Spaces:
Running
on
Zero
Running
on
Zero
Commit
·
1796549
1
Parent(s):
4e1ec1c
refactor
Browse files
app.py
CHANGED
@@ -1,13 +1,16 @@
|
|
1 |
import spaces
|
2 |
import gradio as gr
|
3 |
|
4 |
-
from transformers import Qwen2VLForConditionalGeneration,
|
5 |
from qwen_vl_utils import process_vision_info
|
6 |
import torch
|
7 |
import os
|
8 |
import json
|
|
|
|
|
9 |
|
10 |
os.environ["HF_HUB_ENABLE_HF_TRANSFER"] = "1"
|
|
|
11 |
model = Qwen2VLForConditionalGeneration.from_pretrained(
|
12 |
"Qwen/Qwen2-VL-7B-Instruct",
|
13 |
torch_dtype=torch.bfloat16,
|
@@ -15,8 +18,7 @@ model = Qwen2VLForConditionalGeneration.from_pretrained(
|
|
15 |
device_map="auto",
|
16 |
)
|
17 |
processor = AutoProcessor.from_pretrained("Qwen/Qwen2-VL-7B-Instruct")
|
18 |
-
|
19 |
-
from typing import Tuple
|
20 |
|
21 |
class GeneralRetrievalQuery(BaseModel):
|
22 |
broad_topical_query: str
|
@@ -26,6 +28,7 @@ class GeneralRetrievalQuery(BaseModel):
|
|
26 |
visual_element_query: str
|
27 |
visual_element_explanation: str
|
28 |
|
|
|
29 |
def get_retrieval_prompt(prompt_name: str) -> Tuple[str, GeneralRetrievalQuery]:
|
30 |
if prompt_name != "general":
|
31 |
raise ValueError("Only 'general' prompt is available in this version")
|
@@ -66,11 +69,11 @@ Generate the queries based on this image and provide the response in the specifi
|
|
66 |
return prompt, GeneralRetrievalQuery
|
67 |
|
68 |
|
69 |
-
|
70 |
prompt, pydantic_model = get_retrieval_prompt("general")
|
71 |
|
72 |
-
|
73 |
-
def
|
74 |
messages = [
|
75 |
{
|
76 |
"role": "user",
|
@@ -97,6 +100,12 @@ def generate_response(image):
|
|
97 |
padding=True,
|
98 |
return_tensors="pt",
|
99 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
100 |
inputs = inputs.to("cuda")
|
101 |
|
102 |
generated_ids = model.generate(**inputs, max_new_tokens=200)
|
@@ -116,5 +125,6 @@ def generate_response(image):
|
|
116 |
except Exception:
|
117 |
return {}
|
118 |
|
119 |
-
|
|
|
120 |
demo.launch()
|
|
|
1 |
import spaces
|
2 |
import gradio as gr
|
3 |
|
4 |
+
from transformers import Qwen2VLForConditionalGeneration, AutoProcessor
|
5 |
from qwen_vl_utils import process_vision_info
|
6 |
import torch
|
7 |
import os
|
8 |
import json
|
9 |
+
from pydantic import BaseModel
|
10 |
+
from typing import Tuple
|
11 |
|
12 |
os.environ["HF_HUB_ENABLE_HF_TRANSFER"] = "1"
|
13 |
+
|
14 |
model = Qwen2VLForConditionalGeneration.from_pretrained(
|
15 |
"Qwen/Qwen2-VL-7B-Instruct",
|
16 |
torch_dtype=torch.bfloat16,
|
|
|
18 |
device_map="auto",
|
19 |
)
|
20 |
processor = AutoProcessor.from_pretrained("Qwen/Qwen2-VL-7B-Instruct")
|
21 |
+
|
|
|
22 |
|
23 |
class GeneralRetrievalQuery(BaseModel):
|
24 |
broad_topical_query: str
|
|
|
28 |
visual_element_query: str
|
29 |
visual_element_explanation: str
|
30 |
|
31 |
+
|
32 |
def get_retrieval_prompt(prompt_name: str) -> Tuple[str, GeneralRetrievalQuery]:
|
33 |
if prompt_name != "general":
|
34 |
raise ValueError("Only 'general' prompt is available in this version")
|
|
|
69 |
return prompt, GeneralRetrievalQuery
|
70 |
|
71 |
|
72 |
+
# defined like this so we can later add more prompting options
|
73 |
prompt, pydantic_model = get_retrieval_prompt("general")
|
74 |
|
75 |
+
|
76 |
+
def _prep_data_for_input(image):
|
77 |
messages = [
|
78 |
{
|
79 |
"role": "user",
|
|
|
100 |
padding=True,
|
101 |
return_tensors="pt",
|
102 |
)
|
103 |
+
return inputs
|
104 |
+
|
105 |
+
|
106 |
+
@spaces.GPU
|
107 |
+
def generate_response(image):
|
108 |
+
inputs = _prep_data_for_input(image)
|
109 |
inputs = inputs.to("cuda")
|
110 |
|
111 |
generated_ids = model.generate(**inputs, max_new_tokens=200)
|
|
|
125 |
except Exception:
|
126 |
return {}
|
127 |
|
128 |
+
|
129 |
+
demo = gr.Interface(fn=generate_response, inputs=gr.Image(type="pil"), outputs="json")
|
130 |
demo.launch()
|