lint
Browse files- gme_inference.py +70 -31
gme_inference.py
CHANGED
@@ -24,9 +24,10 @@ from transformers import (
|
|
24 |
import os
|
25 |
from collections.abc import Iterable
|
26 |
|
|
|
27 |
class GmeQwen2VLConfig(Qwen2VLConfig):
|
28 |
model_type: str = "gme_qwen2_vl"
|
29 |
-
|
30 |
def __init__(
|
31 |
self,
|
32 |
min_image_tokens: int = 256,
|
@@ -44,25 +45,27 @@ class GmeQwen2VLConfig(Qwen2VLConfig):
|
|
44 |
class GmeQwen2VLForVision2Seq(PreTrainedModel):
|
45 |
config_class = GmeQwen2VLConfig
|
46 |
base_model_prefix: str = "base"
|
47 |
-
|
48 |
def __init__(self, config: GmeQwen2VLConfig, **kwargs: Any) -> None:
|
49 |
super().__init__(config)
|
50 |
-
model_name: str = getattr(
|
51 |
-
|
|
|
|
|
52 |
self.base = Qwen2VLForConditionalGeneration(config)
|
53 |
self.normalize: bool = True
|
54 |
-
|
55 |
min_pixels: int = config.min_image_tokens * 28 * 28
|
56 |
max_pixels: int = config.max_image_tokens * 28 * 28
|
57 |
-
|
58 |
self.max_length: int = config.max_length
|
59 |
self.processor = AutoProcessor.from_pretrained(
|
60 |
model_name, min_pixels=min_pixels, max_pixels=max_pixels, **kwargs
|
61 |
)
|
62 |
-
self.processor.tokenizer.padding_side =
|
63 |
self.defualt_instruction: str = "You are a helpful assistant."
|
64 |
self.sep: str = " "
|
65 |
-
|
66 |
def forward(
|
67 |
self,
|
68 |
input_ids: Optional[torch.LongTensor] = None,
|
@@ -75,13 +78,15 @@ class GmeQwen2VLForVision2Seq(PreTrainedModel):
|
|
75 |
image_grid_thw: Optional[torch.LongTensor] = None,
|
76 |
# video_grid_thw: Optional[torch.LongTensor] = None,
|
77 |
pooling_mask: Optional[torch.LongTensor] = None,
|
78 |
-
**kwargs
|
79 |
) -> torch.Tensor:
|
80 |
if inputs_embeds is None:
|
81 |
inputs_embeds = self.base.model.embed_tokens(input_ids)
|
82 |
if pixel_values is not None:
|
83 |
pixel_values = pixel_values.type(self.base.visual.get_dtype())
|
84 |
-
image_embeds = self.base.visual(
|
|
|
|
|
85 |
image_mask = input_ids == self.base.config.image_token_id
|
86 |
inputs_embeds[image_mask] = image_embeds
|
87 |
# if pixel_values_videos is not None:
|
@@ -101,37 +106,44 @@ class GmeQwen2VLForVision2Seq(PreTrainedModel):
|
|
101 |
)
|
102 |
|
103 |
pooling_mask = attention_mask if pooling_mask is None else pooling_mask
|
104 |
-
left_padding =
|
105 |
if left_padding:
|
106 |
embeddings = outputs.last_hidden_state[:, -1]
|
107 |
else:
|
108 |
sequence_lengths = pooling_mask.sum(dim=1) - 1
|
109 |
batch_size = outputs.last_hidden_state.shape[0]
|
110 |
-
embeddings = outputs.last_hidden_state[
|
111 |
-
batch_size, device=outputs.last_hidden_state.device
|
112 |
-
|
|
|
113 |
if self.normalize:
|
114 |
embeddings = torch.nn.functional.normalize(embeddings, p=2, dim=1)
|
115 |
return embeddings.contiguous()
|
116 |
|
117 |
-
|
118 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
119 |
self.base.to(self.device)
|
120 |
# Inputs must be batched
|
121 |
input_texts, input_images = list(), list()
|
122 |
for t, i in zip(texts, images):
|
123 |
if not is_query or instruction is None:
|
124 |
instruction = self.defualt_instruction
|
125 |
-
input_str =
|
126 |
if i is None:
|
127 |
input_images = None # All examples in the same batch are consistent
|
128 |
else:
|
129 |
-
input_str +=
|
130 |
i = fetch_image(i)
|
131 |
input_images.append(i)
|
132 |
if t is not None:
|
133 |
input_str += t
|
134 |
-
msg = f
|
135 |
input_texts.append(msg)
|
136 |
|
137 |
inputs = self.processor(
|
@@ -140,7 +152,7 @@ class GmeQwen2VLForVision2Seq(PreTrainedModel):
|
|
140 |
padding=True,
|
141 |
truncation=True,
|
142 |
max_length=self.max_length,
|
143 |
-
return_tensors=
|
144 |
)
|
145 |
inputs = {k: v.to(self.device) for k, v in inputs.items()} # TODO
|
146 |
with torch.no_grad():
|
@@ -148,7 +160,9 @@ class GmeQwen2VLForVision2Seq(PreTrainedModel):
|
|
148 |
return embeddings
|
149 |
|
150 |
def encode(self, sentences: list[str], *, prompt_name=None, **kwargs):
|
151 |
-
return self.get_fused_embeddings(
|
|
|
|
|
152 |
|
153 |
def encode_queries(self, queries: List[str], **kwargs):
|
154 |
embeddings = self.encode(queries, **kwargs)
|
@@ -164,7 +178,9 @@ class GmeQwen2VLForVision2Seq(PreTrainedModel):
|
|
164 |
]
|
165 |
else:
|
166 |
sentences = [
|
167 |
-
(doc["title"] + self.sep + doc["text"]).strip()
|
|
|
|
|
168 |
for doc in corpus
|
169 |
]
|
170 |
embeddings = self.encode(sentences, is_query=False, **kwargs)
|
@@ -176,13 +192,18 @@ class GmeQwen2VLForVision2Seq(PreTrainedModel):
|
|
176 |
def get_text_embeddings(self, texts: list[str], **kwargs):
|
177 |
return self.get_fused_embeddings(texts=texts, **kwargs)
|
178 |
|
179 |
-
def get_fused_embeddings(
|
|
|
|
|
|
|
|
|
|
|
180 |
if isinstance(images, DataLoader):
|
181 |
image_loader = images
|
182 |
batch_size = image_loader.batch_size
|
183 |
image_loader.dataset.transform = None
|
184 |
else:
|
185 |
-
batch_size = kwargs.pop(
|
186 |
if images is None:
|
187 |
image_loader = None
|
188 |
else:
|
@@ -203,10 +224,18 @@ class GmeQwen2VLForVision2Seq(PreTrainedModel):
|
|
203 |
|
204 |
all_embeddings = list()
|
205 |
none_batch = [None] * batch_size
|
206 |
-
show_progress_bar = kwargs.pop(
|
207 |
-
pbar = tqdm(
|
208 |
-
|
209 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
210 |
img_batch = none_batch if img_batch is None else img_batch
|
211 |
embeddings = self.embed(texts=text_batch, images=img_batch, **kwargs)
|
212 |
pbar.update(1)
|
@@ -215,9 +244,11 @@ class GmeQwen2VLForVision2Seq(PreTrainedModel):
|
|
215 |
all_embeddings = torch.cat(all_embeddings, dim=0)
|
216 |
return all_embeddings
|
217 |
|
|
|
218 |
def custom_collate_fn(batch):
|
219 |
return batch
|
220 |
|
|
|
221 |
# Utility functions (copied from your vision processing code)
|
222 |
IMAGE_FACTOR: int = 28
|
223 |
MIN_PIXELS: int = 4 * 28 * 28
|
@@ -241,7 +272,11 @@ def floor_by_factor(number: int, factor: int) -> int:
|
|
241 |
|
242 |
|
243 |
def smart_resize(
|
244 |
-
height: int,
|
|
|
|
|
|
|
|
|
245 |
) -> tuple[int, int]:
|
246 |
"""
|
247 |
Rescales the image so that:
|
@@ -271,11 +306,15 @@ def smart_resize(
|
|
271 |
return h_bar, w_bar
|
272 |
|
273 |
|
274 |
-
def fetch_image(
|
|
|
|
|
275 |
image_obj: Optional[Image.Image] = None
|
276 |
if isinstance(image, Image.Image):
|
277 |
image_obj = image
|
278 |
-
elif isinstance(image, str) and (
|
|
|
|
|
279 |
image_obj = Image.open(requests.get(image, stream=True).raw)
|
280 |
elif isinstance(image, str) and image.startswith("file://"):
|
281 |
image_obj = Image.open(image[7:])
|
|
|
24 |
import os
|
25 |
from collections.abc import Iterable
|
26 |
|
27 |
+
|
28 |
class GmeQwen2VLConfig(Qwen2VLConfig):
|
29 |
model_type: str = "gme_qwen2_vl"
|
30 |
+
|
31 |
def __init__(
|
32 |
self,
|
33 |
min_image_tokens: int = 256,
|
|
|
45 |
class GmeQwen2VLForVision2Seq(PreTrainedModel):
|
46 |
config_class = GmeQwen2VLConfig
|
47 |
base_model_prefix: str = "base"
|
48 |
+
|
49 |
def __init__(self, config: GmeQwen2VLConfig, **kwargs: Any) -> None:
|
50 |
super().__init__(config)
|
51 |
+
model_name: str = getattr(
|
52 |
+
config, "_name_or_path", "Alibaba-NLP/gme-Qwen2-VL-2B-Instruct"
|
53 |
+
)
|
54 |
+
|
55 |
self.base = Qwen2VLForConditionalGeneration(config)
|
56 |
self.normalize: bool = True
|
57 |
+
|
58 |
min_pixels: int = config.min_image_tokens * 28 * 28
|
59 |
max_pixels: int = config.max_image_tokens * 28 * 28
|
60 |
+
|
61 |
self.max_length: int = config.max_length
|
62 |
self.processor = AutoProcessor.from_pretrained(
|
63 |
model_name, min_pixels=min_pixels, max_pixels=max_pixels, **kwargs
|
64 |
)
|
65 |
+
self.processor.tokenizer.padding_side = "right"
|
66 |
self.defualt_instruction: str = "You are a helpful assistant."
|
67 |
self.sep: str = " "
|
68 |
+
|
69 |
def forward(
|
70 |
self,
|
71 |
input_ids: Optional[torch.LongTensor] = None,
|
|
|
78 |
image_grid_thw: Optional[torch.LongTensor] = None,
|
79 |
# video_grid_thw: Optional[torch.LongTensor] = None,
|
80 |
pooling_mask: Optional[torch.LongTensor] = None,
|
81 |
+
**kwargs,
|
82 |
) -> torch.Tensor:
|
83 |
if inputs_embeds is None:
|
84 |
inputs_embeds = self.base.model.embed_tokens(input_ids)
|
85 |
if pixel_values is not None:
|
86 |
pixel_values = pixel_values.type(self.base.visual.get_dtype())
|
87 |
+
image_embeds = self.base.visual(
|
88 |
+
pixel_values, grid_thw=image_grid_thw
|
89 |
+
).to(inputs_embeds.device)
|
90 |
image_mask = input_ids == self.base.config.image_token_id
|
91 |
inputs_embeds[image_mask] = image_embeds
|
92 |
# if pixel_values_videos is not None:
|
|
|
106 |
)
|
107 |
|
108 |
pooling_mask = attention_mask if pooling_mask is None else pooling_mask
|
109 |
+
left_padding = pooling_mask[:, -1].sum() == pooling_mask.shape[0] # TODO
|
110 |
if left_padding:
|
111 |
embeddings = outputs.last_hidden_state[:, -1]
|
112 |
else:
|
113 |
sequence_lengths = pooling_mask.sum(dim=1) - 1
|
114 |
batch_size = outputs.last_hidden_state.shape[0]
|
115 |
+
embeddings = outputs.last_hidden_state[
|
116 |
+
torch.arange(batch_size, device=outputs.last_hidden_state.device),
|
117 |
+
sequence_lengths,
|
118 |
+
]
|
119 |
if self.normalize:
|
120 |
embeddings = torch.nn.functional.normalize(embeddings, p=2, dim=1)
|
121 |
return embeddings.contiguous()
|
122 |
|
123 |
+
def embed(
|
124 |
+
self,
|
125 |
+
texts: list[str],
|
126 |
+
images: list[Image.Image],
|
127 |
+
is_query=True,
|
128 |
+
instruction=None,
|
129 |
+
**kwargs,
|
130 |
+
):
|
131 |
self.base.to(self.device)
|
132 |
# Inputs must be batched
|
133 |
input_texts, input_images = list(), list()
|
134 |
for t, i in zip(texts, images):
|
135 |
if not is_query or instruction is None:
|
136 |
instruction = self.defualt_instruction
|
137 |
+
input_str = ""
|
138 |
if i is None:
|
139 |
input_images = None # All examples in the same batch are consistent
|
140 |
else:
|
141 |
+
input_str += "<|vision_start|><|image_pad|><|vision_end|>"
|
142 |
i = fetch_image(i)
|
143 |
input_images.append(i)
|
144 |
if t is not None:
|
145 |
input_str += t
|
146 |
+
msg = f"<|im_start|>system\n{instruction}<|im_end|>\n<|im_start|>user\n{input_str}<|im_end|>\n<|im_start|>assistant\n<|endoftext|>"
|
147 |
input_texts.append(msg)
|
148 |
|
149 |
inputs = self.processor(
|
|
|
152 |
padding=True,
|
153 |
truncation=True,
|
154 |
max_length=self.max_length,
|
155 |
+
return_tensors="pt",
|
156 |
)
|
157 |
inputs = {k: v.to(self.device) for k, v in inputs.items()} # TODO
|
158 |
with torch.no_grad():
|
|
|
160 |
return embeddings
|
161 |
|
162 |
def encode(self, sentences: list[str], *, prompt_name=None, **kwargs):
|
163 |
+
return self.get_fused_embeddings(
|
164 |
+
texts=sentences, prompt_name=prompt_name, **kwargs
|
165 |
+
)
|
166 |
|
167 |
def encode_queries(self, queries: List[str], **kwargs):
|
168 |
embeddings = self.encode(queries, **kwargs)
|
|
|
178 |
]
|
179 |
else:
|
180 |
sentences = [
|
181 |
+
(doc["title"] + self.sep + doc["text"]).strip()
|
182 |
+
if "title" in doc
|
183 |
+
else doc["text"].strip()
|
184 |
for doc in corpus
|
185 |
]
|
186 |
embeddings = self.encode(sentences, is_query=False, **kwargs)
|
|
|
192 |
def get_text_embeddings(self, texts: list[str], **kwargs):
|
193 |
return self.get_fused_embeddings(texts=texts, **kwargs)
|
194 |
|
195 |
+
def get_fused_embeddings(
|
196 |
+
self,
|
197 |
+
texts: list[str] = None,
|
198 |
+
images: list[Image.Image] | DataLoader = None,
|
199 |
+
**kwargs,
|
200 |
+
):
|
201 |
if isinstance(images, DataLoader):
|
202 |
image_loader = images
|
203 |
batch_size = image_loader.batch_size
|
204 |
image_loader.dataset.transform = None
|
205 |
else:
|
206 |
+
batch_size = kwargs.pop("batch_size", 32)
|
207 |
if images is None:
|
208 |
image_loader = None
|
209 |
else:
|
|
|
224 |
|
225 |
all_embeddings = list()
|
226 |
none_batch = [None] * batch_size
|
227 |
+
show_progress_bar = kwargs.pop("show_progress_bar", True)
|
228 |
+
pbar = tqdm(
|
229 |
+
total=n_batch,
|
230 |
+
disable=not show_progress_bar,
|
231 |
+
mininterval=1,
|
232 |
+
miniters=10,
|
233 |
+
desc="encode",
|
234 |
+
)
|
235 |
+
for n, img_batch in zip(
|
236 |
+
range(0, n_batch * batch_size, batch_size), image_loader
|
237 |
+
):
|
238 |
+
text_batch = none_batch if texts is None else texts[n : n + batch_size]
|
239 |
img_batch = none_batch if img_batch is None else img_batch
|
240 |
embeddings = self.embed(texts=text_batch, images=img_batch, **kwargs)
|
241 |
pbar.update(1)
|
|
|
244 |
all_embeddings = torch.cat(all_embeddings, dim=0)
|
245 |
return all_embeddings
|
246 |
|
247 |
+
|
248 |
def custom_collate_fn(batch):
|
249 |
return batch
|
250 |
|
251 |
+
|
252 |
# Utility functions (copied from your vision processing code)
|
253 |
IMAGE_FACTOR: int = 28
|
254 |
MIN_PIXELS: int = 4 * 28 * 28
|
|
|
272 |
|
273 |
|
274 |
def smart_resize(
|
275 |
+
height: int,
|
276 |
+
width: int,
|
277 |
+
factor: int = IMAGE_FACTOR,
|
278 |
+
min_pixels: int = MIN_PIXELS,
|
279 |
+
max_pixels: int = MAX_PIXELS,
|
280 |
) -> tuple[int, int]:
|
281 |
"""
|
282 |
Rescales the image so that:
|
|
|
306 |
return h_bar, w_bar
|
307 |
|
308 |
|
309 |
+
def fetch_image(
|
310 |
+
image: Union[str, Image.Image], size_factor: int = IMAGE_FACTOR
|
311 |
+
) -> Image.Image:
|
312 |
image_obj: Optional[Image.Image] = None
|
313 |
if isinstance(image, Image.Image):
|
314 |
image_obj = image
|
315 |
+
elif isinstance(image, str) and (
|
316 |
+
image.startswith("http://") or image.startswith("https://")
|
317 |
+
):
|
318 |
image_obj = Image.open(requests.get(image, stream=True).raw)
|
319 |
elif isinstance(image, str) and image.startswith("file://"):
|
320 |
image_obj = Image.open(image[7:])
|