[Model Runner V2] Support VLM (#32546)

Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
This commit is contained in:
Woosuk Kwon
2026-01-18 16:58:51 -08:00
committed by GitHub
parent 6101a26dc9
commit bb1848cd62
6 changed files with 263 additions and 15 deletions
-3
View File
@@ -48,9 +48,6 @@ class EagleSpeculator:
self.input_buffers = InputBuffers(
max_num_reqs=self.max_num_reqs,
max_num_tokens=self.max_num_tokens,
inputs_embeds_size=self.inputs_embeds_size,
vocab_size=self.vocab_size,
dtype=self.dtype,
device=device,
)
self.hidden_states = torch.zeros(