Spaces:
Runtime error
Runtime error
Update convert.py
Browse files- convert.py +62 -117
convert.py
CHANGED
@@ -11,9 +11,7 @@ import torch
|
|
11 |
|
12 |
from huggingface_hub import CommitInfo, CommitOperationAdd, Discussion, HfApi, hf_hub_download
|
13 |
from huggingface_hub.file_download import repo_folder_name
|
14 |
-
from safetensors.torch import load_file, save_file
|
15 |
-
from transformers import AutoConfig
|
16 |
-
from transformers.pipelines.base import infer_framework_load_model
|
17 |
|
18 |
|
19 |
COMMIT_DESCRIPTION = """
|
@@ -40,30 +38,6 @@ class AlreadyExists(Exception):
|
|
40 |
pass
|
41 |
|
42 |
|
43 |
-
def shared_pointers(tensors):
|
44 |
-
ptrs = defaultdict(list)
|
45 |
-
for k, v in tensors.items():
|
46 |
-
ptrs[v.data_ptr()].append(k)
|
47 |
-
failing = []
|
48 |
-
for ptr, names in ptrs.items():
|
49 |
-
if len(names) > 1:
|
50 |
-
failing.append(names)
|
51 |
-
return failing
|
52 |
-
|
53 |
-
|
54 |
-
def check_file_size(sf_filename: str, pt_filename: str):
|
55 |
-
sf_size = os.stat(sf_filename).st_size
|
56 |
-
pt_size = os.stat(pt_filename).st_size
|
57 |
-
|
58 |
-
if (sf_size - pt_size) / pt_size > 0.01:
|
59 |
-
raise RuntimeError(
|
60 |
-
f"""The file size different is more than 1%:
|
61 |
-
- {sf_filename}: {sf_size}
|
62 |
-
- {pt_filename}: {pt_size}
|
63 |
-
"""
|
64 |
-
)
|
65 |
-
|
66 |
-
|
67 |
def rename(pt_filename: str) -> str:
|
68 |
filename, ext = os.path.splitext(pt_filename)
|
69 |
local = f"{filename}.safetensors"
|
@@ -77,29 +51,40 @@ def convert_multi(model_id: str, folder: str) -> ConversionResult:
|
|
77 |
data = json.load(f)
|
78 |
|
79 |
filenames = set(data["weight_map"].values())
|
80 |
-
local_filenames = []
|
81 |
-
for filename in filenames:
|
82 |
-
pt_filename = hf_hub_download(repo_id=model_id, filename=filename)
|
83 |
|
84 |
-
sf_filename = rename(pt_filename)
|
85 |
-
sf_filename = os.path.join(folder, sf_filename)
|
86 |
-
convert_file(pt_filename, sf_filename)
|
87 |
-
local_filenames.append(sf_filename)
|
88 |
|
|
|
89 |
index = os.path.join(folder, "model.safetensors.index.json")
|
90 |
with open(index, "w") as f:
|
91 |
newdata = {k: v for k, v in data.items()}
|
92 |
newmap = {k: rename(v) for k, v in data["weight_map"].items()}
|
93 |
newdata["weight_map"] = newmap
|
94 |
json.dump(newdata, f, indent=4)
|
95 |
-
local_filenames.append(index)
|
96 |
|
97 |
-
|
98 |
-
|
99 |
-
|
100 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
101 |
|
102 |
-
return operations, errors
|
103 |
|
104 |
|
105 |
def convert_single(model_id: str, folder: str) -> ConversionResult:
|
@@ -108,9 +93,15 @@ def convert_single(model_id: str, folder: str) -> ConversionResult:
|
|
108 |
sf_name = "model.safetensors"
|
109 |
sf_filename = os.path.join(folder, sf_name)
|
110 |
convert_file(pt_filename, sf_filename)
|
111 |
-
|
112 |
-
|
113 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
114 |
|
115 |
|
116 |
def convert_file(
|
@@ -120,18 +111,13 @@ def convert_file(
|
|
120 |
loaded = torch.load(pt_filename, map_location="cpu")
|
121 |
if "state_dict" in loaded:
|
122 |
loaded = loaded["state_dict"]
|
123 |
-
|
124 |
-
for shared_weights in shared:
|
125 |
-
for name in shared_weights[1:]:
|
126 |
-
loaded.pop(name)
|
127 |
-
|
128 |
# For tensors to be contiguous
|
129 |
loaded = {k: v.contiguous() for k, v in loaded.items()}
|
130 |
|
131 |
dirname = os.path.dirname(sf_filename)
|
132 |
os.makedirs(dirname, exist_ok=True)
|
133 |
save_file(loaded, sf_filename, metadata={"format": "pt"})
|
134 |
-
check_file_size(sf_filename, pt_filename)
|
135 |
reloaded = load_file(sf_filename)
|
136 |
for k in loaded:
|
137 |
pt_tensor = loaded[k]
|
@@ -156,56 +142,6 @@ def create_diff(pt_infos: Dict[str, List[str]], sf_infos: Dict[str, List[str]])
|
|
156 |
return "\n".join(errors)
|
157 |
|
158 |
|
159 |
-
def check_final_model(model_id: str, folder: str):
|
160 |
-
config = hf_hub_download(repo_id=model_id, filename="config.json")
|
161 |
-
shutil.copy(config, os.path.join(folder, "config.json"))
|
162 |
-
config = AutoConfig.from_pretrained(folder)
|
163 |
-
|
164 |
-
_, (pt_model, pt_infos) = infer_framework_load_model(model_id, config, output_loading_info=True)
|
165 |
-
_, (sf_model, sf_infos) = infer_framework_load_model(folder, config, output_loading_info=True)
|
166 |
-
|
167 |
-
if pt_infos != sf_infos:
|
168 |
-
error_string = create_diff(pt_infos, sf_infos)
|
169 |
-
raise ValueError(f"Different infos when reloading the model: {error_string}")
|
170 |
-
|
171 |
-
pt_params = pt_model.state_dict()
|
172 |
-
sf_params = sf_model.state_dict()
|
173 |
-
|
174 |
-
pt_shared = shared_pointers(pt_params)
|
175 |
-
sf_shared = shared_pointers(sf_params)
|
176 |
-
if pt_shared != sf_shared:
|
177 |
-
raise RuntimeError("The reconstructed model is wrong, shared tensors are different {shared_pt} != {shared_tf}")
|
178 |
-
|
179 |
-
sig = signature(pt_model.forward)
|
180 |
-
input_ids = torch.arange(10).unsqueeze(0)
|
181 |
-
pixel_values = torch.randn(1, 3, 224, 224)
|
182 |
-
input_values = torch.arange(1000).float().unsqueeze(0)
|
183 |
-
kwargs = {}
|
184 |
-
if "input_ids" in sig.parameters:
|
185 |
-
kwargs["input_ids"] = input_ids
|
186 |
-
if "decoder_input_ids" in sig.parameters:
|
187 |
-
kwargs["decoder_input_ids"] = input_ids
|
188 |
-
if "pixel_values" in sig.parameters:
|
189 |
-
kwargs["pixel_values"] = pixel_values
|
190 |
-
if "input_values" in sig.parameters:
|
191 |
-
kwargs["input_values"] = input_values
|
192 |
-
if "bbox" in sig.parameters:
|
193 |
-
kwargs["bbox"] = torch.zeros((1, 10, 4)).long()
|
194 |
-
if "image" in sig.parameters:
|
195 |
-
kwargs["image"] = pixel_values
|
196 |
-
|
197 |
-
if torch.cuda.is_available():
|
198 |
-
pt_model = pt_model.cuda()
|
199 |
-
sf_model = sf_model.cuda()
|
200 |
-
kwargs = {k: v.cuda() for k, v in kwargs.items()}
|
201 |
-
|
202 |
-
pt_logits = pt_model(**kwargs)[0]
|
203 |
-
sf_logits = sf_model(**kwargs)[0]
|
204 |
-
|
205 |
-
torch.testing.assert_close(sf_logits, pt_logits)
|
206 |
-
print(f"Model {model_id} is ok !")
|
207 |
-
|
208 |
-
|
209 |
def previous_pr(api: "HfApi", model_id: str, pr_title: str) -> Optional["Discussion"]:
|
210 |
try:
|
211 |
main_commit = api.list_repo_commits(model_id)[0].commit_id
|
@@ -226,6 +162,8 @@ def convert_generic(model_id: str, folder: str, filenames: Set[str]) -> Conversi
|
|
226 |
errors = []
|
227 |
|
228 |
extensions = set([".bin", ".ckpt"])
|
|
|
|
|
229 |
for filename in filenames:
|
230 |
prefix, ext = os.path.splitext(filename)
|
231 |
if ext in extensions:
|
@@ -240,10 +178,28 @@ def convert_generic(model_id: str, folder: str, filenames: Set[str]) -> Conversi
|
|
240 |
sf_filename = os.path.join(folder, sf_in_repo)
|
241 |
try:
|
242 |
convert_file(pt_filename, sf_filename)
|
243 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
244 |
except Exception as e:
|
245 |
errors.append((pt_filename, e))
|
246 |
-
return
|
247 |
|
248 |
|
249 |
def convert(api: "HfApi", model_id: str, force: bool = False) -> Tuple["CommitInfo", List["Exception"]]:
|
@@ -268,26 +224,15 @@ def convert(api: "HfApi", model_id: str, force: bool = False) -> Tuple["CommitIn
|
|
268 |
raise AlreadyExists(f"Model {model_id} already has an open PR check out {url}")
|
269 |
elif library_name == "transformers":
|
270 |
if "pytorch_model.bin" in filenames:
|
271 |
-
|
272 |
elif "pytorch_model.bin.index.json" in filenames:
|
273 |
-
|
274 |
else:
|
275 |
raise RuntimeError(f"Model {model_id} doesn't seem to be a valid pytorch model. Cannot convert")
|
276 |
-
check_final_model(model_id, folder)
|
277 |
-
else:
|
278 |
-
operations, errors = convert_generic(model_id, folder, filenames)
|
279 |
-
|
280 |
-
if operations:
|
281 |
-
new_pr = api.create_commit(
|
282 |
-
repo_id=model_id,
|
283 |
-
operations=operations,
|
284 |
-
commit_message=pr_title,
|
285 |
-
commit_description=COMMIT_DESCRIPTION,
|
286 |
-
create_pr=True,
|
287 |
-
)
|
288 |
-
print(f"Pr created at {new_pr.pr_url}")
|
289 |
else:
|
290 |
-
|
|
|
|
|
291 |
finally:
|
292 |
shutil.rmtree(folder)
|
293 |
return new_pr, errors
|
|
|
11 |
|
12 |
from huggingface_hub import CommitInfo, CommitOperationAdd, Discussion, HfApi, hf_hub_download
|
13 |
from huggingface_hub.file_download import repo_folder_name
|
14 |
+
from safetensors.torch import load_file, save_file, _remove_duplicate_names
|
|
|
|
|
15 |
|
16 |
|
17 |
COMMIT_DESCRIPTION = """
|
|
|
38 |
pass
|
39 |
|
40 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
41 |
def rename(pt_filename: str) -> str:
|
42 |
filename, ext = os.path.splitext(pt_filename)
|
43 |
local = f"{filename}.safetensors"
|
|
|
51 |
data = json.load(f)
|
52 |
|
53 |
filenames = set(data["weight_map"].values())
|
|
|
|
|
|
|
54 |
|
|
|
|
|
|
|
|
|
55 |
|
56 |
+
|
57 |
index = os.path.join(folder, "model.safetensors.index.json")
|
58 |
with open(index, "w") as f:
|
59 |
newdata = {k: v for k, v in data.items()}
|
60 |
newmap = {k: rename(v) for k, v in data["weight_map"].items()}
|
61 |
newdata["weight_map"] = newmap
|
62 |
json.dump(newdata, f, indent=4)
|
|
|
63 |
|
64 |
+
|
65 |
+
new_pr = api.create_commit(
|
66 |
+
repo_id=model_id,
|
67 |
+
operations=[CommitOperationAdd(path_in_repo=index.split("/")[-1], path_or_fileobj=index)],
|
68 |
+
commit_message=pr_title,
|
69 |
+
commit_description=COMMIT_DESCRIPTION,
|
70 |
+
create_pr=True,
|
71 |
+
)
|
72 |
+
|
73 |
+
for filename in filenames:
|
74 |
+
pt_filename = hf_hub_download(repo_id=model_id, filename=filename)
|
75 |
+
sf_filename = rename(pt_filename)
|
76 |
+
sf_filename = os.path.join(folder, sf_filename)
|
77 |
+
convert_file(pt_filename, sf_filename)
|
78 |
+
api.create_commit(
|
79 |
+
repo_id=repo_id,
|
80 |
+
commit_message=f"Adds {sf_filename}",
|
81 |
+
revision=new_pr.git_reference,
|
82 |
+
operations=[CommitOperationAdd(path_in_repo=sf_filename.split("/")[-1], path_or_fileobj=sf_filename)],
|
83 |
+
create_pr=False,
|
84 |
+
)
|
85 |
+
os.remove(pt_filename)
|
86 |
+
os.remove(sf_filename)
|
87 |
|
|
|
88 |
|
89 |
|
90 |
def convert_single(model_id: str, folder: str) -> ConversionResult:
|
|
|
93 |
sf_name = "model.safetensors"
|
94 |
sf_filename = os.path.join(folder, sf_name)
|
95 |
convert_file(pt_filename, sf_filename)
|
96 |
+
|
97 |
+
new_pr = api.create_commit(
|
98 |
+
repo_id=model_id,
|
99 |
+
operations=[CommitOperationAdd(path_in_repo=sf_name, path_or_fileobj=sf_filename)],
|
100 |
+
commit_message=pr_title,
|
101 |
+
commit_description=COMMIT_DESCRIPTION,
|
102 |
+
create_pr=True,
|
103 |
+
)
|
104 |
+
return new_pr
|
105 |
|
106 |
|
107 |
def convert_file(
|
|
|
111 |
loaded = torch.load(pt_filename, map_location="cpu")
|
112 |
if "state_dict" in loaded:
|
113 |
loaded = loaded["state_dict"]
|
114 |
+
loaded = _remove_duplicate_names(loaded)
|
|
|
|
|
|
|
|
|
115 |
# For tensors to be contiguous
|
116 |
loaded = {k: v.contiguous() for k, v in loaded.items()}
|
117 |
|
118 |
dirname = os.path.dirname(sf_filename)
|
119 |
os.makedirs(dirname, exist_ok=True)
|
120 |
save_file(loaded, sf_filename, metadata={"format": "pt"})
|
|
|
121 |
reloaded = load_file(sf_filename)
|
122 |
for k in loaded:
|
123 |
pt_tensor = loaded[k]
|
|
|
142 |
return "\n".join(errors)
|
143 |
|
144 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
145 |
def previous_pr(api: "HfApi", model_id: str, pr_title: str) -> Optional["Discussion"]:
|
146 |
try:
|
147 |
main_commit = api.list_repo_commits(model_id)[0].commit_id
|
|
|
162 |
errors = []
|
163 |
|
164 |
extensions = set([".bin", ".ckpt"])
|
165 |
+
|
166 |
+
new_pr = None
|
167 |
for filename in filenames:
|
168 |
prefix, ext = os.path.splitext(filename)
|
169 |
if ext in extensions:
|
|
|
178 |
sf_filename = os.path.join(folder, sf_in_repo)
|
179 |
try:
|
180 |
convert_file(pt_filename, sf_filename)
|
181 |
+
|
182 |
+
if new_pr is None:
|
183 |
+
new_pr = api.create_commit(
|
184 |
+
repo_id=model_id,
|
185 |
+
operations=[CommitOperationAdd(path_in_repo=sf_in_repo, path_or_fileobj=sf_filename)],
|
186 |
+
commit_message=pr_title,
|
187 |
+
commit_description=COMMIT_DESCRIPTION,
|
188 |
+
create_pr=True,
|
189 |
+
)
|
190 |
+
else:
|
191 |
+
api.create_commit(
|
192 |
+
repo_id=repo_id,
|
193 |
+
commit_message=f"Adds {sf_filename}",
|
194 |
+
revision=new_pr.git_reference,
|
195 |
+
operations=[CommitOperationAdd(path_in_repo=sf_in_repo, path_or_fileobj=sf_filename)],
|
196 |
+
create_pr=False,
|
197 |
+
)
|
198 |
+
os.remove(pt_filename)
|
199 |
+
os.remove(sf_filename)
|
200 |
except Exception as e:
|
201 |
errors.append((pt_filename, e))
|
202 |
+
return new_pr
|
203 |
|
204 |
|
205 |
def convert(api: "HfApi", model_id: str, force: bool = False) -> Tuple["CommitInfo", List["Exception"]]:
|
|
|
224 |
raise AlreadyExists(f"Model {model_id} already has an open PR check out {url}")
|
225 |
elif library_name == "transformers":
|
226 |
if "pytorch_model.bin" in filenames:
|
227 |
+
new_pr = convert_single(model_id, folder)
|
228 |
elif "pytorch_model.bin.index.json" in filenames:
|
229 |
+
new_pr = convert_multi(model_id, folder)
|
230 |
else:
|
231 |
raise RuntimeError(f"Model {model_id} doesn't seem to be a valid pytorch model. Cannot convert")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
232 |
else:
|
233 |
+
new_pr = convert_generic(model_id, folder, filenames)
|
234 |
+
|
235 |
+
print(f"Pr created at {new_pr.pr_url}")
|
236 |
finally:
|
237 |
shutil.rmtree(folder)
|
238 |
return new_pr, errors
|