tessella-moe-nano
Qwen2-MoE model trained on The Adventures of Sherlock Holmes for Tessella demos.
| Parameter | Value |
|---|---|
| Architecture | Qwen2-MoE |
| Layers | 4 |
| Experts | 8 (2 active/token) |
| Hidden size | 128 |
| Parameters | ~4M |
| Size | ~8MB F16 |
| Shards | 1 |
| Training data | The Adventures of Sherlock Holmes |
# worker.toml
default_model = "victorespada/tessella-moe-nano"
- Downloads last month
- 4
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support