Update README.md
Browse files
README.md
CHANGED
@@ -92,7 +92,7 @@ if __name__ == "__main__":
|
|
92 |
Script to test the vision model
|
93 |
```python
|
94 |
import base64
|
95 |
-
import
|
96 |
import json
|
97 |
import os
|
98 |
|
@@ -109,51 +109,44 @@ def encode_image_to_base64(image_path):
|
|
109 |
print(f"Error encoding image: {str(e)}")
|
110 |
return None
|
111 |
|
112 |
-
def
|
113 |
-
"""Send base64 image to vision model server using
|
114 |
try:
|
115 |
-
# Prepare JSON payload
|
116 |
payload = {
|
117 |
"prompt": "Describe this image in detail.",
|
118 |
"image": base64_image,
|
119 |
"max_tokens": 300
|
120 |
}
|
121 |
-
payload_json = json.dumps(payload)
|
122 |
|
123 |
-
|
124 |
-
"curl",
|
125 |
-
"-X", "POST",
|
126 |
"http://localhost:8000/v1/completions",
|
127 |
-
"
|
128 |
-
|
129 |
-
|
130 |
|
131 |
-
|
132 |
-
|
133 |
-
if result.returncode == 0:
|
134 |
print("Server response:")
|
135 |
-
print(
|
136 |
else:
|
137 |
-
print(f"
|
138 |
-
print(
|
139 |
|
140 |
-
except
|
141 |
-
print(f"Error
|
142 |
except Exception as e:
|
143 |
print(f"Unexpected error: {str(e)}")
|
144 |
|
145 |
def main():
|
146 |
-
image_path = "/path/to/
|
147 |
|
148 |
base64_image = encode_image_to_base64(image_path)
|
149 |
if base64_image:
|
150 |
-
|
151 |
-
send_curl_request(base64_image)
|
152 |
|
153 |
if __name__ == "__main__":
|
154 |
main()
|
155 |
```
|
156 |
-
The quantized model can be loaded using a single GPU with VRAM larger than
|
157 |
```bash
|
158 |
+-----------------------------------------------------------------------------------------+
|
159 |
| NVIDIA-SMI 570.133.20 Driver Version: 570.133.20 CUDA Version: 12.8 |
|
|
|
92 |
Script to test the vision model
|
93 |
```python
|
94 |
import base64
|
95 |
+
import requests
|
96 |
import json
|
97 |
import os
|
98 |
|
|
|
109 |
print(f"Error encoding image: {str(e)}")
|
110 |
return None
|
111 |
|
112 |
+
def send_request(base64_image):
|
113 |
+
"""Send base64 image to vision model server using requests."""
|
114 |
try:
|
|
|
115 |
payload = {
|
116 |
"prompt": "Describe this image in detail.",
|
117 |
"image": base64_image,
|
118 |
"max_tokens": 300
|
119 |
}
|
|
|
120 |
|
121 |
+
response = requests.post(
|
|
|
|
|
122 |
"http://localhost:8000/v1/completions",
|
123 |
+
headers={"Content-Type": "application/json"},
|
124 |
+
json=payload
|
125 |
+
)
|
126 |
|
127 |
+
if response.status_code == 200:
|
|
|
|
|
128 |
print("Server response:")
|
129 |
+
print(response.text)
|
130 |
else:
|
131 |
+
print(f"Request failed with status code {response.status_code}:")
|
132 |
+
print(response.text)
|
133 |
|
134 |
+
except requests.RequestException as e:
|
135 |
+
print(f"Error sending request: {str(e)}")
|
136 |
except Exception as e:
|
137 |
print(f"Unexpected error: {str(e)}")
|
138 |
|
139 |
def main():
|
140 |
+
image_path = "/path/to/sample.jpeg"
|
141 |
|
142 |
base64_image = encode_image_to_base64(image_path)
|
143 |
if base64_image:
|
144 |
+
send_request(base64_image)
|
|
|
145 |
|
146 |
if __name__ == "__main__":
|
147 |
main()
|
148 |
```
|
149 |
+
The quantized model can be loaded using a single GPU with VRAM larger than 12GB (Tested on Tesla T10)
|
150 |
```bash
|
151 |
+-----------------------------------------------------------------------------------------+
|
152 |
| NVIDIA-SMI 570.133.20 Driver Version: 570.133.20 CUDA Version: 12.8 |
|