Commit 7189cfe
committed
feat: Add Eagle3 speculative decoding support for Llama4
- Add Eagle3Llama4ForCausalLM model implementation
- Add SupportsEagle3 interface to Llama4ForConditionalGeneration
- Update eagle.py to support both Llama and Llama4 Eagle3 models
- Register Eagle3Llama4ForCausalLM in model registry
Signed-off-by: Rahul Tuli <rtuli@redhat.com>1 parent 0dc9532 commit 7189cfe
File tree
4 files changed
+569
-4
lines changed- vllm
- model_executor/models
- v1/spec_decode
4 files changed
+569
-4
lines changed
0 commit comments