[Feature] Layer-wise Prefill (#7634)

Signed-off-by: jason-fxz <jason341132@qq.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
This commit is contained in:
Xiaoze Fan
2025-07-17 01:57:46 +08:00
committed by GitHub
parent d9eb5efc71
commit 570d33437b
13 changed files with 464 additions and 2 deletions

View File

@@ -500,6 +500,7 @@ class TboForwardBatchPreparer:
"capture_hidden_mode",
"padded_static_len",
"mrope_positions", # only used by qwen2-vl, thus not care
"split_index", # for split prefill
]:
output_dict[key] = getattr(batch, key)
if not batch.forward_mode.is_target_verify():