-
Notifications
You must be signed in to change notification settings - Fork 3
/
convert_bart_original_pytorch_checkpoint_to_pytorch.py
143 lines (120 loc) · 5.37 KB
/
convert_bart_original_pytorch_checkpoint_to_pytorch.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
# coding=utf-8
# Copyright 2020 The HuggingFace Inc. team.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""Convert BART checkpoint."""
import argparse
import os
from pathlib import Path
import fairseq
import torch
from packaging import version
from transformers import (
BartConfig,
BartForConditionalGeneration,
BartForSequenceClassification,
BartModel,
BartTokenizer,
)
from transformers.models.bart.modeling_bart import _make_linear_from_emb
from transformers.utils import logging
FAIRSEQ_MODELS = ["bart.large", "bart.large.mnli", "bart.large.cnn", "bart_xsum/model.pt"]
extra_arch = {"bart.large": BartModel, "bart.large.mnli": BartForSequenceClassification}
if version.parse(fairseq.__version__) < version.parse("0.9.0"):
raise Exception("requires fairseq >= 0.9.0")
logging.set_verbosity_info()
logger = logging.get_logger(__name__)
SAMPLE_TEXT = " Hello world! cécé herlolip"
mnli_rename_keys = [
("model.classification_heads.mnli.dense.weight", "classification_head.dense.weight"),
("model.classification_heads.mnli.dense.bias", "classification_head.dense.bias"),
("model.classification_heads.mnli.out_proj.weight", "classification_head.out_proj.weight"),
("model.classification_heads.mnli.out_proj.bias", "classification_head.out_proj.bias"),
]
def remove_ignore_keys_(state_dict):
ignore_keys = [
"encoder.version",
"decoder.version",
"model.encoder.version",
"model.decoder.version",
"_float_tensor",
]
for k in ignore_keys:
state_dict.pop(k, None)
def rename_key(dct, old, new):
val = dct.pop(old)
dct[new] = val
def load_xsum_checkpoint(checkpoint_path):
"""Checkpoint path should end in model.pt"""
sd = torch.load(checkpoint_path, map_location="cpu")
hub_interface = torch.hub.load("pytorch/fairseq", "bart.large.cnn").eval()
hub_interface.model.load_state_dict(sd["model"])
return hub_interface
@torch.no_grad()
def convert_bart_checkpoint(checkpoint_path, pytorch_dump_folder_path, hf_checkpoint_name=None):
"""
Copy/paste/tweak model's weights to our BERT structure.
"""
if not os.path.exists(checkpoint_path):
bart = torch.hub.load("pytorch/fairseq", checkpoint_path).eval()
else:
bart = load_xsum_checkpoint(checkpoint_path)
bart.model.upgrade_state_dict(bart.model.state_dict())
if hf_checkpoint_name is None:
hf_checkpoint_name = checkpoint_path.replace(".", "-")
config = BartConfig.from_pretrained(hf_checkpoint_name)
tokens = bart.encode(SAMPLE_TEXT).unsqueeze(0)
tokens2 = BartTokenizer.from_pretrained(hf_checkpoint_name).encode(SAMPLE_TEXT, return_tensors="pt").unsqueeze(0)
assert torch.eq(tokens, tokens2).all()
if checkpoint_path == "bart.large.mnli":
state_dict = bart.state_dict()
remove_ignore_keys_(state_dict)
state_dict["model.shared.weight"] = state_dict["model.decoder.embed_tokens.weight"]
for src, dest in mnli_rename_keys:
rename_key(state_dict, src, dest)
model = BartForSequenceClassification(config).eval()
model.load_state_dict(state_dict)
fairseq_output = bart.predict("mnli", tokens, return_logits=True)
new_model_outputs = model(tokens)[0] # logits
else: # no classification heads to worry about
state_dict = bart.model.state_dict()
remove_ignore_keys_(state_dict)
state_dict["shared.weight"] = state_dict["decoder.embed_tokens.weight"]
fairseq_output = bart.extract_features(tokens)
if hf_checkpoint_name == "facebook/bart-large":
model = BartModel(config).eval()
model.load_state_dict(state_dict)
new_model_outputs = model(tokens).model[0]
else:
model = BartForConditionalGeneration(config).eval() # an existing summarization ckpt
model.model.load_state_dict(state_dict)
if hasattr(model, "lm_head"):
model.lm_head = _make_linear_from_emb(model.model.shared)
new_model_outputs = model.model(tokens)[0]
# Check results
assert fairseq_output.shape == new_model_outputs.shape
assert (fairseq_output == new_model_outputs).all().item()
Path(pytorch_dump_folder_path).mkdir(exist_ok=True)
model.save_pretrained(pytorch_dump_folder_path)
if __name__ == "__main__":
parser = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
"fairseq_path", type=str, help="bart.large, bart.large.cnn or a path to a model.pt on local filesystem."
)
parser.add_argument("pytorch_dump_folder_path", default=None, type=str, help="Path to the output PyTorch model.")
parser.add_argument(
"--hf_config", default=None, type=str, help="Which huggingface architecture to use: bart-large-xsum"
)
args = parser.parse_args()
convert_bart_checkpoint(args.fairseq_path, args.pytorch_dump_folder_path, hf_checkpoint_name=args.hf_config)