[TPU][V1] Remove ragged attention kernel parameter hard coding (#16041)

Signed-off-by: Chengji Yao <[email protected]>

Author: yaochengji

vllm/v1/attention/backends/pallas.py

@@ -11,10 +11,6 @@
                                               AttentionLayer, AttentionType)
 from vllm.attention.backends.utils import CommonAttentionState
 
-# These are the 2 tunable parameters of the paged attention Pallas kernel.
-NUM_QUERIES_PER_BLOCK = 32
-NUM_KV_PAGES_PER_BLOCK = 128
-
 
 class PallasAttentionBackend(AttentionBackend):
 
@@ -115,13 +111,6 @@ def __init__(
         tpu_version = torch_xla.tpu.version()
         if tpu_version < 4:
             raise NotImplementedError("TPU version must be 4 or higher.")
-        # NOTE(chengjiyao): the TPU v4's vmem capacity is 16MB
-        # TODO(chengjiyao): autotune NUM_QUERIES_PER_BLOCK,
-        # NUM_KV_PAGES_PER_BLOCK and vmem_limit_bytes
-        if tpu_version == 4:
-            self.vmem_limit_bytes = 16 * 1024 * 1024
-        else:
-            self.vmem_limit_bytes = 64 * 1024 * 1024
 
     def forward(
         self,
@@ -165,9 +154,12 @@ def forward(
             attn_metadata.block_tables,
             attn_metadata.query_start_loc,
             attn_metadata.num_seqs,
-            num_kv_pages_per_block=NUM_KV_PAGES_PER_BLOCK,
-            num_queries_per_block=NUM_QUERIES_PER_BLOCK,
-            vmem_limit_bytes=self.vmem_limit_bytes,
+            # By default, the system utilizes optimized block size and
+            # vmem_limit_bytes parameters from the kernel repository. However,
+            # these can be manually adjusted for debugging if necessary.
+            num_kv_pages_per_block=None,
+            num_queries_per_block=None,
+            vmem_limit_bytes=None,
             use_kernel=True,
             sm_scale=self.scale,
             sliding_window=self.sliding_window,

vllm/v1/worker/tpu_model_runner.py

@@ -24,8 +24,7 @@
 from vllm.sampling_params import SamplingType
 from vllm.sequence import IntermediateTensors
 from vllm.utils import LayerBlockType, cdiv, is_pin_memory_available
-from vllm.v1.attention.backends.pallas import (NUM_KV_PAGES_PER_BLOCK,
-                                               PallasAttentionBackend,
+from vllm.v1.attention.backends.pallas import (PallasAttentionBackend,
                                                PallasMetadata)
 from vllm.v1.core.encoder_cache_manager import compute_encoder_budget
 from vllm.v1.kv_cache_interface import (FullAttentionSpec, KVCacheConfig,
@@ -155,11 +154,8 @@ def __init__(
                                             dtype=torch.int64,
                                             device="cpu")
         self.slot_mapping_np = self.slot_mapping_cpu.numpy()
-
-        padded_max_num_blocks_per_req = _get_padded_number(
-            self.max_num_blocks_per_req, NUM_KV_PAGES_PER_BLOCK)
         self.block_table_cpu = torch.zeros(
-            (self.max_num_tokens, padded_max_num_blocks_per_req),
+            (self.max_num_tokens, self.max_num_blocks_per_req),
             dtype=self.input_batch.block_table.get_cpu_tensor().dtype,
             device="cpu")