|
--- |
|
license: apache-2.0 |
|
tags: |
|
- mistral |
|
- conversational |
|
- text-generation-inference |
|
base_model: |
|
- starble-dev/Starlight-V3-12B |
|
library_name: transformers |
|
--- |
|
|
|
> [!WARNING] |
|
> **General Use Sampling:**<br> |
|
> Mistral-Nemo-12B is very sensitive to the temperature sampler, try values near **0.3** at first or else you will get some weird results. This is mentioned by MistralAI at their [Transformers](https://huggingface.co./mistralai/Mistral-Nemo-Instruct-2407#transformers) section. |
|
|
|
> [!NOTE] |
|
> **Best Samplers:**<br> |
|
> I found best success using the following for Starlight-V3-12B:<br> |
|
> Temperature: `0.7`-`1.2` (Additional stopping strings will be necessary as you increase the temperature)<br> |
|
> Top K: `-1`<br> |
|
> Min P: `0.05`<br> |
|
> Rep Penalty: `1.03-1.1` |
|
|
|
# Why Version 3? |
|
Currently the other versions resulted in really bad results that I didn't upload them, the version number is just the internal version. |
|
|
|
# Goal |
|
The idea is to keep the strengths of [anthracite-org/magnum-12b-v2](https://huggingface.co./anthracite-org/magnum-12b-v2) while adding some more creativity |
|
that seems to be lacking in the model. Mistral-Nemo by itself seems to behave less sporadic due to the low temperature needed but this gets a bit repetitive, |
|
although it's still the best model I've used so far. |
|
|
|
# Results |
|
I am not entirely pleased with the result of the merge but it seems okay, though base [anthracite-org/magnum-12b-v2](https://huggingface.co./anthracite-org/magnum-12b-v2) |
|
might just be better by itself. However, I'll still experiement on different merge methods. |
|
Leaking of the training data used on both models seems a bit more apparent when using higher temperature values, |
|
especially the use of author notes on the system prompt. Generally I'd advise to create a stopping string for "```" to avoid the generation of the training data. |
|
|
|
**Original Models:** |
|
- [UsernameJustAnother/Nemo-12B-Marlin-v5](https://huggingface.co./UsernameJustAnother/Nemo-12B-Marlin-v5) (Thank you so much for your work ♥) |
|
- [anthracite-org/magnum-12b-v2](https://huggingface.co./anthracite-org/magnum-12b-v2) (Thank you so much for your work ♥) |
|
|
|
**Official Quants:** |
|
PPL = Perplexity, lower is better<br> |
|
Comparisons are done as QX_X Llama-3-8B against FP16 Llama-3-8B, recommended as a guideline and not as fact. |
|
| Quant Type | Note | Size | |
|
| ---- | ---- | ---- | |
|
| [Q2_K](https://huggingface.co./starble-dev/Starlight-V3-12B-GGUF/blob/main/Starlight-V3-12B-Q2_K.gguf) | +3.5199 ppl @ Llama-3-8B | 4.79 GB | |
|
| [Q3_K_S](https://huggingface.co./starble-dev/Starlight-V3-12B-GGUF/blob/main/Starlight-V3-12B-Q3_K_S.gguf) | +1.6321 ppl @ Llama-3-8B | 5.53 GB | |
|
| [Q3_K_M](https://huggingface.co./starble-dev/Starlight-V3-12B-GGUF/blob/main/Starlight-V3-12B-Q3_K_M.gguf) | +0.6569 ppl @ Llama-3-8B | 6.08 GB | |
|
| [Q3_K_L](https://huggingface.co./starble-dev/Starlight-V3-12B-GGUF/blob/main/Starlight-V3-12B-Q3_K_L.gguf) | +0.5562 ppl @ Llama-3-8B | 6.56 GB | |
|
| [Q4_K_S](https://huggingface.co./starble-dev/Starlight-V3-12B-GGUF/blob/main/Starlight-V3-12B-Q4_K_S.gguf) | +0.2689 ppl @ Llama-3-8B | 7.12 GB | |
|
| [Q4_K_M](https://huggingface.co./starble-dev/Starlight-V3-12B-GGUF/blob/main/Starlight-V3-12B-Q4_K_M.gguf) | +0.1754 ppl @ Llama-3-8B | 7.48 GB | |
|
| [Q5_K_S](https://huggingface.co./starble-dev/Starlight-V3-12B-GGUF/blob/main/Starlight-V3-12B-Q5_K_S.gguf) | +0.1049 ppl @ Llama-3-8B | 8.52 GB | |
|
| [Q5_K_M](https://huggingface.co./starble-dev/Starlight-V3-12B-GGUF/blob/main/Starlight-V3-12B-Q5_K_M.gguf) | +0.0569 ppl @ Llama-3-8B | 8.73 GB | |
|
| [Q6_K](https://huggingface.co./starble-dev/Starlight-V3-12B-GGUF/blob/main/Starlight-V3-12B-Q6_K.gguf) | +0.0217 ppl @ Llama-3-8B | 10.1 GB | |
|
| [Q8_0](https://huggingface.co./starble-dev/Starlight-V3-12B-GGUF/blob/main/Starlight-V3-12B-Q8_0.gguf) | +0.0026 ppl @ Llama-3-8B | 13.00 GB | |
|
|
|
**Original Model Licenses & This Model License:** Apache 2.0 |
