[TPU] [V1] fix cases when max_num_reqs is set smaller than MIN_NUM_SEQS (#15583)

Signed-off-by: Chengji Yao <chengjiyao@google.com>

examples/offline_inference/tpu.py

@@ -14,10 +14,7 @@ answers = [
 ]
 N = 1
 # Currently, top-p sampling is disabled. `top_p` should be 1.0.
-sampling_params = SamplingParams(temperature=0.7,
-                                 top_p=1.0,
-                                 n=N,
-                                 max_tokens=16)
+sampling_params = SamplingParams(temperature=0, top_p=1.0, n=N, max_tokens=16)
 
 # Set `enforce_eager=True` to avoid ahead-of-time compilation.
 # In real workloads, `enforace_eager` should be `False`.

vllm/v1/worker/tpu_model_runner.py

@@ -88,7 +88,7 @@ class TPUModelRunner:
         self.max_model_len = model_config.max_model_len
         self.max_num_blocks_per_req = cdiv(self.max_model_len, self.block_size)
         self.max_num_tokens = scheduler_config.max_num_batched_tokens
-        self.max_num_reqs = scheduler_config.max_num_seqs
+        self.max_num_reqs = max(scheduler_config.max_num_seqs, MIN_NUM_SEQS)
 
         # Model-related.
         self.num_attn_layers = model_config.get_num_layers_by_block_type(