Update main README with VRAM Matrix
Browse files
README.md
CHANGED
|
@@ -22,6 +22,7 @@ The model weights are stored in separate branches. **Please switch to a branch t
|
|
| 22 |
| 4.0 | h6 | [4.0bpw_h6](https://huggingface.co/DBMe/Qwen3.5-9B-Instruct-exl3/tree/4.0bpw_h6) | 8.7612 | ~7.49 GB | ~7.49 GB | ~7.49 GB | ~7.49 GB | ~7.49 GB |
|
| 23 |
| 5.0 | h6 | [5.0bpw_h6](https://huggingface.co/DBMe/Qwen3.5-9B-Instruct-exl3/tree/5.0bpw_h6) | 8.7141 | ~8.3 GB | ~8.3 GB | ~8.3 GB | ~8.3 GB | ~8.3 GB |
|
| 24 |
| 6.0 | h6 | [6.0bpw_h6](https://huggingface.co/DBMe/Qwen3.5-9B-Instruct-exl3/tree/6.0bpw_h6) | 8.8005 | ~9.1 GB | ~9.1 GB | ~9.1 GB | ~9.1 GB | ~9.1 GB |
|
|
|
|
| 25 |
|
| 26 |
¹ *Measured on WikiText-2, context length 2048. Lower is better.*
|
| 27 |
*(Higher BPW = higher quality, lower BPW = fits in less VRAM).*
|
|
|
|
| 22 |
| 4.0 | h6 | [4.0bpw_h6](https://huggingface.co/DBMe/Qwen3.5-9B-Instruct-exl3/tree/4.0bpw_h6) | 8.7612 | ~7.49 GB | ~7.49 GB | ~7.49 GB | ~7.49 GB | ~7.49 GB |
|
| 23 |
| 5.0 | h6 | [5.0bpw_h6](https://huggingface.co/DBMe/Qwen3.5-9B-Instruct-exl3/tree/5.0bpw_h6) | 8.7141 | ~8.3 GB | ~8.3 GB | ~8.3 GB | ~8.3 GB | ~8.3 GB |
|
| 24 |
| 6.0 | h6 | [6.0bpw_h6](https://huggingface.co/DBMe/Qwen3.5-9B-Instruct-exl3/tree/6.0bpw_h6) | 8.8005 | ~9.1 GB | ~9.1 GB | ~9.1 GB | ~9.1 GB | ~9.1 GB |
|
| 25 |
+
| 8.0 | h8 | [8.0bpw_h8](https://huggingface.co/DBMe/Qwen3.5-9B-Instruct-exl3/tree/8.0bpw_h8) | 8.7477 | ~10.95 GB | ~10.95 GB | ~10.95 GB | ~10.95 GB | ~10.95 GB |
|
| 26 |
|
| 27 |
¹ *Measured on WikiText-2, context length 2048. Lower is better.*
|
| 28 |
*(Higher BPW = higher quality, lower BPW = fits in less VRAM).*
|