[Bugfix] Support triton==3.3.0+git95326d9f for RTX 5090 (Unsloth + vLLM compatibility) (#15471)

Co-authored-by: ServerAI <ai@exc-mad-ai.com>

vllm/lora/ops/triton_ops/kernel_utils.py

@@ -130,7 +130,7 @@ def do_expand_kernel(
     # Identify A and B block pointers
     offset_k = tl.arange(0, BLOCK_K)
     a_ptr = (cur_input_ptr + ram[:, None] * input_d1_stride +
-             offset_k[None, :] * input_d2_stride, )
+             offset_k[None, :] * input_d2_stride)
     b_ptr = (cur_lora_ptr + cur_lora_d0_stride * lora_index +
              offset_k[:, None] * cur_lora_d2_stride +
              rbn[None, :] * cur_lora_d1_stride)