> This page is for version 1.7.1.
> For other versions, use one of these documentation indexes:
> - 1.8.1 (default): https://docs.nvidia.com/networking/cloudai/1.8.1/llms.txt
> - 1.7.1: https://docs.nvidia.com/networking/cloudai/1.7.1/llms.txt
> - 1.6.1: https://docs.nvidia.com/networking/cloudai/1.6.1/llms.txt
> - dev: https://docs.nvidia.com/networking/cloudai/dev/llms.txt

> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.nvidia.com/networking/cloudai/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.nvidia.com/networking/cloudai/_mcp/server.

This workload (test_template_name is `NixlPerftest`) runs NIXL performance testing suite for comprehensive network performance evaluation.

## Usage Examples

Test TOML example:

```toml
name = "my_nixl_perftest_test"
description = "Example NIXL Perftest test"
test_template_name = "NixlPerftest"

[cmd_args]
docker_image_url = "<docker container url here>"
subtest = "sequential-ct-perftest"
num_user_requests = 1
batch_size = 1
num_prefill_nodes = 1
num_decode_nodes = 1
prefill_tp = 4
decode_tp = 4
isl_mean = 10000
isl_scale = 3000
model = "deepseek-r1-distill-llama-70b"
```

Test Scenario example:

```toml
name = "nixl-perftest-test"

[[Tests]]
id = "perftest.1"
num_nodes = 2
time_limit = "00:20:00"

test_name = "my_nixl_perftest_test"
```

Test-in-Scenario example:

```toml
name = "nixl-perftest-test"

[[Tests]]
id = "perftest.1"
num_nodes = 2
time_limit = "00:20:00"

name = "my_nixl_perftest_test"
description = "Example NIXL Perftest test"
test_template_name = "NixlPerftest"

  [Tests.cmd_args]
  docker_image_url = "<docker container url here>"
  subtest = "sequential-ct-perftest"
  num_user_requests = 100
  batch_size = 1
  num_prefill_nodes = 1
  num_decode_nodes = 1
  prefill_tp = 8
  decode_tp = 8
  model = "deepseek-r1-distill-llama-70b"

  [Tests.extra_env_vars]
  CUDA_VISIBLE_DEVICES = "$SLURM_LOCALID"
```

## API Documentation

### Command Arguments

**&#x20;*pydantic model* cloudai.workloads.nixl_perftest.nixl_perftest.NixlPerftestCmdArgs**

CmdArgs for NIXL Perftest.

**&#x20;*field* subtest*: Literal\['sequential-ct-perftest']* *\[Required]*\*\*

**&#x20;*field* perftest_script*: str = '/workspace/nixl/benchmark/kvbench/main.py'*\*\*

**&#x20;*field* matgen_script*: str = '/workspace/nixl/benchmark/kvbench/test/inference_workload_matgen.py'*\*\*

**&#x20;*field* python_executable*: str = 'python'*\*\*

**&#x20;*field* num_user_requests*: int | list\[int]* *\[Required]*\*\*

**&#x20;*field* batch_size*: int | list\[int]* *\[Required]*\*\*

**&#x20;*field* num_prefill_nodes*: int | list\[int]* *\[Required]*\*\*

**&#x20;*field* num_decode_nodes*: int | list\[int]* *\[Required]*\*\*

**&#x20;*field* isl_mean*: int | list\[int] | None = None*\*\*

**&#x20;*field* isl_scale*: int | list\[int] | None = None*\*\*

**&#x20;*field* prefill_tp*: int | list\[int] = 1*\*\*

**&#x20;*field* prefill_pp*: int | list\[int] = 1*\*\*

**&#x20;*field* prefill_cp*: int | list\[int] = 1*\*\*

**&#x20;*field* decode_tp*: int | list\[int] = 1*\*\*

**&#x20;*field* decode_pp*: int | list\[int] = 1*\*\*

**&#x20;*field* decode_cp*: int | list\[int] = 1*\*\*

**&#x20;*field* model*: str | list\[str] | None = None*\*\*

**&#x20;*field* hidden_size*: int | None = None*\*\*

**&#x20;*field* num_layers*: int | None = None*\*\*

**&#x20;*field* num_heads*: int | None = None*\*\*

**&#x20;*field* num_kv_heads*: int | None = None*\*\*

**&#x20;*field* dtype_size*: int | None = None*\*\*

**&#x20;*field* matgen_args*: MatgenCmdArgs* *\[Optional]*\*\*

**&#x20;*field* docker_image_url*: str* *\[Required]*\*\*

URL of the Docker image to use for the benchmark.

**&#x20;*field* etcd_path*: str = 'etcd'*\*\*

Path to the etcd executable.

**&#x20;*field* wait_etcd_for*: int = 60*\*\*

Number of seconds to wait for etcd to become healthy.

**&#x20;*field* etcd_image_url*: str | None = None*\*\*

Optional URL of the Docker image to use for etcd, by default etcd will be run from the same image as the benchmark.

### Test Definition

**&#x20;*class* cloudai.workloads.nixl_perftest.nixl_perftest.NixlPerftestTestDefinition(*\*, name: str, description: str, test_template_name: str, cmd_args: \~cloudai.workloads.nixl_perftest.nixl_perftest.NixlPerftestCmdArgs, dse_excluded_args: list\[str] = \<factory>, extra_env_vars: dict\[str, str | \~typing.List\[str]] = \{}, extra_cmd_args: dict\[str, str] = \{}, extra_container_mounts: list\[str] = \[], git_repos: list\[\~cloudai.\_core.installables.git_repo.GitRepo] = \[], nsys: \~cloudai.models.workload.NsysConfiguration | None = None, predictor: \~cloudai.models.workload.PredictorConfig | None = None, training_report: \~cloudai.models.workload.TrainingReportConfig | None = None, agent: str = 'grid_search', agent_steps: int = 1, agent_metrics: list\[str] = \['default'], agent_reward_function: str = 'inverse', agent_config: dict\[str, \~typing.Any] | None = None, env_params: dict\[str, \~cloudai.configurator.env_params.EnvParamSpec] = \<factory>*)**

Bases: `NIXLBaseTestDefinition[NixlPerftestCmdArgs]`

TestDefinition for NixlPerftest.

**&#x20;*property* is_domain_randomization_enabled*: bool*\*\*

at least one `env_params` annotation.

**Type:**

Whether the config declares domain randomization

**&#x20;is_dse_excluded_arg(*path: str*) → bool**

Return whether a dot-separated cmd_args path should be ignored by DSE.

**&#x20;is_env_sampled(*cmd_args_path: str*) → bool**

Whether a cmd_args field is env-sampled (env draws it per trial, not the agent).

**&#x20;*validator* validate_env_params*  » \* *all fields**\*

Validate env_params annotations against cmd_args.

`env_params` is an annotation: each key names a `cmd_args` field whose value is the candidate set (the single source of truth), and the entry carries only *how* to sample. So each key must name a real `cmd_args` field whose value is a candidate list; a scalar is already fixed, so annotating it is a meaningless label and is rejected here. When `weights` are declared, the list needs >= 2 values and the weights must align 1:1 with it. Sampling, persistence, the per-trial cmd_args overlay, and the cache key all live in `CloudAIGymEnv`; keeping this shape check in core lets the overlay stay agent- and workload-agnostic rather than re-implemented per workload.