Update README.md
Browse files
README.md
CHANGED
|
@@ -26,12 +26,6 @@ Our GPTQ-based quantization methods achieve **superior quality-compression trade
|
|
| 26 |
- **Error-correcting updates** during calibration for improved accuracy
|
| 27 |
- **Optimized configurations** that allocate bits based on layer sensitivity (EvoPress)
|
| 28 |
|
| 29 |
-
| Method | Avg Bits | C4 PPL | WikiText2 PPL |
|
| 30 |
-
|--------|----------|--------|---------------|
|
| 31 |
-
| GPTQ-4 | 4.50 | 11.35 | 6.89 |
|
| 32 |
-
| EvoPress-GPTQ-4 | 4.50 | 11.35 | 6.89 |
|
| 33 |
-
| EvoPress-GPTQ-5 | 5.51 | 11.13 | 6.79 |
|
| 34 |
-
|
| 35 |
## Usage
|
| 36 |
|
| 37 |
Compatible with llama.cpp and all GGUF-supporting inference engines. No special setup required.
|
|
|
|
| 26 |
- **Error-correcting updates** during calibration for improved accuracy
|
| 27 |
- **Optimized configurations** that allocate bits based on layer sensitivity (EvoPress)
|
| 28 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 29 |
## Usage
|
| 30 |
|
| 31 |
Compatible with llama.cpp and all GGUF-supporting inference engines. No special setup required.
|