-
Notifications
You must be signed in to change notification settings - Fork 2
/
single_turn.py
256 lines (219 loc) · 7.96 KB
/
single_turn.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
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
import json
import os
import warnings
from argparse import ArgumentParser
from typing import Dict, List
from tqdm import tqdm
from src.data import HFDepthQALoader, filter_data_dict, slice_data_dict
from src.model import VLLM
from src.utils import (
SAMPLING_PARAMS,
SYSTEM_PROMPT_CTX,
SYSTEM_PROMPT_ZERO_SHOT,
USER_PROMPT_TEMPLATE_CTX,
USER_PROMPT_TEMPLATE_ZERO_SHOT,
get_output_path,
show_random_inputs,
)
DEBUG = False
# Model inference (Use offline batching)
def batch_completions(
model,
inputs: List[str],
batch_size,
):
batched_outputs = []
# Adjust batch size to fit the number of inputs
# VLLM supports adaptive batch size already
total_batches = len(inputs) // batch_size + (
1 if len(inputs) % batch_size > 0 else 0
)
total_len = len(inputs)
# Process initial batches with progress bar
print("Processing initial batches...")
for i in tqdm(
range(0, len(inputs), batch_size), total=total_batches, desc="Initial Batches"
):
batch_inputs = inputs[i : i + batch_size]
batch_outputs = model.completions(
batch_inputs, **SAMPLING_PARAMS, use_tqdm=True
)
batched_outputs.extend(batch_outputs)
# Final aggregation and printing
outputs_len = len(batched_outputs)
print(f"Processed {outputs_len}/{total_len} instances.")
if outputs_len < total_len:
warnings.warn("Some instances failed.")
warnings.warn("They will be written as None in the output file.")
raise Exception(
f"Failed to generate feedback for {total_len - outputs_len} instances."
)
for i, output in enumerate(batched_outputs):
if output == "":
print("Empty output")
batched_outputs[i] = None
if DEBUG:
print("Checking the results")
for output in batched_outputs[:5]:
print(output)
return batched_outputs
def apply_template_chat(system_prompt, user_prompt, tokenizer):
if tokenizer.chat_template and "system" not in tokenizer.chat_template:
messages = [
{"role": "user", "content": system_prompt + "\n" + user_prompt},
]
else:
messages = [
{"role": "system", "content": system_prompt},
{"role": "user", "content": user_prompt},
]
return (
tokenizer.apply_chat_template( # automatically format to default chat template
messages, tokenize=False, add_generation_prompt=True
)
)
def prepare_inputs_zero_shot(questions: Dict[str, Dict], tokenizer) -> List[str]:
inputs = []
system_prompt = SYSTEM_PROMPT_ZERO_SHOT
for question_d in questions.values():
target_question = question_d["question"]
user_prompt = USER_PROMPT_TEMPLATE_ZERO_SHOT.format(question=target_question)
input_str = apply_template_chat(system_prompt, user_prompt, tokenizer)
inputs.append(input_str)
return inputs
def prepare_inputs_ctx(
questions: Dict[str, Dict],
nodes: Dict[str, Dict],
node_to_q: Dict[str, str],
tokenizer,
use_gold: bool,
) -> List[str]:
inputs = []
system_prompt = SYSTEM_PROMPT_CTX
for nodeid, node in nodes.items():
target_question_d = questions[node_to_q[nodeid]]
target_question = target_question_d["question"]
predecessor_pairs = ""
for predec_nodeid in node["direct_predecessors"]:
predec_question_d = questions[node_to_q[predec_nodeid]]
predec_question = predec_question_d["question"]
if use_gold:
predec_answer = predec_question_d["answer"]
else:
predec_answer = predec_question_d["predicted_answer"]
pair = f"Q: {predec_question}\nA: {predec_answer}\n"
predecessor_pairs += pair
user_prompt = USER_PROMPT_TEMPLATE_CTX.format(
qa_pairs=predecessor_pairs, question=target_question
)
input_str = apply_template_chat(system_prompt, user_prompt, tokenizer)
inputs.append(input_str)
return inputs
def main(args):
global DEBUG
DEBUG = args.debug
# Load data
dataloader = HFDepthQALoader()
if args.task_type == "prompt-pred":
with open(args.input) as f:
questions = json.load(f)
_, nodes, node_to_q = dataloader.load_data(except_questions=True)
else:
questions, nodes, node_to_q = dataloader.load_data()
print(f"Loaded {len(questions)} questions and {len(nodes)} nodes.")
# Load model
model = VLLM(args.model_name, num_gpus=args.num_gpus)
tokenizer = model.get_tokenizer()
# Prepare inputs
if args.task_type == "zero-shot":
if DEBUG:
questions = slice_data_dict(questions, start=0, end=5)
inputs = prepare_inputs_zero_shot(questions, tokenizer)
else:
nodes = filter_data_dict(nodes, lambda node: node["depth"] > 1)
if DEBUG:
nodes = slice_data_dict(nodes, start=0, end=5)
inputs = prepare_inputs_ctx(
questions,
nodes,
node_to_q,
tokenizer,
use_gold=args.task_type == "prompt-gold",
)
show_random_inputs(inputs)
if DEBUG:
inputs = inputs[:5]
# Inference
predictions = batch_completions(model, inputs, args.batch_size)
# Save results
results = {}
if args.task_type == "zero-shot":
for idx, (qid, question_d) in enumerate(questions.items()):
results[qid] = question_d
results[qid].update({"predicted_answer": predictions[idx]})
else:
for idx, nodeid in enumerate(nodes.keys()):
results[nodeid] = questions[node_to_q[nodeid]]
results[nodeid].update({"predicted_answer": predictions[idx]})
output_path = get_output_path(
output_file=args.output_file, default_output_dir="../../outputs/inference"
)
with open(output_path, "w") as f:
json.dump(results, f, indent=4)
if __name__ == "__main__":
parser = ArgumentParser()
# I/O arguments
parser.add_argument(
"--model_name",
type=str,
required=True,
help="Name of model hosted in Hugging Face under AutoModelForCausalLM",
)
parser.add_argument(
"--input",
type=str,
default="kaist-ai/DepthQA",
help="Dataset name in Hugging Face (for zero-shot) or local zero-shot JSON output file (for prompt-*)",
)
parser.add_argument(
"--output_file",
type=str,
required=True,
help="Output JSON file name. Unless the parent directory is specified, will be saved under outputs/inference by default.",
)
parser.add_argument(
"--force_rerun",
action="store_true",
help="Force rerun even if output file exists.",
)
parser.add_argument("--debug", action="store_true", help="Debug mode.")
# Compute arguments
parser.add_argument(
"--batch_size", type=int, default=8, help="Batch size for inference."
)
parser.add_argument(
"--num_gpus",
type=int,
default=2,
help="Number of GPUs to use for inference. Note that we use bfloat16 if available and float16 otherwise.",
)
# Prompt arguments
parser.add_argument(
"--task_type",
type=str,
default="zero-shot",
choices=["zero-shot", "prompt-gold", "prompt-pred"],
help="Task type for the model, which determines the input text.",
)
args = parser.parse_args()
assert not (
args.task_type == "prompt-pred" and not args.input.endswith(".json")
), "Input file for prompt-pred task should be a JSON file that contains zero-shot predictions."
assert args.output_file.endswith(".json"), "Output file must be a JSON file."
output_path = get_output_path(
output_file=args.output_file, default_output_dir="../../outputs/inference"
)
assert not (
os.path.exists(output_path) and not args.force_rerun
), f"Output file {output_path} already exists. Skipping inference."
main(args)