- Rewrite bridge.py to use make_sampler() and make_logits_processors() instead of deprecated direct kwargs (temp, top_p, top_k) - Add repeat_penalty forwarding in backend.go for Generate and Chat - Extract _build_gen_kwargs() helper shared by generate and chat handlers Co-Authored-By: Virgil <virgil@lethean.io> |
||
|---|---|---|
| .. | ||
| testdata | ||
| backend.go | ||
| backend_test.go | ||
| bridge.py | ||