YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
deepseek-v3.2: 2 representative decoder layers
Source: deepseek-ai/DeepSeek-V3.2.
Original dimensions, context/RoPE settings and precision metadata are retained. This directory contains no weights; use SGLang's native implementation with --load-format dummy.
| Layer | Source layer | Attention | FFN | Routing |
|---|---|---|---|---|
| 0 | 0 | DSA | dense | none |
| 1 | 3 | DSA | MoE | noaux_tc |
MTP draft layers retained: 1 (separate from the decoder). MTP requires a speculative draft-and-verify execution path; ordinary one_batch decoding does not run it.
- Downloads last month
- 43
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support