Update README.md
Browse files
README.md
CHANGED
@@ -1,5 +1,32 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
|
|
|
2 |
|
3 |
-
|
4 |
|
5 |
-
- needs [#8604](https://github.com/ggerganov/llama.cpp/pull/8604) & latest master with tekken tokenizer fixes applied
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: apache-2.0
|
3 |
+
language:
|
4 |
+
- en
|
5 |
+
pipeline_tag: text-generation
|
6 |
+
tags:
|
7 |
+
- mistral
|
8 |
+
- instruct
|
9 |
+
- gguf
|
10 |
+
- imatrix
|
11 |
+
base_model: mistralai/Mistral-Nemo-Instruct-2407
|
12 |
+
---
|
13 |
|
14 |
+
# Quant Infos
|
15 |
|
16 |
+
## Updated for all recent llama.cpp fixes (final logit soft capping+sliding window+tokenizer)
|
17 |
|
18 |
+
- needs [#8604](https://github.com/ggerganov/llama.cpp/pull/8604) & latest master with tekken tokenizer fixes applied
|
19 |
+
|
20 |
+
- quants done with an importance matrix for improved quantization loss
|
21 |
+
- Requantized ggufs & imatrix from hf bf16
|
22 |
+
- Wide coverage of different gguf quant types from Q\_8\_0 down to IQ1\_S
|
23 |
+
- experimental custom quant types
|
24 |
+
- `_L` with `--output-tensor-type f16 --token-embedding-type f16` (same as bartowski's)
|
25 |
+
- Imatrix generated with [this](https://gist.github.com/bartowski1182/eb213dccb3571f863da82e99418f81e8) multi-purpose dataset by [bartowski](https://huggingface.co/bartowski).
|
26 |
+
```
|
27 |
+
./imatrix -m $model_name-bf16.gguf -f calibration_datav3.txt -o $model_name.imatrix
|
28 |
+
```
|
29 |
+
|
30 |
+
# Original Model Card:
|
31 |
+
|
32 |
+
TODO
|