuhan/hyper-attn/blob/main/benchmark_patch_llm.py
|
|
#
|
|
|
|
import argparse
|
|
from tqdm import tqdm
|
|
import torch
|
|
from datasets import concatenate_datasets, load_dataset
|
|
from transformers import AutoTokenizer
|
|
|
|
from ppl import BigDLPPL
|
|
from ipex_llm.ggml.quantize import ggml_tensor_qtype
|
|
|
|
import os
|
|
import json
|
|
|
|
def get_arguments():
|
|
parser = argparse.ArgumentParser()
|
|
parser.add_argument("--seq_len", type=int, default=512)
|
|
parser.add_argument("--model_path", required=True, type=str)
|
|
parser.add_argument("--datasets", required=False, type=str, default=None, nargs='*')
|
|
parser.add_argument("--dataset_path", required=False, type=str, default=None)
|
|
parser.add_argument("--language", required=False, type=str, default="en", choices=['en', 'zh', 'all'])
|
|
parser.add_argument("--precisions", required=False, type=str, default=None, nargs='+')
|
|
parser.add_argument("--mixed_precision", action="store_true")
|
|
parser.add_argument("--device", type=str, default="xpu")
|
|
parser.add_argument("--output_path", default=None)
|
|
return parser.parse_args()
|
|
|
|
|
|
@torch.no_grad()
|
|
def main():
|
|
args = get_arguments()
|
|
for arg_name, arg_var in args.__dict__.items():
|
|
print(f"{arg_name:<16} : {arg_var}")
|
|
|
|
tokenizer = AutoTokenizer.from_pretrained(args.model_path, trust_remote_code=True)
|
|
tokenizer.model_max_length = args.seq_len
|
|
|
|
en_datasets = ["narrativeqa", "qasper", "multifieldqa_en", "hotpotqa", "2wikimqa", "musique", "gov_report",
|
|
"qmsum", "multi_news", "trec", "triviaqa", "samsum", "passage_count", "passage_retrieval_en"]
|
|
zh_datasets = ["multifieldqa_zh", "dureader", "vcsum", "lsht", "passage_retrieval_zh"]
|
|
|
|
if args.datasets is None:
|
|
if args.language == 'en':
|
|
datasets = en_datasets
|
|
elif args.language == 'zh':
|
|
datasets = zh_datasets
|
|
else:
|
|
datasets = en_datasets + zh_datasets
|
|
else:
|
|
datasets = args.datasets
|
|
|
|
dataset_all = []
|
|
for dataset_name in datasets:
|
|
data_ = load_dataset(os.path.join(args.dataset_path, dataset_name), split='test') if args.dataset_path \
|
|
else load_dataset('THUDM/LongBench', f'{dataset_name}', split='test')
|
|
dataset_all.append(data_)
|
|
data = concatenate_datasets(dataset_all)
|
|
|
|
encoded_texts = []
|
|
pbar = tqdm(data)
|
|
for i, data_i in enumerate(pbar):
|
|
encoded_text = tokenizer.encode(data_i['context'], return_tensors='pt', truncation=True)
|
|
pbar.set_description(f"seq_len: {len(encoded_text[0])}, n_data: {len(encoded_texts)}")
|
|
if len(encoded_text[0]) < args.seq_len:
|
|
continue
|
|
encoded_texts.append(encoded_text)
|
|
|
|
summary = {}
|
|
output_path = args.output_path if args.output_path else "results"
|
|
model_name = os.path.basename(os.path.realpath(args.model_path))
|
|
for precision in args.precisions:
|
|
model_kwargs = {}
|
|
if precision in ggml_tensor_qtype.keys():
|
|
model_kwargs['load_in_low_bit'] = precision
|
|
else:
|
|
model_kwargs['torch_dtype'] = getattr(torch, precision)
|
|
print(model_kwargs)
|
|
|
|
log_dir = f"{output_path}/{model_name}/{args.device}/{precision}/{args.language}"
|
|
os.makedirs(log_dir, exist_ok=True)
|
|
results = {}
|
|
ppl_evaluator = BigDLPPL(model_path=args.model_path, device=args.device, mixed_precision=args.mixed_precision, **model_kwargs)
|
|
ppl = ppl_evaluator.perplexity_hf(encoded_texts)
|
|
summary[precision] = ppl
|
|
results['results'] = ppl
|
|
results['config'] = {"model": model_name, "precision": precision, "mixed_precision": args.mixed_precision, "device": args.device, "seq_len": args.seq_len, "language": args.language }
|
|
dumped = json.dumps(results, indent=2)
|
|
print(dumped)
|
|
|
|
if output_path:
|
|
with open(f"{log_dir}/result.json", "w") as f:
|
|
f.write(dumped)
|
|
|
|
print(summary)
|
|
|
|
if __name__ == "__main__":
|
|
main() |