Commit
•
00935fb
1
Parent(s):
4cfaeb8
Update README.md
Browse files
README.md
CHANGED
|
@@ -4,13 +4,54 @@ language:
|
|
| 4 |
- en
|
| 5 |
---
|
| 6 |
|
| 7 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 8 |
# "Hathor-v0.1 is a model based on the LLaMA 3 architecture: Designed to seamlessly integrate the qualities of creativity, intelligence, and robust performance. Making it an ideal tool for a wide range of applications; such as creative writing, educational support and human/computer interaction."
|
| 9 |
|
| 10 |
# Recomended ST Presets: [Hathor Presets](https://huggingface.co/Nitral-AI/Hathor-L3-8B-v.01/tree/main/Hathor%20Presets)
|
| 11 |
|
| 12 |
---
|
|
|
|
| 13 |
# Notes: Hathor is trained on 3 epochs of private rp data, synthetic opus instructons, a mix of light/classical novel data. (Heavily wip)
|
|
|
|
| 14 |
---
|
| 15 |
|
| 16 |
- If you want to use vision functionality:
|
|
@@ -22,4 +63,3 @@ language:
|
|
| 22 |
* You can load the **mmproj** by using the corresponding section in the interface:
|
| 23 |
|
| 24 |
|
| 25 |
-
---
|
|
|
|
| 4 |
- en
|
| 5 |
---
|
| 6 |
|
| 7 |
+
My GGUF-IQ-Imatrix quants for [**Nitral-AI/Hathor-L3-8B-v.01**](https://huggingface.co/Nitral-AI/Hathor-L3-8B-v.01).
|
| 8 |
+
|
| 9 |
+
"Designed to seamlessly integrate the qualities of creativity, intelligence, and robust performance."
|
| 10 |
+
|
| 11 |
+
> [!IMPORTANT]
|
| 12 |
+
> **Quantization process:** <br>
|
| 13 |
+
> For future reference, these quants have been done after the fixes from [**#6920**](https://github.com/ggerganov/llama.cpp/pull/6920) have been merged. <br>
|
| 14 |
+
> Imatrix data was generated from the FP16-GGUF and conversions directly from the BF16-GGUF. <br>
|
| 15 |
+
> This was a bit more disk and compute intensive but hopefully avoided any losses during conversion. <br>
|
| 16 |
+
> If you noticed any issues let me know in the discussions.
|
| 17 |
+
|
| 18 |
+
> [!NOTE]
|
| 19 |
+
> **General usage:** <br>
|
| 20 |
+
> Use the latest version of **KoboldCpp**. <br>
|
| 21 |
+
> Remember that you can also use `--flashattention` on KoboldCpp now even with non-RTX cards for reduced VRAM usage. <br>
|
| 22 |
+
> For **8GB VRAM** GPUs, I recommend the **Q4_K_M-imat** quant for up to 12288 context sizes. <br>
|
| 23 |
+
> For **12GB VRAM** GPUs, the **Q5_K_M-imat** quant will give you a great size/quality balance. <br>
|
| 24 |
+
>
|
| 25 |
+
> **Resources:** <br>
|
| 26 |
+
> You can find out more about how each quant stacks up against each other and their types [**here**](https://gist.github.com/Artefact2/b5f810600771265fc1e39442288e8ec9) and [**here**](https://rentry.org/llama-cpp-quants-or-fine-ill-do-it-myself-then-pt-2), respectively.
|
| 27 |
+
>
|
| 28 |
+
> **Presets:** <br>
|
| 29 |
+
> Some compatible SillyTavern presets can be found [**here (Hathor Presets)**](https://huggingface.co/Nitral-AI/Hathor-L3-8B-v.01/tree/main/Hathor%20Presets) or [**here (Virt's Roleplay Presets)**](https://huggingface.co/Virt-io/SillyTavern-Presets). <br>
|
| 30 |
+
<!-- > Check [**discussions such as this one**](https://huggingface.co/Virt-io/SillyTavern-Presets/discussions/5#664d6fb87c563d4d95151baa) for other recommendations and samplers.
|
| 31 |
+
-->
|
| 32 |
+
|
| 33 |
+
> [!TIP]
|
| 34 |
+
> **Personal-support:** <br>
|
| 35 |
+
> I apologize for disrupting your experience. <br>
|
| 36 |
+
> Currently I'm working on moving for a better internet provider. <br>
|
| 37 |
+
> If you **want** and you are **able to**... <br>
|
| 38 |
+
> You can [**spare some change over here (Ko-fi)**](https://ko-fi.com/Lewdiculous). <br>
|
| 39 |
+
>
|
| 40 |
+
> **Author-support:** <br>
|
| 41 |
+
> You can support the author [**at their own page**](https://huggingface.co/Nitral-AI).
|
| 42 |
+
|
| 43 |
+
|
| 44 |
+
|
| 45 |
+
## **Original model text information:**
|
| 46 |
+
|
| 47 |
# "Hathor-v0.1 is a model based on the LLaMA 3 architecture: Designed to seamlessly integrate the qualities of creativity, intelligence, and robust performance. Making it an ideal tool for a wide range of applications; such as creative writing, educational support and human/computer interaction."
|
| 48 |
|
| 49 |
# Recomended ST Presets: [Hathor Presets](https://huggingface.co/Nitral-AI/Hathor-L3-8B-v.01/tree/main/Hathor%20Presets)
|
| 50 |
|
| 51 |
---
|
| 52 |
+
|
| 53 |
# Notes: Hathor is trained on 3 epochs of private rp data, synthetic opus instructons, a mix of light/classical novel data. (Heavily wip)
|
| 54 |
+
|
| 55 |
---
|
| 56 |
|
| 57 |
- If you want to use vision functionality:
|
|
|
|
| 63 |
* You can load the **mmproj** by using the corresponding section in the interface:
|
| 64 |
|
| 65 |
|
|
|