Update README.md
Browse files
README.md
CHANGED
|
@@ -2,6 +2,12 @@
|
|
| 2 |
license: mit
|
| 3 |
---
|
| 4 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 5 |
# HPC-Quantize V2
|
| 6 |
|
| 7 |
## Holographic Phase Contraction, Geometry-Preserving Weight Analogs, and Algebraic Coupled Bilinear Factorization for Ultra-Low-Bit (`Q2_K`) LLM Quantization
|
|
|
|
| 2 |
license: mit
|
| 3 |
---
|
| 4 |
|
| 5 |
+
# Developer Note
|
| 6 |
+
|
| 7 |
+
This requires an absurd amount of resources to quantize, don't even think of attempting it with less than 28 threads.
|
| 8 |
+
|
| 9 |
+
If all goes as expected, it should be a substantial improvement over V1, which alone outperformed GCQ-RSQ; it is not even a contest of quality but time taken to quantize.
|
| 10 |
+
|
| 11 |
# HPC-Quantize V2
|
| 12 |
|
| 13 |
## Holographic Phase Contraction, Geometry-Preserving Weight Analogs, and Algebraic Coupled Bilinear Factorization for Ultra-Low-Bit (`Q2_K`) LLM Quantization
|