Update README.md
Browse files
README.md
CHANGED
|
@@ -2,6 +2,25 @@
|
|
| 2 |
library_name: transformers
|
| 3 |
license: llama3
|
| 4 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 5 |
# Llama-3-8B-Instruct-abliterated-v3 Model Card
|
| 6 |
|
| 7 |
[My Jupyter "cookbook" to replicate the methodology can be found here, refined library coming soon](https://huggingface.co/failspy/llama-3-70B-Instruct-abliterated/blob/main/ortho_cookbook.ipynb)
|
|
|
|
| 2 |
library_name: transformers
|
| 3 |
license: llama3
|
| 4 |
---
|
| 5 |
+
# Exllama v2 mlabonne/NeuralDaredevil-8B-abliterated
|
| 6 |
+
|
| 7 |
+
Using <a href="https://github.com/turboderp/exllamav2/releases/tag/v0.0.21">turboderp's ExLlamaV2 v0.0.21</a> for quantization.
|
| 8 |
+
|
| 9 |
+
<b>The "main" branch only contains the measurement.json, download one of the other branches for the model</b>
|
| 10 |
+
|
| 11 |
+
Each branch contains an individual bits per weight, with the main one containing only the meaurement.json for further conversions.
|
| 12 |
+
|
| 13 |
+
Original model: <a href="https://huggingface.co/mlabonne/NeuralDaredevil-8B-abliterated">mlabonne/NeuralDaredevil-8B-abliterated</a><br>
|
| 14 |
+
Calibration dataset: <a href="https://huggingface.co/datasets/cosmicvalor/toxic-qna">toxic-qna</a>
|
| 15 |
+
|
| 16 |
+
## Available sizes
|
| 17 |
+
|
| 18 |
+
| Branch | Bits | lm_head bits | Description |
|
| 19 |
+
| ----- | ---- | ------- | ------------ |
|
| 20 |
+
| [8_0](https://huggingface.co/Apel-sin/llama-3-NeuralDaredevil-8B-abliterated-exl2/tree/8_0) | 8.0 | 8.0 | Maximum quality that ExLlamaV2 can produce, near unquantized performance. |
|
| 21 |
+
| [6_5](https://huggingface.co/Apel-sin/llama-3-NeuralDaredevil-8B-abliterated-exl2/tree/6_5) | 6.5 | 8.0 | Very similar to 8.0, good tradeoff of size vs performance, **recommended**. |
|
| 22 |
+
| [5_5](https://huggingface.co/Apel-sin/llama-3-NeuralDaredevil-8B-abliterated-exl2/tree/5_5) | 5.5 | 8.0 | Slightly lower quality vs 6.5, but usable on 8GB cards. |
|
| 23 |
+
|
| 24 |
# Llama-3-8B-Instruct-abliterated-v3 Model Card
|
| 25 |
|
| 26 |
[My Jupyter "cookbook" to replicate the methodology can be found here, refined library coming soon](https://huggingface.co/failspy/llama-3-70B-Instruct-abliterated/blob/main/ortho_cookbook.ipynb)
|