Spaces:
Build error
Build error
Upload 2 files
Browse files
app.py
CHANGED
@@ -2,9 +2,6 @@ import os
|
|
2 |
import gradio as gr
|
3 |
import main
|
4 |
|
5 |
-
#os.environ["CUDA_VISIBLE_DEVICES"]='0'
|
6 |
-
#os.environ["USE_GPU"]="True"
|
7 |
-
|
8 |
|
9 |
def predict_from_pdf(pdf_file):
|
10 |
upload_dir = "./catalogue/"
|
@@ -35,9 +32,9 @@ demo = gr.Interface(
|
|
35 |
outputs=["json", "text"],
|
36 |
examples=pdf_examples,
|
37 |
title="Open Source PDF Catalog Parser",
|
38 |
-
description="Efficient PDF catalog processing using
|
39 |
article="Uses MinerU for layout analysis and DeepSeek-7B for structured extraction"
|
40 |
)
|
41 |
|
42 |
if __name__ == "__main__":
|
43 |
-
demo.queue().launch(server_name="0.0.0.0", server_port=7860, share=True)
|
|
|
2 |
import gradio as gr
|
3 |
import main
|
4 |
|
|
|
|
|
|
|
5 |
|
6 |
def predict_from_pdf(pdf_file):
|
7 |
upload_dir = "./catalogue/"
|
|
|
32 |
outputs=["json", "text"],
|
33 |
examples=pdf_examples,
|
34 |
title="Open Source PDF Catalog Parser",
|
35 |
+
description="Efficient PDF catalog processing using MinerU and OpenLLM",
|
36 |
article="Uses MinerU for layout analysis and DeepSeek-7B for structured extraction"
|
37 |
)
|
38 |
|
39 |
if __name__ == "__main__":
|
40 |
+
demo.queue().launch(server_name="0.0.0.0", server_port=7860, share=True)
|
main.py
CHANGED
@@ -42,23 +42,26 @@ class PDFProcessor:
|
|
42 |
self.output_dir.mkdir(exist_ok=True)
|
43 |
|
44 |
def _initialize_emb_model(self, model_name):
|
45 |
-
|
46 |
-
|
47 |
-
|
48 |
-
|
49 |
-
|
|
|
|
|
50 |
|
51 |
-
|
52 |
-
|
53 |
-
|
54 |
|
55 |
def _initialize_llm(self, model_name):
|
56 |
"""Initialize LLM with automatic download if needed"""
|
|
|
57 |
model_path = os.path.join("models/", model_name)
|
58 |
if os.path.exists(model_path):
|
59 |
return Llama(
|
60 |
model_path=model_path,
|
61 |
-
n_ctx=
|
62 |
n_gpu_layers=35 if os.getenv('USE_GPU') else 0,
|
63 |
n_threads=os.cpu_count() - 1,
|
64 |
verbose=False
|
@@ -67,11 +70,16 @@ class PDFProcessor:
|
|
67 |
return Llama.from_pretrained(
|
68 |
repo_id="TheBloke/deepseek-llm-7B-base-GGUF",
|
69 |
filename=model_name,
|
70 |
-
n_ctx=
|
71 |
n_threads=os.cpu_count() - 1,
|
72 |
n_gpu_layers=35 if os.getenv('USE_GPU') else 0,
|
73 |
verbose=False
|
74 |
)
|
|
|
|
|
|
|
|
|
|
|
75 |
|
76 |
def process_pdf(self, pdf_path: str) -> Dict:
|
77 |
"""Process PDF using MinerU pipeline"""
|
|
|
42 |
self.output_dir.mkdir(exist_ok=True)
|
43 |
|
44 |
def _initialize_emb_model(self, model_name):
|
45 |
+
try:
|
46 |
+
model = SentenceTransformer("sentence-transformers/" + model_name)
|
47 |
+
model.save('models/'+ model_name)
|
48 |
+
return model
|
49 |
+
except:
|
50 |
+
# Load model directly
|
51 |
+
from transformers import AutoTokenizer, AutoModel
|
52 |
|
53 |
+
tokenizer = AutoTokenizer.from_pretrained("sentence-transformers/all-MiniLM-L6-v2")
|
54 |
+
model = AutoModel.from_pretrained("sentence-transformers/all-MiniLM-L6-v2")
|
55 |
+
return model
|
56 |
|
57 |
def _initialize_llm(self, model_name):
|
58 |
"""Initialize LLM with automatic download if needed"""
|
59 |
+
"""
|
60 |
model_path = os.path.join("models/", model_name)
|
61 |
if os.path.exists(model_path):
|
62 |
return Llama(
|
63 |
model_path=model_path,
|
64 |
+
n_ctx=2048,
|
65 |
n_gpu_layers=35 if os.getenv('USE_GPU') else 0,
|
66 |
n_threads=os.cpu_count() - 1,
|
67 |
verbose=False
|
|
|
70 |
return Llama.from_pretrained(
|
71 |
repo_id="TheBloke/deepseek-llm-7B-base-GGUF",
|
72 |
filename=model_name,
|
73 |
+
n_ctx=2048,
|
74 |
n_threads=os.cpu_count() - 1,
|
75 |
n_gpu_layers=35 if os.getenv('USE_GPU') else 0,
|
76 |
verbose=False
|
77 |
)
|
78 |
+
"""
|
79 |
+
# Load model directly
|
80 |
+
from transformers import AutoModel
|
81 |
+
model = AutoModel.from_pretrained("TheBloke/deepseek-llm-7B-base-GGUF")
|
82 |
+
return model
|
83 |
|
84 |
def process_pdf(self, pdf_path: str) -> Dict:
|
85 |
"""Process PDF using MinerU pipeline"""
|