forked from meta-llama/llama
-
Notifications
You must be signed in to change notification settings - Fork 0
/
example_ft.py
174 lines (145 loc) · 5.03 KB
/
example_ft.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
161
162
163
164
165
166
167
168
169
170
171
172
173
174
# 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 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,
PromptEmbedding,
PromptTuningConfig,
)
from dataclasses import dataclass, field, asdict
PROMPT_DICT = {
"prompt_input": (
"Below is an instruction that describes a task, paired with an input that provides further context. "
"Write a response that appropriately completes the request.\n\n"
"### Instruction:\n{instruction}\n\n### Input:\n{input}\n\n### Response:"
),
"prompt_no_input": (
# "Below is an instruction that describes a task. "
# "Write a response that appropriately completes the request.\n\n"
"### Instruction:\n{instruction}\n\n### Response:"
),
}
@dataclass
class FTParams:
# Fine-Tuning Model Params
num_virtual_tokens: int = field(default=32)
num_transformer_submodules: int = field(default=1)
peft_type: str = field(default="PROMPT_TUNING")
task_type: str = field(default="CAUSAL_LM")
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))
print(local_rank, world_size)
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,
tuning_ckpt,
) -> 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)
kwargs = asdict(FTParams())
kwargs["token_dim"] = params["dim"]
peft_config = PromptTuningConfig(**kwargs)
prompt_encoder = PromptEmbedding(peft_config, is_inference=True)
prompt_encoder.load_state_dict(tuning_ckpt)
generator = LLaMA(model, tokenizer, prompt_encoder)
print(f"Loaded in {time.time() - start_time:.2f} seconds")
return generator
def main(
ckpt_dir: str,
tuning_ckpt_dir: str,
tokenizer_path: str,
temperature: float = 0.8,
top_p: float = 0.95,
max_seq_len: int = 512,
max_batch_size: int = 32,
):
local_rank, world_size = setup_model_parallel()
if local_rank > 0:
sys.stdout = open(os.devnull, "w")
tuning_ckpt = torch.load(tuning_ckpt_dir, map_location="cpu")
# If Fine-Tuning by HuggingFace
if "prompt_embeddings" in tuning_ckpt:
tuning_ckpt["embedding.weight"] = tuning_ckpt["prompt_embeddings"]
del tuning_ckpt["prompt_embeddings"]
generator = load(
ckpt_dir,
tokenizer_path,
local_rank,
world_size,
max_seq_len,
max_batch_size,
tuning_ckpt,
)
prompts = [
# For these prompts, the expected answer is the natural continuation of the prompt
"I believe the meaning of life is",
"Simply put, the theory of relativity states that ",
"Building a website can be done in 10 simple steps:\n",
# Few shot prompts: https://huggingface.co/blog/few-shot-learning-gpt-neo-and-inference-api
"""Tweet: "I hate it when my phone battery dies."
Sentiment: Negative
###
Tweet: "My day has been 👍"
Sentiment: Positive
###
Tweet: "This is the link to the article"
Sentiment: Neutral
###
Tweet: "This new music video was incredibile"
Sentiment:""",
"""Translate English to French:
sea otter => loutre de mer
peppermint => menthe poivrée
plush girafe => girafe peluche
cheese =>""",
PROMPT_DICT["prompt_no_input"].format_map(
{"instruction": "Give three tips for staying healthy."}
),
]
results = generator.generate(
prompts, max_gen_len=256, temperature=temperature, top_p=top_p
)
for prompt, result in zip(prompts, results):
print(prompt, result)
print("\n==================================\n")
if __name__ == "__main__":
fire.Fire(main)