Update inference.py
This commit is contained in:
191
inference.py
191
inference.py
@@ -1,93 +1,98 @@
|
|||||||
import torch
|
import torch
|
||||||
from transformers import AutoModelForCausalLM, AutoProcessor, AutoTokenizer
|
from transformers import AutoModelForCausalLM, AutoProcessor, AutoTokenizer
|
||||||
from qwen_vl_utils import process_vision_info
|
from qwen_vl_utils import process_vision_info
|
||||||
from dots_ocr.utils import dict_promptmode_to_prompt
|
from dots_ocr.utils import dict_promptmode_to_prompt
|
||||||
from PIL import Image
|
from PIL import Image
|
||||||
import io
|
import io
|
||||||
import json
|
import json
|
||||||
from typing import List, Dict, Tuple
|
from typing import List, Dict, Tuple
|
||||||
import warnings
|
import warnings
|
||||||
warnings.filterwarnings("ignore")
|
warnings.filterwarnings("ignore")
|
||||||
|
|
||||||
class DotsOcr:
|
class DotsOcr:
|
||||||
def __init__(self, model_path, device="cuda"):
|
def __init__(self, model_path, device="cuda"):
|
||||||
self.device = device
|
self.device = device
|
||||||
self.model = AutoModelForCausalLM.from_pretrained(
|
self.model = AutoModelForCausalLM.from_pretrained(
|
||||||
model_path,
|
model_path,
|
||||||
# attn_implementation="flash_attention_2",
|
# attn_implementation="flash_attention_2",
|
||||||
dtype="bfloat16",
|
dtype="bfloat16",
|
||||||
trust_remote_code=True
|
trust_remote_code=True
|
||||||
)
|
)
|
||||||
self.model.to(self.device)
|
self.model.to(self.device)
|
||||||
self.processor = AutoProcessor.from_pretrained(model_path, trust_remote_code=True, use_fast=True)
|
self.processor = AutoProcessor.from_pretrained(model_path, trust_remote_code=True, use_fast=True)
|
||||||
self.prompt = """Please output the layout information from the PDF image, including each layout element's bbox, its category, and the corresponding text content within the bbox.
|
self.prompt = """Please output the layout information from the PDF image, including each layout element's bbox, its category, and the corresponding text content within the bbox.
|
||||||
|
|
||||||
1. Bbox format: [x1, y1, x2, y2]
|
1. Bbox format: [x1, y1, x2, y2]
|
||||||
|
|
||||||
2. Layout Categories: The possible categories are ['Caption', 'Footnote', 'Formula', 'List-item', 'Page-footer', 'Page-header', 'Picture', 'Section-header', 'Table', 'Text', 'Title'].
|
2. Layout Categories: The possible categories are ['Caption', 'Footnote', 'Formula', 'List-item', 'Page-footer', 'Page-header', 'Picture', 'Section-header', 'Table', 'Text', 'Title'].
|
||||||
|
|
||||||
3. Text Extraction & Formatting Rules:
|
3. Text Extraction & Formatting Rules:
|
||||||
- Picture: For the 'Picture' category, the text field should be omitted.
|
- Picture: For the 'Picture' category, the text field should be omitted.
|
||||||
- Formula: Format its text as LaTeX.
|
- Formula: Format its text as LaTeX.
|
||||||
- Table: Format its text as HTML.
|
- Table: Format its text as HTML.
|
||||||
- All Others (Text, Title, etc.): Format their text as Markdown.
|
- All Others (Text, Title, etc.): Format their text as Markdown.
|
||||||
|
|
||||||
4. Constraints:
|
4. Constraints:
|
||||||
- The output text must be the original text from the image, with no translation.
|
- The output text must be the original text from the image, with no translation.
|
||||||
- All layout elements must be sorted according to human reading order.
|
- All layout elements must be sorted according to human reading order.
|
||||||
|
|
||||||
5. Final Output: The entire output must be a single JSON object.
|
5. Final Output: The entire output must be a single JSON object.
|
||||||
"""
|
"""
|
||||||
|
|
||||||
def sample_inference(self, image: Image.Image):
|
def sample_inference(self, image: Image.Image):
|
||||||
"""
|
"""
|
||||||
处理 PIL Image 对象，返回解析后的结果列表。To avoid OOM, process one image at a time.
|
处理 PIL Image 对象，返回解析后的结果列表。To avoid OOM, process one image at a time.
|
||||||
"""
|
"""
|
||||||
messages = [
|
messages = [
|
||||||
{
|
{
|
||||||
"role": "user",
|
"role": "user",
|
||||||
"content": [
|
"content": [
|
||||||
{
|
{
|
||||||
"type": "image",
|
"type": "image",
|
||||||
"image": image
|
"image": image
|
||||||
},
|
},
|
||||||
{"type": "text", "text": self.prompt}
|
{"type": "text", "text": self.prompt}
|
||||||
]
|
]
|
||||||
}
|
}
|
||||||
]
|
]
|
||||||
|
|
||||||
# Preparation for inference
|
# Preparation for inference
|
||||||
text = self.processor.apply_chat_template(
|
text = self.processor.apply_chat_template(
|
||||||
messages,
|
messages,
|
||||||
tokenize=False,
|
tokenize=False,
|
||||||
add_generation_prompt=True
|
add_generation_prompt=True
|
||||||
)
|
)
|
||||||
image_inputs, video_inputs = process_vision_info(messages)
|
image_inputs, video_inputs = process_vision_info(messages)
|
||||||
inputs = self.processor(
|
inputs = self.processor(
|
||||||
text=[text],
|
text=[text],
|
||||||
images=image_inputs,
|
images=image_inputs,
|
||||||
videos=video_inputs,
|
videos=video_inputs,
|
||||||
padding=True,
|
padding=True,
|
||||||
return_tensors="pt",
|
return_tensors="pt",
|
||||||
).to(self.device)
|
).to(self.device)
|
||||||
|
|
||||||
|
|
||||||
# Inference: Generation of the output
|
# Inference: Generation of the output
|
||||||
generated_ids = self.model.generate(**inputs, max_new_tokens=24000)
|
generated_ids = self.model.generate(**inputs, max_new_tokens=24000)
|
||||||
generated_ids_trimmed = [
|
generated_ids_trimmed = [
|
||||||
out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
|
out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
|
||||||
]
|
]
|
||||||
output_text = self.processor.batch_decode(
|
output_text = self.processor.batch_decode(
|
||||||
generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
|
generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
|
||||||
)
|
)
|
||||||
result_str = output_text[0]
|
result_str = output_text[0]
|
||||||
result_list = json.loads(result_str)
|
|
||||||
return result_list
|
try:
|
||||||
|
result_list = json.loads(result_str)
|
||||||
if __name__=="__main__":
|
return result_list
|
||||||
model_path = "DotsOCR"
|
except json.JSONDecodeError:
|
||||||
dots_ocr = DotsOcr(model_path)
|
print(f"无法解析 JSON 输出: {result_str}")
|
||||||
image_paths = ["20260306-065852.webp", "20260306-065909.webp"]
|
return {"error": "无法解析 JSON 输出"}
|
||||||
for image_path in image_paths:
|
|
||||||
output_text = dots_ocr.sample_inference(image_path)
|
if __name__=="__main__":
|
||||||
print(output_text)
|
model_path = "DotsOCR"
|
||||||
|
dots_ocr = DotsOcr(model_path)
|
||||||
|
image_paths = ["20260310-162729.webp"]
|
||||||
|
for image_path in image_paths:
|
||||||
|
output_text = dots_ocr.sample_inference(image_path)
|
||||||
|
print(output_text)
|
||||||
|
|||||||
Reference in New Issue
Block a user