[TRTLLM-13642][feat] Add perf sanity tests for Llama-3.1-8B and Gemma-3-1B and verify cache transceiver V2 support

tests/integration/defs/perf/_model_paths.py

@@ -44,6 +44,7 @@
     "gemma_3_12b_it": "gemma/gemma-3-12b-it",
     "gemma_3_12b_it_fp8": "gemma/gemma-3-12b-it-fp8",
     "gemma_3_12b_it_fp4": "gemma/gemma-3-12b-it-fp4",
+    "gemma_3_1b_it": "gemma/gemma-3-1b-it",
     "deepseek_r1_fp8": "DeepSeek-R1/DeepSeek-R1",
     "deepseek_r1_nvfp4": "DeepSeek-R1/DeepSeek-R1-FP4",
     "deepseek_r1_0528_fp8": "DeepSeek-R1/DeepSeek-R1-0528/",

tests/scripts/perf-sanity/disaggregated/gb200_gemma-3-1b-bf16_1k1k_con256_ctx1_tp1_gen1_tp1_eplb0_mtp0_ccb-NIXL.yaml

@@ -0,0 +1,103 @@
+# SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+metadata:
+  model_name: gemma_3_1b_it
+  precision: bf16
+  model_dir_name: gemma-3-1b-it
+  supported_gpus:
+  - GB200
+  script_file: disaggr_torch.slurm
+  benchmark_type: 1k1k
+slurm:
+  script_file: disaggr_torch.slurm
+  partition: <partition>
+  account: <account>
+  job_time: 02:00:00
+  job_name: unified-benchmark
+  extra_args: --gres=gpu:4
+  numa_bind: true
+benchmark:
+  mode: e2e
+  use_nv_sa_benchmark: false
+  multi_round: 10
+  benchmark_ratio: 0.0
+  streaming: true
+  concurrency_list: '256'
+  input_length: 1024
+  output_length: 1024
+  dataset_file: <dataset_file>
+hardware:
+  gpus_per_node: 4
+  num_ctx_servers: 1
+  num_gen_servers: 1
+environment:
+  container_mount: <container_mount>
+  container_image: <container_image>
+  model_path: <model_path>
+  trtllm_repo: ''
+  build_wheel: false
+  work_dir: <full_path_to_work_dir>
+  worker_env_var: TLLM_LOG_LEVEL=INFO TRTLLM_SERVER_DISABLE_GC=1 TRTLLM_WORKER_DISABLE_GC=1 ENROOT_ALLOW_DEV=yes
+  server_env_var: TRTLLM_SERVER_DISABLE_GC=1
+profiling:
+  nsys_on: false
+worker_config:
+  gen:
+    print_iter_log: true
+    max_batch_size: 256
+    max_num_tokens: 512
+    tensor_parallel_size: 1
+    moe_expert_parallel_size: 1
+    pipeline_parallel_size: 1
+    context_parallel_size: 1
+    enable_attention_dp: false
+    cuda_graph_config:
+      enable_padding: true
+      max_batch_size: 256
+    kv_cache_config:
+      use_kv_cache_manager_v2: false
+      enable_block_reuse: false
+      free_gpu_memory_fraction: 0.85
+      dtype: auto
+      max_attention_window: [512, 512, 512, 512, 512, 32768]
+    cache_transceiver_config:
+      max_tokens_in_buffer: 4096
+      backend: NIXL
+      transceiver_runtime: PYTHON
+    disable_overlap_scheduler: false
+    num_postprocess_workers: 4
+    stream_interval: 20
+  ctx:
+    print_iter_log: true
+    max_batch_size: 32
+    max_num_tokens: 8192
+    tensor_parallel_size: 1
+    moe_expert_parallel_size: 1
+    pipeline_parallel_size: 1
+    context_parallel_size: 1
+    enable_attention_dp: false
+    cuda_graph_config: null
+    kv_cache_config:
+      use_kv_cache_manager_v2: false
+      enable_block_reuse: false
+      free_gpu_memory_fraction: 0.85
+      dtype: auto
+      max_attention_window: [512, 512, 512, 512, 512, 32768]
+    cache_transceiver_config:
+      max_tokens_in_buffer: 4096
+      backend: NIXL
+      transceiver_runtime: PYTHON
+    disable_overlap_scheduler: true

tests/scripts/perf-sanity/disaggregated/gb200_llama-3.1-8b-bf16_1k1k_con256_ctx1_tp1_gen1_tp1_eplb0_mtp0_ccb-NIXL.yaml

@@ -0,0 +1,101 @@
+# SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+metadata:
+  model_name: llama_v3.1_8b_instruct
+  precision: bf16
+  model_dir_name: Llama-3.1-8B-Instruct
+  supported_gpus:
+  - GB200
+  script_file: disaggr_torch.slurm
+  benchmark_type: 1k1k
+slurm:
+  script_file: disaggr_torch.slurm
+  partition: <partition>
+  account: <account>
+  job_time: 02:00:00
+  job_name: unified-benchmark
+  extra_args: --gres=gpu:4
+  numa_bind: true
+benchmark:
+  mode: e2e
+  use_nv_sa_benchmark: false
+  multi_round: 10
+  benchmark_ratio: 0.0
+  streaming: true
+  concurrency_list: '256'
+  input_length: 1024
+  output_length: 1024
+  dataset_file: <dataset_file>
+hardware:
+  gpus_per_node: 4
+  num_ctx_servers: 1
+  num_gen_servers: 1
+environment:
+  container_mount: <container_mount>
+  container_image: <container_image>
+  model_path: <model_path>
+  trtllm_repo: ''
+  build_wheel: false
+  work_dir: <full_path_to_work_dir>
+  worker_env_var: TLLM_LOG_LEVEL=INFO TRTLLM_SERVER_DISABLE_GC=1 TRTLLM_WORKER_DISABLE_GC=1 ENROOT_ALLOW_DEV=yes
+  server_env_var: TRTLLM_SERVER_DISABLE_GC=1
+profiling:
+  nsys_on: false
+worker_config:
+  gen:
+    print_iter_log: true
+    max_batch_size: 256
+    max_num_tokens: 512
+    tensor_parallel_size: 1
+    moe_expert_parallel_size: 1
+    pipeline_parallel_size: 1
+    context_parallel_size: 1
+    enable_attention_dp: false
+    cuda_graph_config:
+      enable_padding: true
+      max_batch_size: 256
+    kv_cache_config:
+      use_kv_cache_manager_v2: false
+      enable_block_reuse: false
+      free_gpu_memory_fraction: 0.85
+      dtype: auto
+    cache_transceiver_config:
+      max_tokens_in_buffer: 4096
+      backend: NIXL
+      transceiver_runtime: PYTHON
+    disable_overlap_scheduler: false
+    num_postprocess_workers: 4
+    stream_interval: 20
+  ctx:
+    print_iter_log: true
+    max_batch_size: 32
+    max_num_tokens: 8192
+    tensor_parallel_size: 1
+    moe_expert_parallel_size: 1
+    pipeline_parallel_size: 1
+    context_parallel_size: 1
+    enable_attention_dp: false
+    cuda_graph_config: null
+    kv_cache_config:
+      use_kv_cache_manager_v2: false
+      enable_block_reuse: false
+      free_gpu_memory_fraction: 0.85
+      dtype: auto
+    cache_transceiver_config:
+      max_tokens_in_buffer: 4096
+      backend: NIXL
+      transceiver_runtime: PYTHON
+    disable_overlap_scheduler: true