Initial: Stheno v3.4 WebGPU
Browse files
README.md
ADDED
|
@@ -0,0 +1,33 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: other
|
| 3 |
+
base_model: Sao10K/Llama-3.1-8B-Stheno-v3.4
|
| 4 |
+
tags:
|
| 5 |
+
- stheno
|
| 6 |
+
- llama-3.1
|
| 7 |
+
- webgpu
|
| 8 |
+
- browser-inference
|
| 9 |
+
- strix-halo
|
| 10 |
+
- amd
|
| 11 |
+
- unified-memory
|
| 12 |
+
- character-voice
|
| 13 |
+
pipeline_tag: text-generation
|
| 14 |
+
---
|
| 15 |
+
|
| 16 |
+
# Stheno v3.4 on WebGPU
|
| 17 |
+
|
| 18 |
+
Sao10K Stheno v3.4 (Llama-3.1-8B) running in browser via WebGPU. Upgrade from the v3.2 MLC package. 4.6 GB Q4_K_M.
|
| 19 |
+
|
| 20 |
+
The voice model behind the Garden entity system. Character embodiment, creative writing, entity voice.
|
| 21 |
+
|
| 22 |
+
## Quick Start
|
| 23 |
+
|
| 24 |
+
1. Download Q4_K_M GGUF from bartowski
|
| 25 |
+
2. Split: llama-gguf-split --split --split-max-size 1G
|
| 26 |
+
3. Place in model_splits/
|
| 27 |
+
4. node serve.js (port 8210)
|
| 28 |
+
5. Open http://localhost:8210
|
| 29 |
+
|
| 30 |
+
## Credits
|
| 31 |
+
|
| 32 |
+
Built by Joshua (LJTSG) and Claude. Model by Sao10K.
|
| 33 |
+
Co-Authored-By: Claude <noreply@anthropic.com>
|