237 lines
8.2 KiB
Python
237 lines
8.2 KiB
Python
from typing import Dict, List, Callable, Union
|
|
from datasets import DatasetDict
|
|
from .tokenizer import BpeTokenizer
|
|
from tqdm import tqdm
|
|
from torch import Tensor
|
|
|
|
import pickle as pkl
|
|
import torch
|
|
import json
|
|
import os
|
|
import re
|
|
|
|
|
|
def fetch_files(directory):
|
|
return [os.path.join(root, f)
|
|
for root, _, files in os.walk(directory) for f in files]
|
|
|
|
|
|
def fetch_folders(root_dir, filter_func=None):
|
|
folders = []
|
|
for root, dirs, _ in os.walk(root_dir):
|
|
for dir_name in dirs:
|
|
folder_path = os.path.join(root, dir_name)
|
|
if filter_func is None or filter_func(folder_path):
|
|
folders.append(folder_path)
|
|
return folders
|
|
|
|
|
|
def comprehensive_normalization(text):
|
|
replacements = {
|
|
'\u2018': "'", '\u2019': "'", '\u0060': "'",
|
|
'\u201C': '"', '\u201D': '"',
|
|
'\u2013': '-', '\u2014': '--', '\u2212': '-',
|
|
'\u00A0': ' ',
|
|
'\u2026': '...'
|
|
}
|
|
pattern = re.compile('|'.join(re.escape(k) for k in replacements))
|
|
return pattern.sub(lambda m: replacements[m.group()], text)
|
|
|
|
|
|
def pack_sequences(sequences: List[Tensor], pack_size: int, pad_value: int) -> List[Tensor]:
|
|
packages = []
|
|
sequences.sort(key=lambda x: x.numel(), reverse=True)
|
|
current_pack = torch.tensor([], dtype=torch.int32)
|
|
|
|
for tensor in sequences:
|
|
if tensor.numel() > pack_size:
|
|
packages.append(tensor[:pack_size])
|
|
continue
|
|
|
|
remaining = pack_size - current_pack.numel()
|
|
|
|
if remaining == 0:
|
|
packages.append(current_pack)
|
|
current_pack = tensor
|
|
elif tensor.numel() <= remaining:
|
|
current_pack = torch.cat([current_pack, tensor])
|
|
else:
|
|
padding = torch.full((remaining,), pad_value, dtype=torch.int32)
|
|
current_pack = torch.cat([current_pack, padding])
|
|
packages.append(current_pack)
|
|
current_pack = tensor
|
|
|
|
if current_pack.numel() > 0:
|
|
if current_pack.numel() < pack_size:
|
|
padding = torch.full(
|
|
(pack_size - current_pack.numel(),),
|
|
pad_value,
|
|
dtype=torch.int32
|
|
)
|
|
current_pack = torch.cat([current_pack, padding])
|
|
else:
|
|
current_pack = current_pack[:pack_size]
|
|
|
|
packages.append(current_pack)
|
|
|
|
return packages
|
|
|
|
|
|
def dump_pkl_files(
|
|
files: List[str],
|
|
base_out_dir: str,
|
|
process_func: Callable[[dict], dict],
|
|
output_keys: List[str],
|
|
packing_size: int = -1,
|
|
pad_value: int = 0
|
|
):
|
|
|
|
for file_path in files:
|
|
out_file_name = os.path.basename(file_path).replace(".jsonl", ".pkl")
|
|
out_file_path = os.path.join(base_out_dir, out_file_name)
|
|
file_name = os.path.basename(file_path)
|
|
os.makedirs(os.path.dirname(out_file_path), exist_ok=True)
|
|
|
|
arrows: List[Dict[str, Tensor]] = []
|
|
|
|
with open(file_path, "r") as f:
|
|
lines = f.readlines()
|
|
|
|
for line in tqdm(lines, desc=f"Processing {file_name}", leave=False):
|
|
line_dict = json.loads(line)
|
|
arrow = process_func(line_dict)
|
|
arrows.append(arrow)
|
|
|
|
package: Dict[str, List[Tensor]] = {}
|
|
for key in output_keys:
|
|
list_tensor = [arrow[key] for arrow in arrows]
|
|
package[key] = list_tensor
|
|
|
|
output_package: Dict[str, Tensor] = {}
|
|
|
|
for key in output_keys:
|
|
if packing_size > 0:
|
|
print(f"Packaging key: '{key}'")
|
|
package[key] = pack_sequences(package[key], packing_size, pad_value)
|
|
sequence = torch.cat(package[key])
|
|
output_package[key] = sequence
|
|
|
|
with open(out_file_path, "wb") as f:
|
|
pkl.dump(output_package, f)
|
|
|
|
|
|
def get_pt_processor(tokenizer: BpeTokenizer):
|
|
def processor(intput_dict: dict) -> dict:
|
|
segment = intput_dict["text"]
|
|
ids = tokenizer.encode(f"{segment} <eos>")
|
|
t_ids = torch.tensor(ids, dtype=torch.int32)
|
|
|
|
return {'sequence': t_ids}
|
|
|
|
return processor
|
|
|
|
|
|
def get_sft_processor(tokenizer: BpeTokenizer):
|
|
def processor(input_dict: dict):
|
|
query, response = input_dict["query"], input_dict["response"]
|
|
prefix_seg = f"<|user|> {query} <|system|> <bos>"
|
|
suffix_seg = f"{response}<eos>\n"
|
|
prefix_ids = tokenizer.encode(prefix_seg)
|
|
suffix_ids = tokenizer.encode(suffix_seg)
|
|
|
|
tokens = prefix_ids + suffix_ids
|
|
tokens = torch.tensor(tokens, dtype=torch.int32)
|
|
masks = torch.zeros_like(tokens, dtype=torch.bool)
|
|
masks[len(prefix_ids):] = True
|
|
|
|
return {"sequence": tokens, "mask": masks}
|
|
|
|
return processor
|
|
|
|
def get_dpo_processor(tokenizer: BpeTokenizer):
|
|
def processor(input_dict: dict):
|
|
prompt, chosen, rejected = input_dict["prompt"], input_dict["chosen"], input_dict["rejected"]
|
|
prefix_seg = f"<|user|> {prompt} <|system|> <bos>"
|
|
chosen_seg = f"{chosen}<eos>\n"
|
|
rejected_seg = f"{rejected}<eos>\n"
|
|
prefix_ids = tokenizer.encode(prefix_seg)
|
|
chosen_ids = tokenizer.encode(chosen_seg)
|
|
rejected_ids = tokenizer.encode(rejected_seg)
|
|
|
|
chosen_seq = torch.tensor(prefix_ids + chosen_ids, dtype=torch.int32)
|
|
chosen_mask = torch.zeros_like(chosen_seq, dtype=torch.bool)
|
|
chosen_mask[len(prefix_ids):] = True
|
|
|
|
rejected_seq = torch.tensor(prefix_ids + rejected_ids, dtype=torch.int32)
|
|
resjected_mask = torch.zeros_like(rejected_seq, dtype=torch.bool)
|
|
resjected_mask[len(prefix_ids):] = True
|
|
|
|
return {"chosen": chosen_seq, "chosen_mask": chosen_mask, "rejected": rejected_seq, "rejected_mask": resjected_mask}
|
|
|
|
return processor
|
|
|
|
|
|
def cache_files(tokenizer, files, base_out_dir, cache_type):
|
|
processor = None
|
|
keys = []
|
|
if cache_type == "pt":
|
|
processor = get_pt_processor(tokenizer)
|
|
keys = ["text"]
|
|
elif cache_type == "sft":
|
|
processor = get_sft_processor(tokenizer)
|
|
keys = ["query", "response"]
|
|
elif cache_type == "dpo":
|
|
processor = get_dpo_processor(tokenizer)
|
|
keys = ["chosen", "chosen_mask", "rejected", "rejected_mask"]
|
|
else:
|
|
raise ValueError("Invalid cache type")
|
|
|
|
dump_pkl_files(files, base_out_dir, processor, keys)
|
|
|
|
|
|
def process_dataset(
|
|
dataset_dict: DatasetDict,
|
|
output_subdir: str,
|
|
max_chunk_num: int = None,
|
|
chunk_size: int = 1000000,
|
|
split_name: str = "train",
|
|
column_name: str = "text",
|
|
process_func: Union[Callable[[dict], dict], Callable[[List[dict]], List[dict]]] = None,
|
|
normalization_func=comprehensive_normalization,
|
|
output_dir: str = None,
|
|
):
|
|
train_dataset = dataset_dict[split_name]
|
|
total_samples = len(train_dataset)
|
|
num_chunks = (total_samples // chunk_size) + 1
|
|
lim_chunks = min(max_chunk_num, num_chunks) if max_chunk_num else num_chunks
|
|
|
|
if output_dir is None:
|
|
output_dir = os.path.join(os.getcwd(), "dataset", output_subdir)
|
|
|
|
os.makedirs(output_dir, exist_ok=True)
|
|
|
|
for i in range(lim_chunks):
|
|
start_idx = i * chunk_size
|
|
end_idx = min((i + 1) * chunk_size, total_samples)
|
|
chunk = train_dataset.select(range(start_idx, end_idx))
|
|
|
|
output_path = os.path.join(output_dir, f"{output_subdir}_text_chunk_{i}.jsonl")
|
|
with open(output_path, "w", encoding="utf-8") as f:
|
|
for example in chunk:
|
|
if process_func:
|
|
processed_example = process_func(example)
|
|
else:
|
|
text = example[column_name]
|
|
if normalization_func:
|
|
text = normalization_func(text)
|
|
processed_example = {column_name: text}
|
|
|
|
if isinstance(processed_example, dict):
|
|
f.write(json.dumps(processed_example, ensure_ascii=False) + "\n")
|
|
elif isinstance(processed_example, list):
|
|
for item in processed_example:
|
|
f.write(json.dumps(item, ensure_ascii=False) + "\n")
|
|
|
|
|
|
print(f"Saved text chunk {i} to {output_path}")
|