/
llama_official.py
160 lines (128 loc) · 5.6 KB
/
llama_official.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
# Copyright (c) Meta Platforms, Inc. and affiliates.
# This software may be used and distributed according to the terms of the GNU General Public License version 3.
from typing import Tuple
import os
import sys
import pdb
import torch
import fire
import time
import json
from pathlib import Path
from fairscale.nn.model_parallel.initialize import initialize_model_parallel
from llama import ModelArgs, Transformer, Tokenizer, LLaMA
from argparse import ArgumentParser
import tqdm
from tools import DynamicDataset, read_jsonlines
import jsonlines
from torch.utils.data import DataLoader
def define_hps():
parser = ArgumentParser(description='LLaMA')
parser.add_argument('--mode', type=str, default='few_shot')
parser.add_argument('--task', type=str, default='copa')
parser.add_argument('--data_dir', type=str, default='./data/jsonlines')
parser.add_argument('--checkpoint', type=str, default='/ssd1/huggingface_transformers/llama/')
parser.add_argument('--model_size', type=str, default='7B')
parser.add_argument('--tokenizer', type=str, default='/ssd1/huggingface_transformers/llama/tokenizer.model')
parser.add_argument('--prompt_dir', type=str, default='./data/prompts')
parser.add_argument('--log_dir', type=str, default='./output/logger')
parser.add_argument('--log_name', type=str, default='')
parser.add_argument('--output_dir', type=str, default='./output/few_shot_LLaMA')
parser.add_argument('--max_seq_len', type=int, default=650)
parser.add_argument('--max_tokens', type=int, default=20)
parser.add_argument('--batch_size', type=int, default=10)
parser.add_argument('--temperature', type=float, default=1.0)
parser.add_argument('--top_p', type=float, default=1.0)
hps = parser.parse_args()
return hps
def setup_model_parallel() -> Tuple[int, int]:
local_rank = int(os.environ.get("LOCAL_RANK", -1))
world_size = int(os.environ.get("WORLD_SIZE", -1))
torch.distributed.init_process_group("nccl")
initialize_model_parallel(world_size)
torch.cuda.set_device(local_rank)
# seed must be the same in all processes
torch.manual_seed(1)
return local_rank, world_size
def load(
ckpt_dir: str,
tokenizer_path: str,
local_rank: int,
world_size: int,
max_seq_len: int,
max_batch_size: int,
) -> LLaMA:
start_time = time.time()
checkpoints = sorted(Path(ckpt_dir).glob("*.pth"))
assert world_size == len(
checkpoints
), f"Loading a checkpoint for MP={len(checkpoints)} but world size is {world_size}"
ckpt_path = checkpoints[local_rank]
print("Loading")
checkpoint = torch.load(ckpt_path, map_location="cpu")
with open(Path(ckpt_dir) / "params.json", "r") as f:
params = json.loads(f.read())
model_args: ModelArgs = ModelArgs(
max_seq_len=max_seq_len, max_batch_size=max_batch_size, **params
)
tokenizer = Tokenizer(model_path=tokenizer_path)
model_args.vocab_size = tokenizer.n_words
torch.set_default_tensor_type(torch.cuda.HalfTensor)
model = Transformer(model_args)
torch.set_default_tensor_type(torch.FloatTensor)
model.load_state_dict(checkpoint, strict=False)
generator = LLaMA(model, tokenizer)
print(f"Loaded in {time.time() - start_time:.2f} seconds")
return generator
def main():
parser = ArgumentParser()
parser.add_argument("--task", type=str, default="anli")
opt = parser.parse_args()
checkpoint = "/ssd1/huggingface_transformers/llama/"
model_size = '13B_2'
# task = 'anli'
task = opt.task
batch_size = 24
max_seq_len = 1224
data_dir = "./data/jsonlines"
prompt_dir = "./data/prompts"
output_dir = "./output/LLaMA/13B"
max_tokens = 256
temperature = 0.0
top_p = 1
print(f"[Model]: {model_size}")
local_rank, world_size = setup_model_parallel()
if local_rank > 0:
sys.stdout = open(os.devnull, "w")
generator = load(os.path.join(checkpoint, model_size), os.path.join(checkpoint, 'tokenizer.model'), local_rank, world_size, max_seq_len, batch_size)
data = read_jsonlines(os.path.join(data_dir, '{}.jsonl'.format(task)))
DATA = DynamicDataset(*data)
dataloader = DataLoader(DATA, batch_size=batch_size)
if task in ['singleeq', 'multiarith', 'svamp', 'addsub', 'gsm8k']:
prompt = open(os.path.join(prompt_dir, 'manual_cot_{}.txt'.format('mwp')), 'r').read()
else:
prompt = open(os.path.join(prompt_dir, 'manual_cot_{}.txt'.format(task)), 'r').read()
if not os.path.exists(os.path.join(output_dir)):
os.makedirs(os.path.join(output_dir))
bar = tqdm.trange(len(dataloader))
fo = jsonlines.open(os.path.join(output_dir, '{}.jsonl'.format(task)), 'w')
bar.set_postfix()
for batch, _ in zip(dataloader, bar):
questions, options, answers, responses = batch
if task != 'strategyqa':
overall_prompt = ["{}\n\nQ: {} Answer Choices: {}\nA:".format(prompt, q, o) for q, o in zip(questions, options)]
else:
overall_prompt = ["{}\n\nQ: {}\nA:".format(prompt, q) for q in questions]
results = generator.generate(overall_prompt, max_gen_len=max_tokens, temperature=temperature, top_p=top_p)
for r, q, o, a, _, p in zip(results, *batch, overall_prompt):
# pdb.set_trace()
r = r[len(p):].strip().split('\n\n')[0]
# r = r[len(p):]
if isinstance(a, torch.Tensor):
fo.write({'Q': q, 'A': a.item(), 'O': o, 'R': {'text': r}})
else:
fo.write({'Q': q, 'A': a, 'O': o, 'R': {'text': r}})
fo.close()
del generator
if __name__ == "__main__":
fire.Fire(main)