Update README.md
Browse files
README.md
CHANGED
|
@@ -6,6 +6,9 @@ pipeline_tag: text-generation
|
|
| 6 |
GGUF importance matrix (imatrix) quants for https://huggingface.co/codellama/CodeLlama-70b-Instruct-hf
|
| 7 |
The importance matrix was trained for 100K tokens (200 batches of 512 tokens) using wiki.train.raw.
|
| 8 |
|
|
|
|
|
|
|
|
|
|
| 9 |
| Layers | Context | Template |
|
| 10 |
| --- | --- | --- |
|
| 11 |
| <pre>0</pre> | <pre>4096</pre> | <pre>\<s\> Source: system<br><br> {instructions}\<step\> Source: user<br><br> {prompt}\<step\> Source: assistant<br>Destination: user<br><br> {response}</pre> |
|
|
|
|
| 6 |
GGUF importance matrix (imatrix) quants for https://huggingface.co/codellama/CodeLlama-70b-Instruct-hf
|
| 7 |
The importance matrix was trained for 100K tokens (200 batches of 512 tokens) using wiki.train.raw.
|
| 8 |
|
| 9 |
+
The template for this model is very sensitive and must be set very precisely.
|
| 10 |
+
All whitespace are intended, and special tokens `<s>` and `<step>` must be encodded properly.
|
| 11 |
+
|
| 12 |
| Layers | Context | Template |
|
| 13 |
| --- | --- | --- |
|
| 14 |
| <pre>0</pre> | <pre>4096</pre> | <pre>\<s\> Source: system<br><br> {instructions}\<step\> Source: user<br><br> {prompt}\<step\> Source: assistant<br>Destination: user<br><br> {response}</pre> |
|