mirror of
https://github.com/hanzoai/open-instruct.git
synced 2026-10-30 18:26:32 +00:00
74069db498
* Reorganize the data preparation scripts for tulu v1 and v2. * Minor improvement * Remove open_platypus_commercial subset from Daring-Anteater * Use hard-coded examples repo. * Fix some bugs. * Add OpenMathInstruct. * Add a few more v3.5.x SFT mix ablations for the cleaner datasets. * More experiments on mixes. * help merge * prep for merge * reapply changes * fix naming --------- Co-authored-by: Nathan Lambert <[email protected]>
116 lines
3.5 KiB
Python
116 lines
3.5 KiB
Python
"""
|
|
This script is largely copied from the Vicuna repo: https://github.com/lm-sys/FastChat/blob/main/fastchat/data/split_long_conversation.py
|
|
We fixed a bug in `split_one_sample`, which previously includes long conversations in the processed data. Now we skip these long conversations.
|
|
"""
|
|
import argparse
|
|
from concurrent.futures import ProcessPoolExecutor
|
|
import json
|
|
import transformers
|
|
from tqdm import tqdm
|
|
|
|
|
|
def make_sample(sample, start_idx, end_idx):
|
|
assert (end_idx - start_idx) % 2 == 0
|
|
return {
|
|
"id": sample["id"] + "_" + str(start_idx),
|
|
"conversations": sample["conversations"][start_idx:end_idx],
|
|
}
|
|
|
|
|
|
tokenizer = max_length = None
|
|
|
|
|
|
def split_one_sample(sample):
|
|
tokenized_lens = []
|
|
conversations = sample["conversations"]
|
|
conversations = conversations[: len(conversations) // 2 * 2]
|
|
for c in conversations:
|
|
length = len(tokenizer(c["value"]).input_ids) + 6
|
|
tokenized_lens.append(length)
|
|
|
|
start_idx = 0
|
|
cur_len = 0
|
|
|
|
if len(conversations) % 2 != 0 or len(conversations) < 2:
|
|
return []
|
|
|
|
new_samples = []
|
|
for i in range(0, len(conversations), 2):
|
|
tmp_len = tokenized_lens[i] + tokenized_lens[i + 1]
|
|
if cur_len + tmp_len > max_length:
|
|
new_samples.append(make_sample(sample, start_idx, i))
|
|
if tmp_len > max_length: # if the current conversation is too long, we should skip it
|
|
start_idx = i + 2
|
|
else:
|
|
start_idx = i
|
|
cur_len = 0
|
|
elif i == len(conversations) - 2:
|
|
new_samples.append(make_sample(sample, start_idx, i + 2))
|
|
|
|
cur_len += tmp_len
|
|
|
|
return new_samples
|
|
|
|
|
|
def split_all(content, begin, end, tokenizer_, max_length_):
|
|
"""
|
|
Keep the maximum round of conversations within the max token length constraint
|
|
"""
|
|
global tokenizer, max_length
|
|
tokenizer = tokenizer_
|
|
max_length = max_length_
|
|
|
|
content = content[begin:end]
|
|
new_content = []
|
|
|
|
with ProcessPoolExecutor(max_workers=128) as executor:
|
|
for result in tqdm(executor.map(split_one_sample, content), total=len(content)):
|
|
new_content.extend(result)
|
|
|
|
return new_content
|
|
|
|
|
|
def filter_invalid_roles(content):
|
|
new_content = []
|
|
for i, c in enumerate(content):
|
|
roles = ["human", "gpt"]
|
|
if len(c["conversations"]) <= 0:
|
|
continue
|
|
|
|
valid = True
|
|
for j, s in enumerate(c["conversations"]):
|
|
if s["from"] != roles[j % 2]:
|
|
valid = False
|
|
break
|
|
|
|
if valid:
|
|
new_content.append(c)
|
|
|
|
return new_content
|
|
|
|
|
|
def main(args):
|
|
content = []
|
|
for file in args.in_files:
|
|
content.extend(json.load(open(file)))
|
|
tokenizer = transformers.AutoTokenizer.from_pretrained(
|
|
args.model_name_or_path,
|
|
use_fast=False,
|
|
)
|
|
new_content = split_all(content, args.begin, args.end, tokenizer, args.max_length)
|
|
new_content = filter_invalid_roles(new_content)
|
|
|
|
print(f"total: {len(content)}, new: {len(new_content)}")
|
|
json.dump(new_content, open(args.out_file, "w"), indent=2)
|
|
|
|
|
|
if __name__ == "__main__":
|
|
parser = argparse.ArgumentParser()
|
|
parser.add_argument("--in-files", nargs="+", type=str)
|
|
parser.add_argument("--out-file", type=str, default="sharegpt_split.json")
|
|
parser.add_argument("--begin", type=int)
|
|
parser.add_argument("--end", type=int)
|
|
parser.add_argument("--model-name-or-path", type=str, required=True)
|
|
parser.add_argument("--max-length", type=int, default=4096)
|
|
args = parser.parse_args()
|
|
main(args) |