support device type `cpu` generate
#26
by
weege007
- opened
- modeling_GOT.py +200 -134
modeling_GOT.py
CHANGED
@@ -1,3 +1,4 @@
|
|
|
|
1 |
from transformers import Qwen2Config, Qwen2Model, Qwen2ForCausalLM, StoppingCriteria, TextStreamer
|
2 |
from transformers.modeling_outputs import BaseModelOutputWithPast, CausalLMOutputWithPast
|
3 |
from typing import List, Optional, Tuple, Union
|
@@ -19,7 +20,7 @@ DEFAULT_IMAGE_PATCH_TOKEN = '<imgpad>'
|
|
19 |
DEFAULT_IM_START_TOKEN = '<img>'
|
20 |
DEFAULT_IM_END_TOKEN = '</img>'
|
21 |
|
22 |
-
|
23 |
class SeparatorStyle(Enum):
|
24 |
"""Different separator style."""
|
25 |
SINGLE = auto()
|
@@ -65,7 +66,7 @@ class Conversation:
|
|
65 |
return ret
|
66 |
if self.sep_style == SeparatorStyle.MPT:
|
67 |
if self.system:
|
68 |
-
ret = self.system + self.sep
|
69 |
else:
|
70 |
ret = ''
|
71 |
for role, message in self.messages:
|
@@ -79,7 +80,6 @@ class Conversation:
|
|
79 |
else:
|
80 |
raise ValueError(f"Invalid style: {self.sep_style}")
|
81 |
|
82 |
-
|
83 |
def append_message(self, role, message):
|
84 |
self.messages.append([role, message])
|
85 |
|
@@ -94,12 +94,12 @@ class Conversation:
|
|
94 |
sep2=self.sep2)
|
95 |
|
96 |
|
97 |
-
|
98 |
class KeywordsStoppingCriteria(StoppingCriteria):
|
99 |
def __init__(self, keywords, tokenizer, input_ids):
|
100 |
self.keywords = keywords
|
101 |
self.keyword_ids = [tokenizer(keyword).input_ids for keyword in keywords]
|
102 |
-
self.keyword_ids = [keyword_id[0] for keyword_id in self.keyword_ids if type(
|
|
|
103 |
self.tokenizer = tokenizer
|
104 |
self.start_len = None
|
105 |
self.input_ids = input_ids
|
@@ -111,12 +111,13 @@ class KeywordsStoppingCriteria(StoppingCriteria):
|
|
111 |
for keyword_id in self.keyword_ids:
|
112 |
if output_ids[0, -1] == keyword_id:
|
113 |
return True
|
114 |
-
outputs = self.tokenizer.batch_decode(
|
|
|
115 |
for keyword in self.keywords:
|
116 |
if keyword in outputs:
|
117 |
return True
|
118 |
return False
|
119 |
-
|
120 |
|
121 |
class GOTImageEvalProcessor:
|
122 |
def __init__(self, image_size=384, mean=None, std=None):
|
@@ -136,11 +137,11 @@ class GOTImageEvalProcessor:
|
|
136 |
self.normalize,
|
137 |
]
|
138 |
)
|
|
|
139 |
def __call__(self, item):
|
140 |
return self.transform(item)
|
141 |
|
142 |
|
143 |
-
|
144 |
class GOTConfig(Qwen2Config):
|
145 |
model_type = "GOT"
|
146 |
|
@@ -153,28 +154,25 @@ class GOTQwenModel(Qwen2Model):
|
|
153 |
|
154 |
self.vision_tower_high = build_GOT_vit_b()
|
155 |
|
156 |
-
self.mm_projector_vary =
|
157 |
-
|
158 |
|
159 |
def initialize_vision_modules(
|
160 |
-
self,
|
161 |
vision_tower,
|
162 |
pretrained_stage1_model=None,
|
163 |
freeze_vision_tower=False,
|
164 |
use_im_start_end=False,
|
165 |
vision_select_layer=-1,
|
166 |
dtype=torch.float16,
|
167 |
-
device="cuda"
|
168 |
):
|
169 |
|
170 |
-
|
171 |
image_processor_high = GOTImageEvalProcessor(image_size=1024)
|
172 |
-
|
173 |
self.vision_tower_high = self.vision_tower_high.to(dtype=dtype, device=device)
|
174 |
|
175 |
self.mm_projector_vary = self.mm_projector_vary.to(dtype=dtype, device=device)
|
176 |
|
177 |
-
|
178 |
image_token_len = 256
|
179 |
|
180 |
self.config.vision_tower = vision_tower
|
@@ -184,13 +182,12 @@ class GOTQwenModel(Qwen2Model):
|
|
184 |
|
185 |
self.config.vision_select_layer = vision_select_layer
|
186 |
self.config.freeze_vision_tower = freeze_vision_tower
|
187 |
-
|
188 |
return dict(
|
189 |
image_processor_high=image_processor_high,
|
190 |
image_token_len=image_token_len,
|
191 |
)
|
192 |
-
|
193 |
-
|
194 |
def forward(
|
195 |
self,
|
196 |
input_ids: torch.LongTensor = None,
|
@@ -209,16 +206,17 @@ class GOTQwenModel(Qwen2Model):
|
|
209 |
orig_embeds_params = getattr(self, 'orig_embeds_params', None)
|
210 |
if orig_embeds_params is not None:
|
211 |
with torch.no_grad():
|
212 |
-
self.get_input_embeddings().weight[:-
|
|
|
|
|
213 |
|
214 |
if inputs_embeds is None:
|
215 |
inputs_embeds = self.embed_tokens(input_ids)
|
216 |
|
217 |
-
|
218 |
vision_tower_high = getattr(self, 'vision_tower_high', None)
|
219 |
|
220 |
-
|
221 |
-
|
222 |
use_im_start_end = getattr(self.config, "use_im_start_end", -1)
|
223 |
|
224 |
vision_select_layer = getattr(self.config, "vision_select_layer", -1)
|
@@ -232,15 +230,15 @@ class GOTQwenModel(Qwen2Model):
|
|
232 |
im_start_token = 151857
|
233 |
|
234 |
im_end_token = 151858
|
235 |
-
|
236 |
image_features = []
|
237 |
-
|
238 |
for image in images:
|
239 |
P, C, H, W = image.shape
|
240 |
if P == 1:
|
241 |
with torch.set_grad_enabled(False):
|
242 |
cnn_feature = vision_tower_high(image)
|
243 |
-
cnn_feature = cnn_feature.flatten(2).permute(0, 2, 1)
|
244 |
image_feature = self.mm_projector_vary(cnn_feature)
|
245 |
image_features.append(image_feature)
|
246 |
|
@@ -249,7 +247,7 @@ class GOTQwenModel(Qwen2Model):
|
|
249 |
image_patches_features = []
|
250 |
for image_patch in image_patches:
|
251 |
image_p = torch.stack([image_patch])
|
252 |
-
|
253 |
with torch.set_grad_enabled(False):
|
254 |
cnn_feature_p = vision_tower_high(image_p)
|
255 |
cnn_feature_p = cnn_feature_p.flatten(2).permute(0, 2, 1)
|
@@ -258,39 +256,44 @@ class GOTQwenModel(Qwen2Model):
|
|
258 |
image_feature = torch.cat(image_patches_features, dim=1)
|
259 |
image_features.append(image_feature)
|
260 |
|
261 |
-
|
262 |
-
|
263 |
dummy_image_features = dummy_image_features_2
|
264 |
use_im_start_end = True
|
265 |
new_input_embeds = []
|
266 |
-
for cur_input_ids, cur_input_embeds, cur_image_features in zip(
|
|
|
267 |
if (cur_input_ids == im_patch_token).sum() == 0:
|
268 |
cur_input_embeds = cur_input_embeds + (0. * dummy_image_features).sum()
|
269 |
new_input_embeds.append(cur_input_embeds)
|
270 |
continue
|
271 |
|
272 |
if use_im_start_end:
|
273 |
-
if (cur_input_ids == im_start_token).sum() != (
|
274 |
-
|
275 |
-
|
|
|
|
|
276 |
image_start_tokens = torch.where(cur_input_ids == im_start_token)[0]
|
277 |
-
for image_start_token_pos, per_cur_image_features in zip(
|
278 |
-
|
|
|
|
|
279 |
num_patches = per_cur_image_features.shape[0]
|
280 |
|
281 |
if cur_input_ids[image_start_token_pos + num_patches + 1] != im_end_token:
|
282 |
-
raise ValueError(
|
283 |
-
|
|
|
284 |
cur_input_embeds = torch.cat(
|
285 |
(
|
286 |
-
cur_input_embeds[:image_start_token_pos+1],
|
287 |
-
per_cur_image_features,
|
288 |
cur_input_embeds[image_start_token_pos + num_patches + 1:]
|
289 |
-
),
|
290 |
dim=0
|
291 |
)
|
292 |
|
293 |
-
|
294 |
new_input_embeds.append(cur_input_embeds)
|
295 |
else:
|
296 |
raise NotImplementedError
|
@@ -299,13 +302,12 @@ class GOTQwenModel(Qwen2Model):
|
|
299 |
|
300 |
return super(GOTQwenModel, self).forward(
|
301 |
input_ids=None, attention_mask=attention_mask, past_key_values=past_key_values,
|
302 |
-
inputs_embeds=inputs_embeds, use_cache=use_cache, position_ids
|
303 |
output_attentions=output_attentions, output_hidden_states=output_hidden_states,
|
304 |
return_dict=return_dict
|
305 |
)
|
306 |
|
307 |
|
308 |
-
|
309 |
class GOTQwenForCausalLM(Qwen2ForCausalLM):
|
310 |
config_class = GOTConfig
|
311 |
# supports_gradient_checkpointing = True
|
@@ -336,15 +338,14 @@ class GOTQwenForCausalLM(Qwen2ForCausalLM):
|
|
336 |
output_hidden_states: Optional[bool] = None,
|
337 |
images: Optional[torch.FloatTensor] = None,
|
338 |
return_dict: Optional[bool] = None,
|
339 |
-
|
340 |
) -> Union[Tuple, CausalLMOutputWithPast]:
|
341 |
output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
|
342 |
output_hidden_states = (
|
343 |
-
output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
|
344 |
-
)
|
345 |
return_dict = return_dict if return_dict is not None else self.config.use_return_dict
|
346 |
|
347 |
-
outputs
|
348 |
input_ids=input_ids,
|
349 |
past_key_values=past_key_values,
|
350 |
attention_mask=attention_mask,
|
@@ -355,7 +356,7 @@ class GOTQwenForCausalLM(Qwen2ForCausalLM):
|
|
355 |
output_hidden_states=output_hidden_states,
|
356 |
images=images,
|
357 |
return_dict=return_dict
|
358 |
-
|
359 |
)
|
360 |
|
361 |
hidden_states = outputs[0]
|
@@ -389,7 +390,6 @@ class GOTQwenForCausalLM(Qwen2ForCausalLM):
|
|
389 |
attentions=outputs.attentions,
|
390 |
)
|
391 |
|
392 |
-
|
393 |
def prepare_inputs_for_generation(
|
394 |
self, input_ids, past_key_values=None, attention_mask=None, inputs_embeds=None, **kwargs
|
395 |
):
|
@@ -408,14 +408,16 @@ class GOTQwenForCausalLM(Qwen2ForCausalLM):
|
|
408 |
# some of the inputs are exclusively passed as part of the cache (e.g. when passing input_embeds as
|
409 |
# input)
|
410 |
if attention_mask is not None and attention_mask.shape[1] > input_ids.shape[1]:
|
411 |
-
input_ids = input_ids[:, -(attention_mask.shape[1] - past_length)
|
412 |
# 2 - If the past_length is smaller than input_ids', then input_ids holds all input tokens. We can discard
|
413 |
# input_ids based on the past_length.
|
414 |
elif past_length < input_ids.shape[1]:
|
415 |
input_ids = input_ids[:, past_length:]
|
416 |
-
# 3 - Otherwise (past_length >= input_ids.shape[1]), let's assume
|
|
|
417 |
|
418 |
-
# If we are about to go beyond the maximum cache length, we need to crop
|
|
|
419 |
if (
|
420 |
max_cache_length is not None
|
421 |
and attention_mask is not None
|
@@ -429,7 +431,7 @@ class GOTQwenForCausalLM(Qwen2ForCausalLM):
|
|
429 |
position_ids = attention_mask.long().cumsum(-1) - 1
|
430 |
position_ids.masked_fill_(attention_mask == 0, 1)
|
431 |
if past_key_values:
|
432 |
-
position_ids = position_ids[:, -input_ids.shape[1]
|
433 |
|
434 |
# if `inputs_embeds` are passed, we only want to use them in the 1st generation step
|
435 |
if inputs_embeds is not None and past_key_values is None:
|
@@ -449,15 +451,13 @@ class GOTQwenForCausalLM(Qwen2ForCausalLM):
|
|
449 |
return model_inputs
|
450 |
|
451 |
def initialize_vision_tokenizer(
|
452 |
-
self,
|
453 |
-
tokenizer,
|
454 |
-
freeze_lm_model=False,
|
455 |
pretrained_stage1_model=None,
|
456 |
-
device="cuda"
|
457 |
):
|
458 |
config = self.get_model().config
|
459 |
|
460 |
-
|
461 |
self.resize_token_embeddings(len(tokenizer))
|
462 |
|
463 |
config.im_patch_token = 151859
|
@@ -484,12 +484,23 @@ class GOTQwenForCausalLM(Qwen2ForCausalLM):
|
|
484 |
setattr(torch.nn.Linear, "reset_parameters", lambda self: None)
|
485 |
setattr(torch.nn.LayerNorm, "reset_parameters", lambda self: None)
|
486 |
|
487 |
-
def chat(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
488 |
|
489 |
self.disable_torch_init()
|
490 |
|
491 |
-
|
492 |
-
image_processor_high = GOTImageEvalProcessor(image_size=1024)
|
493 |
|
494 |
use_im_start_end = True
|
495 |
|
@@ -501,7 +512,7 @@ class GOTQwenForCausalLM(Qwen2ForCausalLM):
|
|
501 |
image = self.load_image(image_file)
|
502 |
|
503 |
w, h = image.size
|
504 |
-
|
505 |
if ocr_type == 'format':
|
506 |
qs = 'OCR with format: '
|
507 |
else:
|
@@ -510,13 +521,13 @@ class GOTQwenForCausalLM(Qwen2ForCausalLM):
|
|
510 |
if ocr_box:
|
511 |
bbox = eval(ocr_box)
|
512 |
if len(bbox) == 2:
|
513 |
-
bbox[0] = int(bbox[0]/w*1000)
|
514 |
-
bbox[1] = int(bbox[1]/h*1000)
|
515 |
if len(bbox) == 4:
|
516 |
-
bbox[0] = int(bbox[0]/w*1000)
|
517 |
-
bbox[1] = int(bbox[1]/h*1000)
|
518 |
-
bbox[2] = int(bbox[2]/w*1000)
|
519 |
-
bbox[3] = int(bbox[3]/h*1000)
|
520 |
if ocr_type == 'format':
|
521 |
qs = str(bbox) + ' ' + 'OCR with format: '
|
522 |
else:
|
@@ -529,11 +540,11 @@ class GOTQwenForCausalLM(Qwen2ForCausalLM):
|
|
529 |
qs = '[' + ocr_color + ']' + ' ' + 'OCR: '
|
530 |
|
531 |
if use_im_start_end:
|
532 |
-
qs = DEFAULT_IM_START_TOKEN + DEFAULT_IMAGE_PATCH_TOKEN*
|
|
|
533 |
else:
|
534 |
qs = DEFAULT_IMAGE_TOKEN + '\n' + qs
|
535 |
|
536 |
-
|
537 |
conv_mpt = Conversation(
|
538 |
system="""<|im_start|>system
|
539 |
You should follow the instructions carefully and explain your answers in detail.""",
|
@@ -558,40 +569,42 @@ class GOTQwenForCausalLM(Qwen2ForCausalLM):
|
|
558 |
|
559 |
image_tensor_1 = image_processor_high(image)
|
560 |
|
561 |
-
input_ids = torch.as_tensor(inputs.input_ids).
|
562 |
|
563 |
stop_str = conv.sep if conv.sep_style != SeparatorStyle.TWO else conv.sep2
|
564 |
keywords = [stop_str]
|
565 |
stopping_criteria = KeywordsStoppingCriteria(keywords, tokenizer, input_ids)
|
566 |
-
streamer = TextStreamer(
|
|
|
567 |
|
|
|
568 |
if stream_flag:
|
569 |
-
with torch.autocast(
|
570 |
output_ids = self.generate(
|
571 |
input_ids,
|
572 |
-
images=[image_tensor_1.unsqueeze(0).half().
|
573 |
do_sample=False,
|
574 |
-
num_beams
|
575 |
-
no_repeat_ngram_size
|
576 |
streamer=streamer,
|
577 |
max_new_tokens=4096,
|
578 |
stopping_criteria=[stopping_criteria]
|
579 |
-
|
580 |
else:
|
581 |
-
with torch.autocast(
|
582 |
output_ids = self.generate(
|
583 |
input_ids,
|
584 |
-
images=[image_tensor_1.unsqueeze(0).half().
|
585 |
do_sample=False,
|
586 |
-
num_beams
|
587 |
-
no_repeat_ngram_size
|
588 |
# streamer=streamer,
|
589 |
max_new_tokens=4096,
|
590 |
stopping_criteria=[stopping_criteria]
|
591 |
-
|
592 |
-
|
593 |
outputs = tokenizer.decode(output_ids[0, input_ids.shape[1]:]).strip()
|
594 |
-
|
595 |
if outputs.endswith(stop_str):
|
596 |
outputs = outputs[:-len(stop_str)]
|
597 |
outputs = outputs.strip()
|
@@ -606,8 +619,8 @@ class GOTQwenForCausalLM(Qwen2ForCausalLM):
|
|
606 |
tk = verovio.toolkit()
|
607 |
tk.loadData(outputs)
|
608 |
tk.setOptions({"pageWidth": 2100, "footer": 'none',
|
609 |
-
|
610 |
-
|
611 |
tk.getPageCount()
|
612 |
svg = tk.renderToSVG()
|
613 |
svg = svg.replace("overflow=\"inherit\"", "overflow=\"visible\"")
|
@@ -616,35 +629,52 @@ class GOTQwenForCausalLM(Qwen2ForCausalLM):
|
|
616 |
|
617 |
if ocr_type == 'format' and '**kern' not in outputs:
|
618 |
|
619 |
-
|
620 |
-
if '\\begin{tikzpicture}' not in outputs:
|
621 |
html_path_2 = save_render_file
|
622 |
right_num = outputs.count('\\right')
|
623 |
-
left_num = outputs.count('
|
624 |
|
625 |
if right_num != left_num:
|
626 |
-
outputs = outputs.replace(
|
627 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
628 |
|
629 |
outputs = outputs.replace('"', '``').replace('$', '')
|
630 |
|
631 |
outputs_list = outputs.split('\n')
|
632 |
-
gt= ''
|
633 |
for out in outputs_list:
|
634 |
-
gt +=
|
635 |
-
|
636 |
-
gt = gt[:-2]
|
637 |
|
|
|
638 |
|
639 |
lines = content_mmd_to_html
|
640 |
lines = lines.split("const text =")
|
641 |
-
new_web = lines[0] + 'const text ='
|
642 |
|
643 |
else:
|
644 |
html_path_2 = save_render_file
|
645 |
outputs = outputs.translate(translation_table)
|
646 |
outputs_list = outputs.split('\n')
|
647 |
-
gt= ''
|
648 |
for out in outputs_list:
|
649 |
if out:
|
650 |
if '\\begin{tikzpicture}' not in out and '\\end{tikzpicture}' not in out:
|
@@ -652,7 +682,7 @@ class GOTQwenForCausalLM(Qwen2ForCausalLM):
|
|
652 |
out = out[:-1]
|
653 |
if out is None:
|
654 |
break
|
655 |
-
|
656 |
if out:
|
657 |
if out[-1] != ';':
|
658 |
gt += out[:-1] + ';\n'
|
@@ -661,7 +691,6 @@ class GOTQwenForCausalLM(Qwen2ForCausalLM):
|
|
661 |
else:
|
662 |
gt += out + '\n'
|
663 |
|
664 |
-
|
665 |
lines = tik_html
|
666 |
lines = lines.split("const text =")
|
667 |
new_web = lines[0] + gt + lines[1]
|
@@ -671,7 +700,7 @@ class GOTQwenForCausalLM(Qwen2ForCausalLM):
|
|
671 |
return response_str
|
672 |
|
673 |
def dynamic_preprocess(self, image, min_num=1, max_num=6, image_size=1024, use_thumbnail=True):
|
674 |
-
|
675 |
def find_closest_aspect_ratio(aspect_ratio, target_ratios, width, height, image_size):
|
676 |
best_ratio_diff = float('inf')
|
677 |
best_ratio = (1, 1)
|
@@ -687,14 +716,25 @@ class GOTQwenForCausalLM(Qwen2ForCausalLM):
|
|
687 |
best_ratio = ratio
|
688 |
# print(f'width: {width}, height: {height}, best_ratio: {best_ratio}')
|
689 |
return best_ratio
|
690 |
-
|
691 |
orig_width, orig_height = image.size
|
692 |
aspect_ratio = orig_width / orig_height
|
693 |
|
694 |
# calculate the existing image aspect ratio
|
695 |
target_ratios = set(
|
696 |
-
(i,
|
697 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
698 |
# print(target_ratios)
|
699 |
target_ratios = sorted(target_ratios, key=lambda x: x[0] * x[1])
|
700 |
|
@@ -727,18 +767,25 @@ class GOTQwenForCausalLM(Qwen2ForCausalLM):
|
|
727 |
processed_images.append(thumbnail_img)
|
728 |
return processed_images
|
729 |
|
730 |
-
|
731 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
732 |
# Model
|
733 |
self.disable_torch_init()
|
734 |
-
multi_page=False
|
735 |
|
736 |
-
|
737 |
-
image_processor_high = GOTImageEvalProcessor(image_size=1024)
|
738 |
|
739 |
use_im_start_end = True
|
740 |
|
741 |
-
|
742 |
image_token_len = 256
|
743 |
|
744 |
image_list = []
|
@@ -778,18 +825,16 @@ class GOTQwenForCausalLM(Qwen2ForCausalLM):
|
|
778 |
image_tensor_1 = image_processor_high(image)
|
779 |
image_list.append(image_tensor_1)
|
780 |
|
781 |
-
|
782 |
image_list = torch.stack(image_list)
|
783 |
|
784 |
-
print('====new images batch size======: \n',image_list.shape)
|
785 |
-
|
786 |
|
787 |
if use_im_start_end:
|
788 |
-
qs = DEFAULT_IM_START_TOKEN + DEFAULT_IMAGE_PATCH_TOKEN*
|
|
|
789 |
else:
|
790 |
qs = DEFAULT_IMAGE_TOKEN + '\n' + qs
|
791 |
|
792 |
-
|
793 |
conv_mpt = Conversation(
|
794 |
system="""<|im_start|>system
|
795 |
You should follow the instructions carefully and explain your answers in detail.""",
|
@@ -812,43 +857,45 @@ class GOTQwenForCausalLM(Qwen2ForCausalLM):
|
|
812 |
|
813 |
inputs = tokenizer([prompt])
|
814 |
|
815 |
-
input_ids = torch.as_tensor(inputs.input_ids).
|
816 |
|
817 |
stop_str = conv.sep if conv.sep_style != SeparatorStyle.TWO else conv.sep2
|
818 |
keywords = [stop_str]
|
819 |
stopping_criteria = KeywordsStoppingCriteria(keywords, tokenizer, input_ids)
|
820 |
-
streamer = TextStreamer(
|
|
|
821 |
|
|
|
822 |
if stream_flag:
|
823 |
-
with torch.autocast(
|
824 |
output_ids = self.generate(
|
825 |
input_ids,
|
826 |
-
images=[image_list.half().
|
827 |
do_sample=False,
|
828 |
-
num_beams
|
829 |
# no_repeat_ngram_size = 20,
|
830 |
streamer=streamer,
|
831 |
max_new_tokens=4096,
|
832 |
stopping_criteria=[stopping_criteria]
|
833 |
-
|
834 |
else:
|
835 |
-
with torch.autocast(
|
836 |
output_ids = self.generate(
|
837 |
input_ids,
|
838 |
-
images=[image_list.half().
|
839 |
do_sample=False,
|
840 |
-
num_beams
|
841 |
# no_repeat_ngram_size = 20,
|
842 |
# streamer=streamer,
|
843 |
max_new_tokens=4096,
|
844 |
stopping_criteria=[stopping_criteria]
|
845 |
-
|
846 |
|
847 |
outputs = tokenizer.decode(output_ids[0, input_ids.shape[1]:]).strip()
|
848 |
-
|
849 |
if outputs.endswith(stop_str):
|
850 |
outputs = outputs[:-len(stop_str)]
|
851 |
-
outputs = outputs.strip()
|
852 |
response_str = outputs
|
853 |
|
854 |
if render:
|
@@ -856,26 +903,45 @@ class GOTQwenForCausalLM(Qwen2ForCausalLM):
|
|
856 |
from .render_tools import content_mmd_to_html
|
857 |
html_path_2 = save_render_file
|
858 |
right_num = outputs.count('\\right')
|
859 |
-
left_num = outputs.count('
|
860 |
|
861 |
if right_num != left_num:
|
862 |
-
outputs = outputs.replace(
|
863 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
864 |
|
865 |
outputs = outputs.replace('"', '``').replace('$', '')
|
866 |
|
867 |
outputs_list = outputs.split('\n')
|
868 |
-
gt= ''
|
869 |
for out in outputs_list:
|
870 |
-
gt +=
|
871 |
-
|
872 |
gt = gt[:-2]
|
873 |
|
874 |
lines = content_mmd_to_html
|
875 |
lines = lines.split("const text =")
|
876 |
-
new_web = lines[0] + 'const text ='
|
877 |
-
|
878 |
with open(html_path_2, 'w') as web_f_new:
|
879 |
web_f_new.write(new_web)
|
880 |
|
881 |
-
return response_str
|
|
|
1 |
+
from enum import auto, Enum
|
2 |
from transformers import Qwen2Config, Qwen2Model, Qwen2ForCausalLM, StoppingCriteria, TextStreamer
|
3 |
from transformers.modeling_outputs import BaseModelOutputWithPast, CausalLMOutputWithPast
|
4 |
from typing import List, Optional, Tuple, Union
|
|
|
20 |
DEFAULT_IM_START_TOKEN = '<img>'
|
21 |
DEFAULT_IM_END_TOKEN = '</img>'
|
22 |
|
23 |
+
|
24 |
class SeparatorStyle(Enum):
|
25 |
"""Different separator style."""
|
26 |
SINGLE = auto()
|
|
|
66 |
return ret
|
67 |
if self.sep_style == SeparatorStyle.MPT:
|
68 |
if self.system:
|
69 |
+
ret = self.system + self.sep
|
70 |
else:
|
71 |
ret = ''
|
72 |
for role, message in self.messages:
|
|
|
80 |
else:
|
81 |
raise ValueError(f"Invalid style: {self.sep_style}")
|
82 |
|
|
|
83 |
def append_message(self, role, message):
|
84 |
self.messages.append([role, message])
|
85 |
|
|
|
94 |
sep2=self.sep2)
|
95 |
|
96 |
|
|
|
97 |
class KeywordsStoppingCriteria(StoppingCriteria):
|
98 |
def __init__(self, keywords, tokenizer, input_ids):
|
99 |
self.keywords = keywords
|
100 |
self.keyword_ids = [tokenizer(keyword).input_ids for keyword in keywords]
|
101 |
+
self.keyword_ids = [keyword_id[0] for keyword_id in self.keyword_ids if type(
|
102 |
+
keyword_id) is list and len(keyword_id) == 1]
|
103 |
self.tokenizer = tokenizer
|
104 |
self.start_len = None
|
105 |
self.input_ids = input_ids
|
|
|
111 |
for keyword_id in self.keyword_ids:
|
112 |
if output_ids[0, -1] == keyword_id:
|
113 |
return True
|
114 |
+
outputs = self.tokenizer.batch_decode(
|
115 |
+
output_ids[:, self.start_len:], skip_special_tokens=True)[0]
|
116 |
for keyword in self.keywords:
|
117 |
if keyword in outputs:
|
118 |
return True
|
119 |
return False
|
120 |
+
|
121 |
|
122 |
class GOTImageEvalProcessor:
|
123 |
def __init__(self, image_size=384, mean=None, std=None):
|
|
|
137 |
self.normalize,
|
138 |
]
|
139 |
)
|
140 |
+
|
141 |
def __call__(self, item):
|
142 |
return self.transform(item)
|
143 |
|
144 |
|
|
|
145 |
class GOTConfig(Qwen2Config):
|
146 |
model_type = "GOT"
|
147 |
|
|
|
154 |
|
155 |
self.vision_tower_high = build_GOT_vit_b()
|
156 |
|
157 |
+
self.mm_projector_vary = nn.Linear(1024, 1024)
|
|
|
158 |
|
159 |
def initialize_vision_modules(
|
160 |
+
self,
|
161 |
vision_tower,
|
162 |
pretrained_stage1_model=None,
|
163 |
freeze_vision_tower=False,
|
164 |
use_im_start_end=False,
|
165 |
vision_select_layer=-1,
|
166 |
dtype=torch.float16,
|
|
|
167 |
):
|
168 |
|
169 |
+
device = self.device
|
170 |
image_processor_high = GOTImageEvalProcessor(image_size=1024)
|
171 |
+
|
172 |
self.vision_tower_high = self.vision_tower_high.to(dtype=dtype, device=device)
|
173 |
|
174 |
self.mm_projector_vary = self.mm_projector_vary.to(dtype=dtype, device=device)
|
175 |
|
|
|
176 |
image_token_len = 256
|
177 |
|
178 |
self.config.vision_tower = vision_tower
|
|
|
182 |
|
183 |
self.config.vision_select_layer = vision_select_layer
|
184 |
self.config.freeze_vision_tower = freeze_vision_tower
|
185 |
+
|
186 |
return dict(
|
187 |
image_processor_high=image_processor_high,
|
188 |
image_token_len=image_token_len,
|
189 |
)
|
190 |
+
|
|
|
191 |
def forward(
|
192 |
self,
|
193 |
input_ids: torch.LongTensor = None,
|
|
|
206 |
orig_embeds_params = getattr(self, 'orig_embeds_params', None)
|
207 |
if orig_embeds_params is not None:
|
208 |
with torch.no_grad():
|
209 |
+
self.get_input_embeddings().weight[:-
|
210 |
+
self.num_new_tokens] = orig_embeds_params[:-
|
211 |
+
self.num_new_tokens].data
|
212 |
|
213 |
if inputs_embeds is None:
|
214 |
inputs_embeds = self.embed_tokens(input_ids)
|
215 |
|
|
|
216 |
vision_tower_high = getattr(self, 'vision_tower_high', None)
|
217 |
|
218 |
+
if vision_tower_high is not None and (
|
219 |
+
input_ids.shape[1] != 1 or self.training) and images is not None:
|
220 |
use_im_start_end = getattr(self.config, "use_im_start_end", -1)
|
221 |
|
222 |
vision_select_layer = getattr(self.config, "vision_select_layer", -1)
|
|
|
230 |
im_start_token = 151857
|
231 |
|
232 |
im_end_token = 151858
|
233 |
+
|
234 |
image_features = []
|
235 |
+
|
236 |
for image in images:
|
237 |
P, C, H, W = image.shape
|
238 |
if P == 1:
|
239 |
with torch.set_grad_enabled(False):
|
240 |
cnn_feature = vision_tower_high(image)
|
241 |
+
cnn_feature = cnn_feature.flatten(2).permute(0, 2, 1) # 256*1024
|
242 |
image_feature = self.mm_projector_vary(cnn_feature)
|
243 |
image_features.append(image_feature)
|
244 |
|
|
|
247 |
image_patches_features = []
|
248 |
for image_patch in image_patches:
|
249 |
image_p = torch.stack([image_patch])
|
250 |
+
|
251 |
with torch.set_grad_enabled(False):
|
252 |
cnn_feature_p = vision_tower_high(image_p)
|
253 |
cnn_feature_p = cnn_feature_p.flatten(2).permute(0, 2, 1)
|
|
|
256 |
image_feature = torch.cat(image_patches_features, dim=1)
|
257 |
image_features.append(image_feature)
|
258 |
|
259 |
+
dummy_image_features_2 = torch.zeros(
|
260 |
+
256, 1024, device=inputs_embeds.device, dtype=inputs_embeds.dtype)
|
261 |
dummy_image_features = dummy_image_features_2
|
262 |
use_im_start_end = True
|
263 |
new_input_embeds = []
|
264 |
+
for cur_input_ids, cur_input_embeds, cur_image_features in zip(
|
265 |
+
input_ids, inputs_embeds, image_features):
|
266 |
if (cur_input_ids == im_patch_token).sum() == 0:
|
267 |
cur_input_embeds = cur_input_embeds + (0. * dummy_image_features).sum()
|
268 |
new_input_embeds.append(cur_input_embeds)
|
269 |
continue
|
270 |
|
271 |
if use_im_start_end:
|
272 |
+
if (cur_input_ids == im_start_token).sum() != (
|
273 |
+
cur_input_ids == im_end_token).sum():
|
274 |
+
raise ValueError(
|
275 |
+
"The number of image start tokens and image end tokens should be the same.")
|
276 |
+
|
277 |
image_start_tokens = torch.where(cur_input_ids == im_start_token)[0]
|
278 |
+
for image_start_token_pos, per_cur_image_features in zip(
|
279 |
+
image_start_tokens, cur_image_features):
|
280 |
+
per_cur_image_features = per_cur_image_features.to(
|
281 |
+
device=cur_input_embeds.device)
|
282 |
num_patches = per_cur_image_features.shape[0]
|
283 |
|
284 |
if cur_input_ids[image_start_token_pos + num_patches + 1] != im_end_token:
|
285 |
+
raise ValueError(
|
286 |
+
"The image end token should follow the image start token.")
|
287 |
+
|
288 |
cur_input_embeds = torch.cat(
|
289 |
(
|
290 |
+
cur_input_embeds[:image_start_token_pos + 1],
|
291 |
+
per_cur_image_features,
|
292 |
cur_input_embeds[image_start_token_pos + num_patches + 1:]
|
293 |
+
),
|
294 |
dim=0
|
295 |
)
|
296 |
|
|
|
297 |
new_input_embeds.append(cur_input_embeds)
|
298 |
else:
|
299 |
raise NotImplementedError
|
|
|
302 |
|
303 |
return super(GOTQwenModel, self).forward(
|
304 |
input_ids=None, attention_mask=attention_mask, past_key_values=past_key_values,
|
305 |
+
inputs_embeds=inputs_embeds, use_cache=use_cache, position_ids=position_ids,
|
306 |
output_attentions=output_attentions, output_hidden_states=output_hidden_states,
|
307 |
return_dict=return_dict
|
308 |
)
|
309 |
|
310 |
|
|
|
311 |
class GOTQwenForCausalLM(Qwen2ForCausalLM):
|
312 |
config_class = GOTConfig
|
313 |
# supports_gradient_checkpointing = True
|
|
|
338 |
output_hidden_states: Optional[bool] = None,
|
339 |
images: Optional[torch.FloatTensor] = None,
|
340 |
return_dict: Optional[bool] = None,
|
341 |
+
|
342 |
) -> Union[Tuple, CausalLMOutputWithPast]:
|
343 |
output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
|
344 |
output_hidden_states = (
|
345 |
+
output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states)
|
|
|
346 |
return_dict = return_dict if return_dict is not None else self.config.use_return_dict
|
347 |
|
348 |
+
outputs = self.model(
|
349 |
input_ids=input_ids,
|
350 |
past_key_values=past_key_values,
|
351 |
attention_mask=attention_mask,
|
|
|
356 |
output_hidden_states=output_hidden_states,
|
357 |
images=images,
|
358 |
return_dict=return_dict
|
359 |
+
|
360 |
)
|
361 |
|
362 |
hidden_states = outputs[0]
|
|
|
390 |
attentions=outputs.attentions,
|
391 |
)
|
392 |
|
|
|
393 |
def prepare_inputs_for_generation(
|
394 |
self, input_ids, past_key_values=None, attention_mask=None, inputs_embeds=None, **kwargs
|
395 |
):
|
|
|
408 |
# some of the inputs are exclusively passed as part of the cache (e.g. when passing input_embeds as
|
409 |
# input)
|
410 |
if attention_mask is not None and attention_mask.shape[1] > input_ids.shape[1]:
|
411 |
+
input_ids = input_ids[:, -(attention_mask.shape[1] - past_length):]
|
412 |
# 2 - If the past_length is smaller than input_ids', then input_ids holds all input tokens. We can discard
|
413 |
# input_ids based on the past_length.
|
414 |
elif past_length < input_ids.shape[1]:
|
415 |
input_ids = input_ids[:, past_length:]
|
416 |
+
# 3 - Otherwise (past_length >= input_ids.shape[1]), let's assume
|
417 |
+
# input_ids only has unprocessed tokens.
|
418 |
|
419 |
+
# If we are about to go beyond the maximum cache length, we need to crop
|
420 |
+
# the input attention mask.
|
421 |
if (
|
422 |
max_cache_length is not None
|
423 |
and attention_mask is not None
|
|
|
431 |
position_ids = attention_mask.long().cumsum(-1) - 1
|
432 |
position_ids.masked_fill_(attention_mask == 0, 1)
|
433 |
if past_key_values:
|
434 |
+
position_ids = position_ids[:, -input_ids.shape[1]:]
|
435 |
|
436 |
# if `inputs_embeds` are passed, we only want to use them in the 1st generation step
|
437 |
if inputs_embeds is not None and past_key_values is None:
|
|
|
451 |
return model_inputs
|
452 |
|
453 |
def initialize_vision_tokenizer(
|
454 |
+
self,
|
455 |
+
tokenizer,
|
456 |
+
freeze_lm_model=False,
|
457 |
pretrained_stage1_model=None,
|
|
|
458 |
):
|
459 |
config = self.get_model().config
|
460 |
|
|
|
461 |
self.resize_token_embeddings(len(tokenizer))
|
462 |
|
463 |
config.im_patch_token = 151859
|
|
|
484 |
setattr(torch.nn.Linear, "reset_parameters", lambda self: None)
|
485 |
setattr(torch.nn.LayerNorm, "reset_parameters", lambda self: None)
|
486 |
|
487 |
+
def chat(
|
488 |
+
self,
|
489 |
+
tokenizer,
|
490 |
+
image_file,
|
491 |
+
ocr_type,
|
492 |
+
ocr_box='',
|
493 |
+
ocr_color='',
|
494 |
+
render=False,
|
495 |
+
save_render_file=None,
|
496 |
+
print_prompt=False,
|
497 |
+
gradio_input=False,
|
498 |
+
stream_flag=False,
|
499 |
+
streamer=None):
|
500 |
|
501 |
self.disable_torch_init()
|
502 |
|
503 |
+
image_processor_high = GOTImageEvalProcessor(image_size=1024)
|
|
|
504 |
|
505 |
use_im_start_end = True
|
506 |
|
|
|
512 |
image = self.load_image(image_file)
|
513 |
|
514 |
w, h = image.size
|
515 |
+
|
516 |
if ocr_type == 'format':
|
517 |
qs = 'OCR with format: '
|
518 |
else:
|
|
|
521 |
if ocr_box:
|
522 |
bbox = eval(ocr_box)
|
523 |
if len(bbox) == 2:
|
524 |
+
bbox[0] = int(bbox[0] / w * 1000)
|
525 |
+
bbox[1] = int(bbox[1] / h * 1000)
|
526 |
if len(bbox) == 4:
|
527 |
+
bbox[0] = int(bbox[0] / w * 1000)
|
528 |
+
bbox[1] = int(bbox[1] / h * 1000)
|
529 |
+
bbox[2] = int(bbox[2] / w * 1000)
|
530 |
+
bbox[3] = int(bbox[3] / h * 1000)
|
531 |
if ocr_type == 'format':
|
532 |
qs = str(bbox) + ' ' + 'OCR with format: '
|
533 |
else:
|
|
|
540 |
qs = '[' + ocr_color + ']' + ' ' + 'OCR: '
|
541 |
|
542 |
if use_im_start_end:
|
543 |
+
qs = DEFAULT_IM_START_TOKEN + DEFAULT_IMAGE_PATCH_TOKEN * \
|
544 |
+
image_token_len + DEFAULT_IM_END_TOKEN + '\n' + qs
|
545 |
else:
|
546 |
qs = DEFAULT_IMAGE_TOKEN + '\n' + qs
|
547 |
|
|
|
548 |
conv_mpt = Conversation(
|
549 |
system="""<|im_start|>system
|
550 |
You should follow the instructions carefully and explain your answers in detail.""",
|
|
|
569 |
|
570 |
image_tensor_1 = image_processor_high(image)
|
571 |
|
572 |
+
input_ids = torch.as_tensor(inputs.input_ids).to(self.model.device)
|
573 |
|
574 |
stop_str = conv.sep if conv.sep_style != SeparatorStyle.TWO else conv.sep2
|
575 |
keywords = [stop_str]
|
576 |
stopping_criteria = KeywordsStoppingCriteria(keywords, tokenizer, input_ids)
|
577 |
+
streamer = streamer if streamer else TextStreamer(
|
578 |
+
tokenizer, skip_prompt=True, skip_special_tokens=True)
|
579 |
|
580 |
+
device = "cuda" if "cuda" in str(self.model.device) else "cpu"
|
581 |
if stream_flag:
|
582 |
+
with torch.autocast(device, dtype=torch.bfloat16):
|
583 |
output_ids = self.generate(
|
584 |
input_ids,
|
585 |
+
images=[image_tensor_1.unsqueeze(0).half().to(self.model.device)],
|
586 |
do_sample=False,
|
587 |
+
num_beams=1,
|
588 |
+
no_repeat_ngram_size=20,
|
589 |
streamer=streamer,
|
590 |
max_new_tokens=4096,
|
591 |
stopping_criteria=[stopping_criteria]
|
592 |
+
)
|
593 |
else:
|
594 |
+
with torch.autocast(device, dtype=torch.bfloat16):
|
595 |
output_ids = self.generate(
|
596 |
input_ids,
|
597 |
+
images=[image_tensor_1.unsqueeze(0).half().to(self.model.device)],
|
598 |
do_sample=False,
|
599 |
+
num_beams=1,
|
600 |
+
no_repeat_ngram_size=20,
|
601 |
# streamer=streamer,
|
602 |
max_new_tokens=4096,
|
603 |
stopping_criteria=[stopping_criteria]
|
604 |
+
)
|
605 |
+
|
606 |
outputs = tokenizer.decode(output_ids[0, input_ids.shape[1]:]).strip()
|
607 |
+
|
608 |
if outputs.endswith(stop_str):
|
609 |
outputs = outputs[:-len(stop_str)]
|
610 |
outputs = outputs.strip()
|
|
|
619 |
tk = verovio.toolkit()
|
620 |
tk.loadData(outputs)
|
621 |
tk.setOptions({"pageWidth": 2100, "footer": 'none',
|
622 |
+
'barLineWidth': 0.5, 'beamMaxSlope': 15,
|
623 |
+
'staffLineWidth': 0.2, 'spacingStaff': 6})
|
624 |
tk.getPageCount()
|
625 |
svg = tk.renderToSVG()
|
626 |
svg = svg.replace("overflow=\"inherit\"", "overflow=\"visible\"")
|
|
|
629 |
|
630 |
if ocr_type == 'format' and '**kern' not in outputs:
|
631 |
|
632 |
+
if '\\begin{tikzpicture}' not in outputs:
|
|
|
633 |
html_path_2 = save_render_file
|
634 |
right_num = outputs.count('\\right')
|
635 |
+
left_num = outputs.count('\\left')
|
636 |
|
637 |
if right_num != left_num:
|
638 |
+
outputs = outputs.replace(
|
639 |
+
'\\left(',
|
640 |
+
'(').replace(
|
641 |
+
'\\right)',
|
642 |
+
')').replace(
|
643 |
+
'\\left[',
|
644 |
+
'[').replace(
|
645 |
+
'\\right]',
|
646 |
+
']').replace(
|
647 |
+
'\\left{',
|
648 |
+
'{').replace(
|
649 |
+
'\\right}',
|
650 |
+
'}').replace(
|
651 |
+
'\\left|',
|
652 |
+
'|').replace(
|
653 |
+
'\\right|',
|
654 |
+
'|').replace(
|
655 |
+
'\\left.',
|
656 |
+
'.').replace(
|
657 |
+
'\\right.',
|
658 |
+
'.')
|
659 |
|
660 |
outputs = outputs.replace('"', '``').replace('$', '')
|
661 |
|
662 |
outputs_list = outputs.split('\n')
|
663 |
+
gt = ''
|
664 |
for out in outputs_list:
|
665 |
+
gt += '"' + out.replace('\\', '\\\\') + r'\n' + '"' + '+' + '\n'
|
|
|
|
|
666 |
|
667 |
+
gt = gt[:-2]
|
668 |
|
669 |
lines = content_mmd_to_html
|
670 |
lines = lines.split("const text =")
|
671 |
+
new_web = lines[0] + 'const text =' + gt + lines[1]
|
672 |
|
673 |
else:
|
674 |
html_path_2 = save_render_file
|
675 |
outputs = outputs.translate(translation_table)
|
676 |
outputs_list = outputs.split('\n')
|
677 |
+
gt = ''
|
678 |
for out in outputs_list:
|
679 |
if out:
|
680 |
if '\\begin{tikzpicture}' not in out and '\\end{tikzpicture}' not in out:
|
|
|
682 |
out = out[:-1]
|
683 |
if out is None:
|
684 |
break
|
685 |
+
|
686 |
if out:
|
687 |
if out[-1] != ';':
|
688 |
gt += out[:-1] + ';\n'
|
|
|
691 |
else:
|
692 |
gt += out + '\n'
|
693 |
|
|
|
694 |
lines = tik_html
|
695 |
lines = lines.split("const text =")
|
696 |
new_web = lines[0] + gt + lines[1]
|
|
|
700 |
return response_str
|
701 |
|
702 |
def dynamic_preprocess(self, image, min_num=1, max_num=6, image_size=1024, use_thumbnail=True):
|
703 |
+
|
704 |
def find_closest_aspect_ratio(aspect_ratio, target_ratios, width, height, image_size):
|
705 |
best_ratio_diff = float('inf')
|
706 |
best_ratio = (1, 1)
|
|
|
716 |
best_ratio = ratio
|
717 |
# print(f'width: {width}, height: {height}, best_ratio: {best_ratio}')
|
718 |
return best_ratio
|
719 |
+
|
720 |
orig_width, orig_height = image.size
|
721 |
aspect_ratio = orig_width / orig_height
|
722 |
|
723 |
# calculate the existing image aspect ratio
|
724 |
target_ratios = set(
|
725 |
+
(i,
|
726 |
+
j) for n in range(
|
727 |
+
min_num,
|
728 |
+
max_num +
|
729 |
+
1) for i in range(
|
730 |
+
1,
|
731 |
+
n +
|
732 |
+
1) for j in range(
|
733 |
+
1,
|
734 |
+
n +
|
735 |
+
1) if i *
|
736 |
+
j <= max_num and i *
|
737 |
+
j >= min_num)
|
738 |
# print(target_ratios)
|
739 |
target_ratios = sorted(target_ratios, key=lambda x: x[0] * x[1])
|
740 |
|
|
|
767 |
processed_images.append(thumbnail_img)
|
768 |
return processed_images
|
769 |
|
770 |
+
def chat_crop(
|
771 |
+
self,
|
772 |
+
tokenizer,
|
773 |
+
image_file,
|
774 |
+
ocr_type,
|
775 |
+
render=False,
|
776 |
+
save_render_file=None,
|
777 |
+
print_prompt=False,
|
778 |
+
gradio_input=False,
|
779 |
+
stream_flag=False,
|
780 |
+
streamer=None):
|
781 |
# Model
|
782 |
self.disable_torch_init()
|
783 |
+
multi_page = False
|
784 |
|
785 |
+
image_processor_high = GOTImageEvalProcessor(image_size=1024)
|
|
|
786 |
|
787 |
use_im_start_end = True
|
788 |
|
|
|
789 |
image_token_len = 256
|
790 |
|
791 |
image_list = []
|
|
|
825 |
image_tensor_1 = image_processor_high(image)
|
826 |
image_list.append(image_tensor_1)
|
827 |
|
|
|
828 |
image_list = torch.stack(image_list)
|
829 |
|
830 |
+
print('====new images batch size======: \n', image_list.shape)
|
|
|
831 |
|
832 |
if use_im_start_end:
|
833 |
+
qs = DEFAULT_IM_START_TOKEN + DEFAULT_IMAGE_PATCH_TOKEN * \
|
834 |
+
image_token_len * ll + DEFAULT_IM_END_TOKEN + '\n' + qs
|
835 |
else:
|
836 |
qs = DEFAULT_IMAGE_TOKEN + '\n' + qs
|
837 |
|
|
|
838 |
conv_mpt = Conversation(
|
839 |
system="""<|im_start|>system
|
840 |
You should follow the instructions carefully and explain your answers in detail.""",
|
|
|
857 |
|
858 |
inputs = tokenizer([prompt])
|
859 |
|
860 |
+
input_ids = torch.as_tensor(inputs.input_ids).to(self.model.device)
|
861 |
|
862 |
stop_str = conv.sep if conv.sep_style != SeparatorStyle.TWO else conv.sep2
|
863 |
keywords = [stop_str]
|
864 |
stopping_criteria = KeywordsStoppingCriteria(keywords, tokenizer, input_ids)
|
865 |
+
streamer = streamer if streamer else TextStreamer(
|
866 |
+
tokenizer, skip_prompt=True, skip_special_tokens=True)
|
867 |
|
868 |
+
device = "cuda" if "cuda" in str(self.model.device) else "cpu"
|
869 |
if stream_flag:
|
870 |
+
with torch.autocast(device, dtype=torch.bfloat16):
|
871 |
output_ids = self.generate(
|
872 |
input_ids,
|
873 |
+
images=[image_list.half().to(self.model.device)],
|
874 |
do_sample=False,
|
875 |
+
num_beams=1,
|
876 |
# no_repeat_ngram_size = 20,
|
877 |
streamer=streamer,
|
878 |
max_new_tokens=4096,
|
879 |
stopping_criteria=[stopping_criteria]
|
880 |
+
)
|
881 |
else:
|
882 |
+
with torch.autocast(device, dtype=torch.bfloat16):
|
883 |
output_ids = self.generate(
|
884 |
input_ids,
|
885 |
+
images=[image_list.half().to(self.model.device)],
|
886 |
do_sample=False,
|
887 |
+
num_beams=1,
|
888 |
# no_repeat_ngram_size = 20,
|
889 |
# streamer=streamer,
|
890 |
max_new_tokens=4096,
|
891 |
stopping_criteria=[stopping_criteria]
|
892 |
+
)
|
893 |
|
894 |
outputs = tokenizer.decode(output_ids[0, input_ids.shape[1]:]).strip()
|
895 |
+
|
896 |
if outputs.endswith(stop_str):
|
897 |
outputs = outputs[:-len(stop_str)]
|
898 |
+
outputs = outputs.strip()
|
899 |
response_str = outputs
|
900 |
|
901 |
if render:
|
|
|
903 |
from .render_tools import content_mmd_to_html
|
904 |
html_path_2 = save_render_file
|
905 |
right_num = outputs.count('\\right')
|
906 |
+
left_num = outputs.count('\\left')
|
907 |
|
908 |
if right_num != left_num:
|
909 |
+
outputs = outputs.replace(
|
910 |
+
'\\left(',
|
911 |
+
'(').replace(
|
912 |
+
'\\right)',
|
913 |
+
')').replace(
|
914 |
+
'\\left[',
|
915 |
+
'[').replace(
|
916 |
+
'\\right]',
|
917 |
+
']').replace(
|
918 |
+
'\\left{',
|
919 |
+
'{').replace(
|
920 |
+
'\\right}',
|
921 |
+
'}').replace(
|
922 |
+
'\\left|',
|
923 |
+
'|').replace(
|
924 |
+
'\\right|',
|
925 |
+
'|').replace(
|
926 |
+
'\\left.',
|
927 |
+
'.').replace(
|
928 |
+
'\\right.',
|
929 |
+
'.')
|
930 |
|
931 |
outputs = outputs.replace('"', '``').replace('$', '')
|
932 |
|
933 |
outputs_list = outputs.split('\n')
|
934 |
+
gt = ''
|
935 |
for out in outputs_list:
|
936 |
+
gt += '"' + out.replace('\\', '\\\\') + r'\n' + '"' + '+' + '\n'
|
937 |
+
|
938 |
gt = gt[:-2]
|
939 |
|
940 |
lines = content_mmd_to_html
|
941 |
lines = lines.split("const text =")
|
942 |
+
new_web = lines[0] + 'const text =' + gt + lines[1]
|
943 |
+
|
944 |
with open(html_path_2, 'w') as web_f_new:
|
945 |
web_f_new.write(new_web)
|
946 |
|
947 |
+
return response_str
|