Introduce cpu agents for general-purpose tasks that don't require TPU (#1012)

Signed-off-by: Johnny Yang <johnnyyang@google.com>

Author: jcyang43

.buildkite/features/Chunked_Prefill.yml

@@ -14,7 +14,7 @@ steps:
       CI_TARGET: "Chunked Prefill"
       CI_STAGE: "CorrectnessTest"
     agents:
-      queue: tpu_v6e_queue
+      queue: cpu
     commands:
       - |
         .buildkite/scripts/record_step_result.sh Chunked_Prefill_CorrectnessTest
@@ -34,7 +34,7 @@ steps:
       CI_TARGET: "Chunked Prefill"
       CI_STAGE: "PerformanceTest"
     agents:
-      queue: tpu_v6e_queue
+      queue: cpu
     commands:
       - |
         .buildkite/scripts/record_step_result.sh Chunked_Prefill_PerformanceTest

.buildkite/features/Collective_Communication_Matmul.yml

@@ -14,7 +14,7 @@ steps:
       CI_TARGET: "Collective Communication Matmul"
       CI_STAGE: "CorrectnessTest"
     agents:
-      queue: tpu_v6e_queue
+      queue: cpu
     commands:
       - |
         .buildkite/scripts/record_step_result.sh Collective_Communication_Matmul_CorrectnessTest

.buildkite/features/JAX-Path_Qxix_Quantization.yml

@@ -14,7 +14,7 @@ steps:
       CI_TARGET: "JAX-Path Qxix Quantization"
       CI_STAGE: "CorrectnessTest"
     agents:
-      queue: tpu_v6e_queue
+      queue: cpu
     commands:
       - |
         .buildkite/scripts/record_step_result.sh JAX-Path_Qxix_Quantization_CorrectnessTest
@@ -36,7 +36,7 @@ steps:
       CI_TARGET: "JAX-Path Qxix Quantization"
       CI_STAGE: "PerformanceTest"
     agents:
-      queue: tpu_v6e_queue
+      queue: cpu
     commands:
       - |
         .buildkite/scripts/record_step_result.sh JAX-Path_Qxix_Quantization_PerformanceTest

.buildkite/features/Multimodal_Inputs.yml

@@ -14,7 +14,7 @@ steps:
       CI_TARGET: Multimodal Inputs
       CI_STAGE: "CorrectnessTest"
     agents:
-      queue: tpu_v6e_queue
+      queue: cpu
     commands:
       - |
         .buildkite/scripts/record_step_result.sh Multimodal_Inputs_CorrectnessTest
@@ -34,7 +34,7 @@ steps:
       CI_TARGET: Multimodal Inputs
       CI_STAGE: "PerformanceTest"
     agents:
-      queue: tpu_v6e_queue
+      queue: cpu
     commands:
       - |
         .buildkite/scripts/record_step_result.sh Multimodal_Inputs_PerformanceTest

.buildkite/features/Prefix_Caching.yml

@@ -4,7 +4,7 @@ steps:
     key: "Prefix_Caching_CorrectnessTest"
     soft_fail: true
     agents:
-      queue: tpu_v6e_queue
+      queue: cpu
     commands:
       - echo "default on in vLLM"
   - label: "Record correctness test result for Prefix Caching"
@@ -14,7 +14,7 @@ steps:
       CI_TARGET: "Prefix Caching"
       CI_STAGE: "CorrectnessTest"
     agents:
-      queue: tpu_v6e_queue
+      queue: cpu
     commands:
       - |
         .buildkite/scripts/record_step_result.sh Prefix_Caching_CorrectnessTest
@@ -24,7 +24,7 @@ steps:
     depends_on: "record_Prefix_Caching_CorrectnessTest"
     soft_fail: true
     agents:
-      queue: tpu_v6e_queue
+      queue: cpu
     commands:
       - echo "default on in vLLM"
   - label: "Record performance test result for Prefix Caching"
@@ -34,7 +34,7 @@ steps:
       CI_TARGET: "Prefix Caching"
       CI_STAGE: "PerformanceTest"
     agents:
-      queue: tpu_v6e_queue
+      queue: cpu
     commands:
       - |
         .buildkite/scripts/record_step_result.sh Prefix_Caching_PerformanceTest

.buildkite/features/Quantized_Matmul_Attention_and_KV_Cache.yml

@@ -4,7 +4,7 @@ steps:
     key: "Quantized_Matmul_Attention_and_KV_Cache_CorrectnessTest"
     soft_fail: true
     agents:
-      queue: tpu_v6e_queue
+      queue: cpu
     commands:
       - echo "covered by performance test"
   - label: "Record correctness test result for Quantized Matmul Attention and KV Cache"
@@ -14,7 +14,7 @@ steps:
       CI_TARGET: "Quantized Matmul Attention and KV Cache"
       CI_STAGE: "CorrectnessTest"
     agents:
-      queue: tpu_v6e_queue
+      queue: cpu
     commands:
       - |
         .buildkite/scripts/record_step_result.sh Quantized_Matmul_Attention_and_KV_Cache_CorrectnessTest
@@ -45,7 +45,7 @@ steps:
       CI_TARGET: "Quantized Matmul Attention and KV Cache"
       CI_STAGE: "PerformanceTest"
     agents:
-      queue: tpu_v6e_queue
+      queue: cpu
     commands:
       - |
         .buildkite/scripts/record_step_result.sh Quantized_Matmul_Attention_and_KV_Cache_PerformanceTest

.buildkite/features/Ragged_Paged_Attention_V3.yml

@@ -4,7 +4,7 @@ steps:
     key: "Ragged_Paged_Attention_V3_CorrectnessTest"
     soft_fail: true
     agents:
-      queue: tpu_v6e_queue
+      queue: cpu
     commands:
       - echo "default on in vLLM"
   - label: "Record correctness test result for Ragged Paged Attention V3"
@@ -14,7 +14,7 @@ steps:
       CI_TARGET: "Ragged Paged Attention V3"
       CI_STAGE: "CorrectnessTest"
     agents:
-      queue: tpu_v6e_queue
+      queue: cpu
     commands:
       - |
         .buildkite/scripts/record_step_result.sh Ragged_Paged_Attention_V3_CorrectnessTest
@@ -24,7 +24,7 @@ steps:
     depends_on: "record_Ragged_Paged_Attention_V3_CorrectnessTest"
     soft_fail: true
     agents:
-      queue: tpu_v6e_queue
+      queue: cpu
     commands:
       - echo "default on in vLLM"
   - label: "Record performance test result for Ragged Paged Attention V3"
@@ -34,7 +34,7 @@ steps:
       CI_TARGET: "Ragged Paged Attention V3"
       CI_STAGE: "PerformanceTest"
     agents:
-      queue: tpu_v6e_queue
+      queue: cpu
     commands:
       - |
         .buildkite/scripts/record_step_result.sh Ragged_Paged_Attention_V3_PerformanceTest

.buildkite/features/Single_Program_Multi_Data.yml

@@ -4,7 +4,7 @@ steps:
     key: "Single_Program_Multi_Data__CorrectnessTest"
     soft_fail: true
     agents:
-      queue: tpu_v6e_queue
+      queue: cpu
     commands:
       - echo "default on in vLLM"
   - label: "Record correctness test result for Single Program Multi Data"
@@ -14,7 +14,7 @@ steps:
       CI_TARGET: "Single Program Multi Data"
       CI_STAGE: "CorrectnessTest"
     agents:
-      queue: tpu_v6e_queue
+      queue: cpu
     commands:
       - |
         .buildkite/scripts/record_step_result.sh Single_Program_Multi_Data__CorrectnessTest
@@ -24,7 +24,7 @@ steps:
     depends_on: "record_Single_Program_Multi_Data__CorrectnessTest"
     soft_fail: true
     agents:
-      queue: tpu_v6e_queue
+      queue: cpu
     commands:
       - echo "default on in vLLM"
   - label: "Record performance test result for Single Program Multi Data"
@@ -34,7 +34,7 @@ steps:
       CI_TARGET: "Single Program Multi Data"
       CI_STAGE: "PerformanceTest"
     agents:
-      queue: tpu_v6e_queue
+      queue: cpu
     commands:
       - |
         .buildkite/scripts/record_step_result.sh Single_Program_Multi_Data__PerformanceTest

.buildkite/features/Speculative_Decoding-_Ngram.yml

@@ -4,7 +4,7 @@ steps:
     key: "Speculative_Decoding-_Ngram_CorrectnessTest"
     soft_fail: true
     agents:
-      queue: tpu_v6e_queue
+      queue: cpu
     commands:
       - echo "covered by performance test"
   - label: "Record correctness test result for Speculative Decoding: Ngram"
@@ -14,7 +14,7 @@ steps:
       CI_TARGET: "Speculative Decoding: Ngram"
       CI_STAGE: "CorrectnessTest"
     agents:
-      queue: tpu_v6e_queue
+      queue: cpu
     commands:
       - |
         .buildkite/scripts/record_step_result.sh Speculative_Decoding-_Ngram_CorrectnessTest
@@ -24,7 +24,7 @@ steps:
     depends_on: "record_Speculative_Decoding-_Ngram_CorrectnessTest"
     soft_fail: true
     agents:
-      queue: tpu_v6e_queue
+      queue: cpu
     commands:
       - .buildkite/scripts/run_in_docker.sh python3 -m pytest -s -v /workspace/tpu_inference/tests/e2e/test_speculative_decoding.py
   - label: "Record performance test result for Speculative Decoding: Ngram"
@@ -34,7 +34,7 @@ steps:
       CI_TARGET: "Speculative Decoding: Ngram"
       CI_STAGE: "PerformanceTest"
     agents:
-      queue: tpu_v6e_queue
+      queue: cpu
     commands:
       - |
         .buildkite/scripts/record_step_result.sh Speculative_Decoding-_Ngram_PerformanceTest

.buildkite/features/Structured_Decoding.yml

@@ -14,7 +14,7 @@ steps:
       CI_TARGET: Structured Decoding
       CI_STAGE: "CorrectnessTest"
     agents:
-      queue: tpu_v6e_queue
+      queue: cpu
     commands:
       - |
         .buildkite/scripts/record_step_result.sh Structured_Decoding_CorrectnessTest