-
Notifications
You must be signed in to change notification settings - Fork 1
/
run.py
270 lines (224 loc) · 10.6 KB
/
run.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
257
258
259
260
261
262
263
264
265
266
267
268
269
270
import argparse
import os
from path_config import SAVEPATH, CACHEDIR
def tf_str(tf):
s = "true" if tf else "false"
return s
def run(args):
base_cmd = f"python -B -m src.train"
if "debug" not in args.model:
if args.load_path != '':
assert args.model in args.load_path, "Check model and load_path"
if args.eval_path != '':
assert args.train is False
assert args.model in args.eval_path, "Check model and eval_path"
else:
args.no_wandb = True
# Config file
model = args.model
base_cmd = f"{base_cmd} +{args.dataset}={model}"
base_cmd = f"{base_cmd} training.do_train={tf_str(args.train)}"
if args.dataset == 'metaicl':
base_cmd = f"{base_cmd} data.k={args.k}"
if args.dataset == 'lamp':
base_cmd = f"{base_cmd} data.dataset_name=lamp2"
base_cmd = f"{base_cmd} data.k={args.k}"
max_length = 1024
max_length += 256
if args.comp_type != "no":
max_length += args.n_tok * args.k
base_cmd = f"{base_cmd} training.generation_max_length={max_length}"
# Compression config
if args.attn_type == "gist":
args.comp_type = "fixed"
if args.comp_type in ["no", "neg_control"]:
base_cmd = f"{base_cmd} training.comp.add_comp_token=false"
args.cond_lora = args.sepembed = False
base_cmd = f"{base_cmd} training.comp.relative_embedding=base"
base_cmd = f"{base_cmd} training.comp.comp_type={args.comp_type}"
method_tag = args.comp_type
if args.comp_type not in ["no", "neg_control"]:
base_cmd = f"{base_cmd} training.comp.attn_type={args.attn_type}"
method_tag += f"-{args.attn_type}"
if args.eval_path != '':
for i in range(1, 17):
if (f"-ntok{i}" in args.eval_path):
args.n_tok = i
print(f"Update n_tok as {args.n_tok}")
if args.n_tok > 1:
base_cmd = f"{base_cmd} training.comp.num_comp_tokens={args.n_tok}"
method_tag += f"-ntok{args.n_tok}"
# Model config
model_tag = ''
if model.startswith("llama"):
if args.lora_r > 0:
base_cmd = f"{base_cmd} training.lora_r={args.lora_r}"
method_tag = f"r{args.lora_r}-{method_tag}"
if args.max_steps > 0:
st = args.max_steps
args.tag += f"_{st}"
base_cmd = f"{base_cmd} training.max_steps={st}"
base_cmd = f"{base_cmd} training.save_steps={st//4} training.eval_steps={st//2}"
if args.embed:
base_cmd = f"{base_cmd} training.finetune_embed=true"
method_tag = f"embed-{method_tag}"
if args.per_device_train_batch_size >= 1:
bs = args.per_device_train_batch_size
base_cmd = f"{base_cmd} training.per_device_train_batch_size={bs}"
base_cmd = f"{base_cmd} training.gradient_accumulation_steps={128//bs}"
args.tag += f"_batch{bs}"
if model.startswith('flan'):
if args.max_steps > 0:
st = args.max_steps
args.tag += f"_{st}"
base_cmd = f"{base_cmd} training.max_steps={st}"
base_cmd = f"{base_cmd} training.save_steps={st//2} training.eval_steps={st//2}"
if args.lr > 0:
args.tag += f"_lr{args.lr}"
base_cmd = f"{base_cmd} training.learning_rate={args.lr}"
# Conditional Lora
if not args.cond_lora: base_cmd = f"{base_cmd} training.comp.cond_lora=false"
if not args.sepembed: base_cmd = f"{base_cmd} training.comp.separate_embed=false"
if args.seed != 42:
base_cmd = f"{base_cmd} training.seed={args.seed}"
args.tag += f"_seed{args.seed}"
# Logging path
if args.pretrain_dataset is None:
wandb_group = wandb_group_out = args.dataset
else:
wandb_group = wandb_group_out = args.pretrain_dataset
args.tag += f"_{args.dataset}"
if model.startswith('llama') and args.max_length > 1024:
base_cmd = f"{base_cmd} data.max_length={args.max_length}"
wandb_group_out = f"{args.dataset}-len{args.max_length}"
generation_max_length = args.max_length + args.n_tok * args.k + 128
base_cmd = f"{base_cmd} training.generation_max_length={generation_max_length}"
if args.dataset in ["metaicl", "lamp"] and (args.k != 16 or args.eval_path != ''):
method_tag += f"-k{args.k}"
if args.tag != '':
method_tag += f"{args.tag}"
# Evaluation path
if args.eval_path != '':
subfolder = 'test'
training_output_dir = f"{SAVEPATH}/{wandb_group_out}/{subfolder}/{args.eval_path}"
training_output_dir += f"-{method_tag}"
wandb_name = f"{subfolder}-{os.path.basename(training_output_dir)}"
# Load pretrained model for evaluation
if args.eval_path.startswith("finetune"):
assert args.load_path != '', "Check args.load_path! (e.g., llama-7b-no)"
if args.load_path != '':
load_path = f"{SAVEPATH}/{wandb_group}/{args.load_path}"
base_cmd = f"{base_cmd} training.load_path={load_path}"
eval_path = f"{SAVEPATH}/{wandb_group}/{args.eval_path}"
base_cmd = f"{base_cmd} training.eval_path={eval_path}"
# Load pretrained model for finetuning
elif args.load_path != '':
subfolder = 'finetune'
training_output_dir = f"{SAVEPATH}/{wandb_group_out}/{subfolder}/{args.load_path}"
training_output_dir += f"{model_tag}-{method_tag}"
wandb_name = f"{subfolder}-{os.path.basename(training_output_dir)}"
load_path = f"{SAVEPATH}/{wandb_group}/{args.load_path}"
base_cmd = f"{base_cmd} training.load_path={load_path}"
else:
wandb_name = f"{model}{model_tag}-{method_tag}"
training_output_dir = f"{SAVEPATH}/{wandb_group_out}/{wandb_name}"
if "debug" in model:
wandb_name = "debug"
training_output_dir = f"{SAVEPATH}/{wandb_group_out}/{wandb_name}"
base_cmd = f"{base_cmd} wandb.group={wandb_group}"
base_cmd = f"{base_cmd} wandb.name={wandb_name}"
base_cmd = f"{base_cmd} training.output_dir={training_output_dir}"
base_cmd = f"{base_cmd} model.cache_dir={CACHEDIR}"
# No wandb for testing
if (args.no_wandb) or not args.train:
base_cmd = f"{base_cmd} wandb.log=false"
if args.override is not None:
override = " ".join(args.override.split(","))
base_cmd = f"{base_cmd} {override}"
cmd = base_cmd.split()
print(cmd)
os.execvp(cmd[0], cmd)
if __name__ == "__main__":
def str2bool(v):
if isinstance(v, bool):
return v
if v.lower() in ('yes', 'true', 't', 'y', '1'):
return True
elif v.lower() in ('no', 'false', 'f', 'n', '0'):
return False
else:
raise argparse.ArgumentTypeError('Boolean value expected.')
user = os.environ.get("USER", "janghyun")
parser = argparse.ArgumentParser()
parser.add_argument("--save_dir", type=str, default="", help="See path_config.py")
# Model
parser.add_argument("--model", "-m", default="llama-7b")
parser.add_argument("--sepembed",
type=str2bool,
default=True,
help="Train only embeddings for COMP tokens")
parser.add_argument("--embed",
action="store_true",
help="Train full embedding vectors during LoRA finetuning")
parser.add_argument("--cond_lora", type=str2bool, default=True, help="Conditional LoRA")
parser.add_argument("--lora_r",
"-r",
type=int,
default=-1,
help="LoRA rank size (default settings are in src/config)")
# Compression
parser.add_argument("--comp_type",
default="online",
choices=["no", "pos_control", "neg_control", "fixed", "online"],
help="Compression type")
## Notes on attention types ##
## 1. concat_recur/merge_recur refers to CCM-concat/-merge
## 2. concat/merge are the variants of CCM-concat/-merge.
## During compression, they do not attend the previous memory state, attending only to the current context.
## When generating outputs, they attend to the previous memory state.
## This strategy show slightly better performance in MetaICL.
## 3. gist refers to the Gisting compression method.
parser.add_argument("--attn_type",
default="concat_recur",
choices=["concat_recur", "merge_recur", "concat", "merge", "gist"],
help="Attention type")
parser.add_argument("--n_tok",
type=int,
default=1,
help="Number of COMP tokens for each context")
# Data
parser.add_argument("--dataset",
"-d",
default='metaicl',
choices=['all', 'metaicl', 'dialog', 'soda', 'lamp'],
help="The 'all' dataset refers to the mixture of MetaICL and SODA.")
parser.add_argument("--pretrain_dataset",
type=str,
default=None,
help="Train dataset. Use when it is differ from the evaluation dataset")
parser.add_argument("--k",
type=int,
default=16,
help="Max number of context time steps (metaicl, LaMP)")
parser.add_argument("--max_length", type=int, default=1024, help="Max length for data sample")
parser.add_argument("--generation_max_length",
type=int,
default=-1,
help="Generation max length")
# Training
parser.add_argument("--train", action="store_true", help="Conduct finetuning")
parser.add_argument("--max_steps", type=int, default=-1, help="Max finetuning steps")
parser.add_argument("--lr", type=float, default=-1, help="Learning rate")
parser.add_argument("--per_device_train_batch_size", '-b', type=int, default=-1)
parser.add_argument("--seed", type=int, default=42, help="Random seed")
# Eval
parser.add_argument("--load_path",
type=str,
default='',
help="Full-context finetuned adapter path (not required for LLaMA-2-chat)")
parser.add_argument("--eval_path", type=str, default='', help="Compression adapter path")
parser.add_argument("--no_wandb", action="store_true")
parser.add_argument("--override")
parser.add_argument("--tag", type=str, default='')
args = parser.parse_args()
run(args)