Update to new version of base image

vllm/attention/ops/paged_attn.py

@@ -4,7 +4,10 @@ from typing import List, Optional, Tuple
 import torch
 
 from vllm import _custom_ops as ops
-from vllm.attention.ops.prefix_prefill import context_attention_fwd
+from vllm.triton_utils import HAS_TRITON
+
+if HAS_TRITON:
+    from vllm.attention.ops.prefix_prefill import context_attention_fwd
 
 # Should be the same as PARTITION_SIZE in `paged_attention_v2_launcher`.
 _PARTITION_SIZE = 512

vllm/attention/ops/prefix_prefill.py

@@ -808,8 +808,6 @@ if triton.__version__ >= "2.1.0":
             )
             return
 
-        import time
-        ts_beg = time.time()
         _fwd_kernel[grid](
             q,
             k,
@@ -860,6 +858,4 @@ if triton.__version__ >= "2.1.0":
             num_warps=NUM_WARPS,
             num_stages=1,
         )
-        elapsed = time.time() - ts_beg
-        #print(f'{elapsed}: {BLOCK=}, {Lk=}, {Lk_padded=}, {BLOCK=}, {sliding_window=}, {NUM_WARPS=}')
         return