| license: cc-by-nc-2.0 | |
| GGUF importance matrix (imatrix) quants for https://huggingface.co/wolfram/miquliz-120b-v2.0 | |
| The importance matrix was trained for 100K tokens (200 batches of 512 tokens) using wiki.train.raw. | |
| Using IQ2_XXS it seems to fit 100/141 layers using 2K context on a 24GB card. | |
| | Layers | Context | Template | | |
| | --- | --- | --- | | |
| | <pre>140</pre> | <pre>32768</pre> | <pre>[INST] {prompt} [/INST]<br>{response}</pre> | |