Update README.md
Browse files
README.md
CHANGED
|
@@ -1,6 +1,11 @@
|
|
| 1 |
---
|
| 2 |
license: llama2
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 4 |
Experiment for DARE(Drop and REscale), most of the delta parameters can be directly set to zeros without affecting the capabilities of SFT LMs and larger models can tolerate a higher proportion of discarded parameters.
|
| 5 |
|
| 6 |
weight_mask_rate: 0.85 / use_weight_rescale: True / mask_stratery: random / scaling_coefficient: 1.0
|
|
|
|
| 1 |
---
|
| 2 |
license: llama2
|
| 3 |
---
|
| 4 |
+
|
| 5 |
+
* [AWQ model(s) for GPU inference.](https://huggingface.co/TheBloke/SynthIA-7B-v1.3-dare-0.85-AWQ)
|
| 6 |
+
* [GPTQ models for GPU inference, with multiple quantisation parameter options.](https://huggingface.co/TheBloke/SynthIA-7B-v1.3-dare-0.85-GPTQ)
|
| 7 |
+
* [2, 3, 4, 5, 6 and 8-bit GGUF models for CPU+GPU inference](https://huggingface.co/TheBloke/SynthIA-7B-v1.3-dare-0.85-GGUF)
|
| 8 |
+
|
| 9 |
Experiment for DARE(Drop and REscale), most of the delta parameters can be directly set to zeros without affecting the capabilities of SFT LMs and larger models can tolerate a higher proportion of discarded parameters.
|
| 10 |
|
| 11 |
weight_mask_rate: 0.85 / use_weight_rescale: True / mask_stratery: random / scaling_coefficient: 1.0
|