Lewdiculous commited on
Commit
7c4516e
1 Parent(s): e7f749b

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +9 -0
README.md CHANGED
@@ -20,6 +20,15 @@ base_model:
20
  ---
21
 
22
  This repository hosts GGUF-Imatrix quantizations for [ResplendentAI/Paradigm_7B](https://huggingface.co/ResplendentAI/Paradigm_7B).
 
 
 
 
 
 
 
 
 
23
  ```
24
  Base⇢ GGUF(F16)⇢ Imatrix-Data(F16)⇢ GGUF(Imatrix-Quants)
25
  ```
 
20
  ---
21
 
22
  This repository hosts GGUF-Imatrix quantizations for [ResplendentAI/Paradigm_7B](https://huggingface.co/ResplendentAI/Paradigm_7B).
23
+
24
+ **What does "Imatrix" mean?**
25
+
26
+ It stands for **Importance Matrix**, a technique used to improve the quality of quantized models.
27
+ The **Imatrix** is calculated based on calibration data, and it helps determine the importance of different model activations during the quantization process.
28
+ The idea is to preserve the most important information during quantization, which can help reduce the loss of model performance, especially when the calibration data is diverse.
29
+ [[1]](https://github.com/ggerganov/llama.cpp/discussions/5006) [[2]](https://github.com/ggerganov/llama.cpp/discussions/5263#discussioncomment-8395384)
30
+
31
+ **Steps:**
32
  ```
33
  Base⇢ GGUF(F16)⇢ Imatrix-Data(F16)⇢ GGUF(Imatrix-Quants)
34
  ```