Remove redundant architecture prose and restore standard introduction
Browse files
README.md
CHANGED
|
@@ -3,9 +3,7 @@ base_model:
|
|
| 3 |
- meta-models/Muse-Spark-Open-816B-A42B
|
| 4 |
---
|
| 5 |
|
| 6 |
-
This tiny model is
|
| 7 |
-
|
| 8 |
-
Built with Muse Spark Open. Random weights preserve sigmoid top-8 routed MoE, expert pooling, centered norms, residual/attention gates, sliding RoPE and full NoPE attention, NaViT image/video encoding, and the four-stage waveform audio encoder. Vocabulary rows and source precision are unchanged.
|
| 9 |
|
| 10 |
| Feature | Source | Tiny |
|
| 11 |
|---|---|---|
|
|
|
|
| 3 |
- meta-models/Muse-Spark-Open-816B-A42B
|
| 4 |
---
|
| 5 |
|
| 6 |
+
This tiny model is intended for debugging. It is randomly initialized using the configuration adapted from [meta-models/Muse-Spark-Open-816B-A42B](https://huggingface.co/meta-models/Muse-Spark-Open-816B-A42B).
|
|
|
|
|
|
|
| 7 |
|
| 8 |
| Feature | Source | Tiny |
|
| 9 |
|---|---|---|
|