From bfb443a6b6f670e68e211112a141d089f1cf956f Mon Sep 17 00:00:00 2001 From: Wentao Ye <44945378+yewentao256@users.noreply.github.com> Date: Tue, 8 Sep 2026 17:07:54 -0400 Subject: [PATCH] [Kimi Bug] Fix kda ima `Triton Error [CUDA]: an illegal memory access was encountered` (#55924) Signed-off-by: yewentao256 --- vllm/models/kimi_k3/nvidia/kda.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/vllm/models/kimi_k3/nvidia/kda.py b/vllm/models/kimi_k3/nvidia/kda.py index c5c556d745..52039acab1 100644 --- a/vllm/models/kimi_k3/nvidia/kda.py +++ b/vllm/models/kimi_k3/nvidia/kda.py @@ -415,7 +415,7 @@ def _store_cache_checkpoints_kernel( # store checkpoints to cache seq_idx = tl.program_id(0) cols = tl.program_id(1) * BLOCK_SIZE + tl.arange(0, BLOCK_SIZE) - state_idx = tl.load(checkpoint_state_indices_ptr + seq_idx) + state_idx = tl.load(checkpoint_state_indices_ptr + seq_idx).to(tl.int64) checkpoint_offset = tl.load( checkpoint_offsets_ptr + seq_idx * checkpoint_offset_stride )