Skip to content

+text 模式下的pending logit问题 #60

Description

@ljjTYJR

if mode == "TEXT":
# text-only mode: get logits from the last token
self.pending_logits, _ = self.decoder.feed(text_embeds, return_logits=True)
else:
# mixed mode: just feed, let other modality get logits
self.decoder.feed(text_embeds)

你好,
请问一下如果是 video/audio + text的prefill情况下,pending_logit 不是依据最新的kv cache解码出,而仍然是其他模态下解码出来的,这么设计有什么原因么?

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions