Update README.md
Browse files
README.md
CHANGED
|
@@ -222,6 +222,12 @@ curl --location 'http://localhost:8000/completion' \
|
|
| 222 |
}'
|
| 223 |
```
|
| 224 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 225 |
|
| 226 |
### Authors
|
| 227 |
* Kobkrit Viriyayudhakorn (kobkrit@aieat.or.th)
|
|
|
|
| 222 |
}'
|
| 223 |
```
|
| 224 |
|
| 225 |
+
### GPU Memory Requirements
|
| 226 |
+
| **ขนาดของ LLMs** | **FP 16 bits** | **8 bits (Quantized)** | **4 bits (Quantized)** | **ตัวอย่างการ์ดจอ สำหรับ 4 bits Quantized** |
|
| 227 |
+
|------------------|----------------|------------------------|------------------------|---------------------------------------------|
|
| 228 |
+
| **7b** | 24 GB | 12 GB | 6 GB | Nvidia RTX 4060 8GB |
|
| 229 |
+
| **13b** | 48 GB | 24 GB | 12 GB | Nvidia RTX 4070 16GB |
|
| 230 |
+
| **70b** | 192 GB | 96 GB | 48 GB | Nvidia RTX 4090 24GB x 2 ใบ |
|
| 231 |
|
| 232 |
### Authors
|
| 233 |
* Kobkrit Viriyayudhakorn (kobkrit@aieat.or.th)
|