mirror of
https://github.com/huggingface/text-generation-inference.git
synced 2025-04-22 15:32:08 +00:00
c1638a56f1
Deepseek V2 is a MoE model from Deepseek. Relevant variations
compared to other models:
- Grouped top-K in expert selection.
- mscale in yarn is calculated using the `mscale` and `mscale_all_dim`
configuration options.
- `mscale_all_dim` is also used in scaling attention softmax.
- Permuting of the query/key representations before applying rotary
embeddings.
- Some projections cannot be sharded (`q_a_proj`, `kv_a_proj_with_mqa`).
So, we need weight loads that supports quantized weights. To this
end `{Weights,WeightLoader}.get_weight` was added.
- The query/key head dimensionality differs from that of the value,
so we need to pad during attention.
- Heads with size 192, needs an extension to our paged attention
fork and we need to ensure that the KV cache is allocated with the
correct size.
- Shared experts.
|
||
|---|---|---|
| .. | ||
| adapters | ||
| layers | ||
| models | ||
| pb | ||
| utils | ||
| __init__.py | ||
| cache.py | ||
| cli.py | ||
| interceptor.py | ||
| server.py | ||
| tracing.py | ||