Add README for muffin model
Browse files- muffin/README.md +24 -0
muffin/README.md
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# muffin
|
| 2 |
+
|
| 3 |
+
## GenAI WebGPU Model
|
| 4 |
+
|
| 5 |
+
Extracted from Microsoft Edge Canary's on-device model.
|
| 6 |
+
|
| 7 |
+
- **Model name**: Muffin
|
| 8 |
+
- **Architecture**: Qwen3 (28 layers, 24 attention heads, 8 KV heads, hidden size 2048)
|
| 9 |
+
- **Precision**: int4
|
| 10 |
+
- **Execution Provider**: webgpu
|
| 11 |
+
- **Context length**: 8192
|
| 12 |
+
- **Vocab size**: 200029
|
| 13 |
+
|
| 14 |
+
### How to obtain this model
|
| 15 |
+
|
| 16 |
+
1. Install [Microsoft Edge Canary](https://www.microsoft.com/en-us/edge/download/canary)
|
| 17 |
+
2. Enable the following flags in `edge://flags`:
|
| 18 |
+
- `edge://flags/#edge-llm-prompt-api-for-phi-mini`
|
| 19 |
+
- `edge://flags/#edge-llm-on-device-model-prerelease`
|
| 20 |
+
3. Use the [Prompt API](https://learn.microsoft.com/en-us/microsoft-edge/web-platform/prompt-api) in a webpage — Edge will automatically download the model
|
| 21 |
+
4. The downloaded model files can be found at:
|
| 22 |
+
```
|
| 23 |
+
C:\Users\<username>\AppData\Local\Microsoft\Edge SxS\User Data\EdgeLLMOnDeviceModel\
|
| 24 |
+
```
|