Skip to content

Commit

Permalink
revert tests
Browse files Browse the repository at this point in the history
  • Loading branch information
@nnshah1
nnshah1 committed Dec 6, 2024
1 parent 1b2514a commit 7a9e0bd
Show file tree
Hide file tree
Showing 6 changed files with 134 additions and 85 deletions.
20 changes: 13 additions & 7 deletions genai-perf/genai_perf/parser.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -165,19 +165,26 @@ def _check_conditional_args(
Check for conditional args and raise an error if they are not set.
"""

if args.endpoint_type not in _endpoint_type_map:
parser.error(f"Invalid endpoint type {args.endpoint_type}")

endpoint_config = _endpoint_type_map[args.endpoint_type]
args.output_format = endpoint_config.output_format

# Endpoint and output format checks
if args.service_kind == "openai":
print(args.endpoint_type)
if args.endpoint_type is None:
parser.error(
"The --endpoint-type option is required when using the 'openai' service-kind."
)

if args.service_kind == "triton" and args.endpoint_type is None:
args.endpoint_type = "kserve"

if args.service_kind == "tensorrtllm_engine" and args.endpoint_type is None:
args.endpoint_type = "tensorrtllm_engine"

if args.endpoint_type and args.endpoint_type not in _endpoint_type_map:
parser.error(f"Invalid endpoint type {args.endpoint_type}")

endpoint_config = _endpoint_type_map[args.endpoint_type]
args.output_format = endpoint_config.output_format

if args.endpoint is not None:
args.endpoint = args.endpoint.lstrip(" /")
else:
Expand Down Expand Up @@ -708,7 +715,6 @@ def _add_endpoint_args(parser):
"--endpoint-type",
type=str,
choices=list(_endpoint_type_map.keys()),
default="kserve",
required=False,
help=f"The endpoint-type to send requests to on the " "server.",
)
Expand Down
147 changes: 79 additions & 68 deletions genai-perf/tests/test_cli.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -84,6 +84,8 @@ def test_help_version_arguments_output_and_exit(
"5",
"--endpoint-type",
"embeddings",
"--service-kind",
"openai",
],
{"batch_size_text": 5},
),
Expand All @@ -93,6 +95,8 @@ def test_help_version_arguments_output_and_exit(
"5",
"--endpoint-type",
"image_retrieval",
"--service-kind",
"openai",
],
{"batch_size_image": 5},
),
Expand All @@ -102,30 +106,34 @@ def test_help_version_arguments_output_and_exit(
"5",
"--endpoint-type",
"embeddings",
"--service-kind",
"openai",
],
{"batch_size_text": 5},
),
(["--concurrency", "3"], {"concurrency": 3}),
(
["--endpoint-type", "completions"],
["--endpoint-type", "completions", "--service-kind", "openai"],
{"endpoint": "v1/completions"},
),
(
["--endpoint-type", "chat"],
["--endpoint-type", "chat", "--service-kind", "openai"],
{"endpoint": "v1/chat/completions"},
),
(
["--endpoint-type", "rankings"],
["--endpoint-type", "rankings", "--service-kind", "openai"],
{"endpoint": "v1/ranking"},
),
(
["--endpoint-type", "image_retrieval"],
["--endpoint-type", "image_retrieval", "--service-kind", "openai"],
{"endpoint": "v1/infer"},
),
(
[
"--endpoint-type",
"chat",
"--service-kind",
"openai",
"--endpoint",
"custom/address",
],
Expand All @@ -135,6 +143,8 @@ def test_help_version_arguments_output_and_exit(
[
"--endpoint-type",
"chat",
"--service-kind",
"openai",
"--endpoint",
" /custom/address",
],
Expand All @@ -144,6 +154,8 @@ def test_help_version_arguments_output_and_exit(
[
"--endpoint-type",
"completions",
"--service-kind",
"openai",
"--endpoint",
"custom/address",
],
Expand Down Expand Up @@ -210,13 +222,13 @@ def test_help_version_arguments_output_and_exit(
(["--warmup-request-count", "100"], {"warmup_request_count": 100}),
(["--request-rate", "9.0"], {"request_rate": 9.0}),
(["-s", "99.5"], {"stability_percentage": 99.5}),
(["--endpoint-type", "kserve"], {"service_kind": "triton"}),
(["--service-kind", "triton"], {"service_kind": "triton"}),
(
["--endpoint-type", "tensorrtllm_engine"],
["--service-kind", "tensorrtllm_engine"],
{"service_kind": "tensorrtllm_engine"},
),
(
["--endpoint-type", "chat"],
["--service-kind", "openai", "--endpoint-type", "chat"],
{"service_kind": "openai", "endpoint": "v1/chat/completions"},
),
(["--stability-percentage", "99.5"], {"stability_percentage": 99.5}),
Expand Down Expand Up @@ -342,33 +354,33 @@ def test_file_flags_parsed(self, monkeypatch, mocker):
"arg, expected_path",
[
(
["--endpoint-type", "chat"],
["--service-kind", "openai", "--endpoint-type", "chat"],
"artifacts/test_model-openai-chat-concurrency1",
),
(
["--endpoint-type", "completions"],
["--service-kind", "openai", "--endpoint-type", "completions"],
"artifacts/test_model-openai-completions-concurrency1",
),
(
["--endpoint-type", "rankings"],
["--service-kind", "openai", "--endpoint-type", "rankings"],
"artifacts/test_model-openai-rankings-concurrency1",
),
(
["--endpoint-type", "image_retrieval"],
["--service-kind", "openai", "--endpoint-type", "image_retrieval"],
"artifacts/test_model-openai-image_retrieval-concurrency1",
),
(
["--endpoint-type", "kserve", "--backend", "tensorrtllm"],
["--service-kind", "triton", "--backend", "tensorrtllm"],
"artifacts/test_model-triton-tensorrtllm-concurrency1",
),
(
["--endpoint-type", "kserve", "--backend", "vllm"],
["--service-kind", "triton", "--backend", "vllm"],
"artifacts/test_model-triton-vllm-concurrency1",
),
(
[
"--endpoint-type",
"kserve",
"--service-kind",
"triton",
"--backend",
"vllm",
"--concurrency",
Expand Down Expand Up @@ -403,6 +415,8 @@ def test_default_profile_export_filepath(
[
"--model",
"hello/world/test_model",
"--service-kind",
"openai",
"--endpoint-type",
"chat",
],
Expand Down Expand Up @@ -493,44 +507,29 @@ def test_unrecognized_arg(self, monkeypatch, capsys):
"args, expected_output",
[
(
["genai-perf",
"profile",
"-m",
"test_model",
"--endpoint-type",
"chat",
"--service-kind",
"triton",
],
"Invalid service-kind triton for endpoint-type chat. service-kind is now inferred from endpoint-type and the argument will be removed in a future release."
),
(
["genai-perf",
"profile",
"-m",
"test_model",
"--endpoint-type",
"kserve",
"--service-kind",
"openai",
],
"Invalid service-kind openai for endpoint-type kserve. service-kind is now inferred from endpoint-type and the argument will be removed in a future release."
),
(
["genai-perf",
"profile",
"-m",
"test_model",
"--endpoint-type",
"tensorrtllm_engine",
"--service-kind",
"openai",
],
"Invalid service-kind openai for endpoint-type tensorrtllm_engine. service-kind is now inferred from endpoint-type and the argument will be removed in a future release."
),
[
"genai-perf",
"profile",
"-m",
"test_model",
"--service-kind",
"openai",
],
"The --endpoint-type option is required when using the 'openai' service-kind.",
),
(
[
"genai-perf",
"profile",
"-m",
"test_model",
"--service-kind",
"openai",
"--endpoint",
"custom/address",
],
"The --endpoint-type option is required when using the 'openai' service-kind.",
),
(
[
"genai-perf",
Expand Down Expand Up @@ -568,20 +567,24 @@ def test_unrecognized_arg(self, monkeypatch, capsys):
"profile",
"-m",
"test_model",
"--service-kind",
"openai",
"--endpoint-type",
"chat",
"--output-tokens-mean",
"100",
"--output-tokens-mean-deterministic",
],
"The --output-tokens-mean-deterministic option is only supported with the 'kserve' and 'tensorrtllm_engine' endpoint-types",
"The --output-tokens-mean-deterministic option is only supported with the Triton and TensorRT-LLM Engine service-kind",
),
(
[
"genai-perf",
"profile",
"-m",
"test_model",
"--service-kind",
"openai",
"--endpoint-type",
"embeddings",
"--generate-plots",
Expand All @@ -594,6 +597,8 @@ def test_unrecognized_arg(self, monkeypatch, capsys):
"profile",
"-m",
"test_model",
"--service-kind",
"openai",
"--endpoint-type",
"rankings",
"--generate-plots",
Expand All @@ -606,6 +611,8 @@ def test_unrecognized_arg(self, monkeypatch, capsys):
"profile",
"-m",
"test_model",
"--service-kind",
"openai",
"--endpoint-type",
"image_retrieval",
"--generate-plots",
Expand All @@ -618,8 +625,8 @@ def test_unrecognized_arg(self, monkeypatch, capsys):
"profile",
"--model",
"test_model",
"--endpoint-type",
"kserve",
"--service-kind",
"triton",
"--server-metrics-url",
"invalid_url",
],
Expand Down Expand Up @@ -656,12 +663,14 @@ def test_unrecognized_arg(self, monkeypatch, capsys):
"profile",
"-m",
"test_model",
"--service-kind",
"openai",
"--endpoint-type",
"rankings",
"--backend",
"vllm",
],
"The --backend option should only be used when using the 'kserve' endpoint-type",
"The --backend option should only be used when using the 'triton' service-kind and 'kserve' endpoint-type.",
),
],
)
Expand All @@ -679,15 +688,17 @@ def test_conditional_errors(self, args, expected_output, monkeypatch, capsys):
"args, expected_format",
[
(
["--endpoint-type", "chat"],
["--service-kind", "openai", "--endpoint-type", "chat"],
OutputFormat.OPENAI_CHAT_COMPLETIONS,
),
(
["--endpoint-type", "completions"],
["--service-kind", "openai", "--endpoint-type", "completions"],
OutputFormat.OPENAI_COMPLETIONS,
),
(
[
"--service-kind",
"openai",
"--endpoint-type",
"completions",
"--endpoint",
Expand All @@ -696,19 +707,19 @@ def test_conditional_errors(self, args, expected_output, monkeypatch, capsys):
OutputFormat.OPENAI_COMPLETIONS,
),
(
["--endpoint-type", "rankings"],
["--service-kind", "openai", "--endpoint-type", "rankings"],
OutputFormat.RANKINGS,
),
(
["--endpoint-type", "image_retrieval"],
["--service-kind", "openai", "--endpoint-type", "image_retrieval"],
OutputFormat.IMAGE_RETRIEVAL,
),
(
["--endpoint-type", "kserve", "--backend", "tensorrtllm"],
["--service-kind", "triton", "--backend", "tensorrtllm"],
OutputFormat.TENSORRTLLM,
),
(["--endpoint-type", "kserve", "--backend", "vllm"], OutputFormat.VLLM),
(["--endpoint-type", "tensorrtllm_engine"], OutputFormat.TENSORRTLLM_ENGINE),
(["--service-kind", "triton", "--backend", "vllm"], OutputFormat.VLLM),
(["--service-kind", "tensorrtllm_engine"], OutputFormat.TENSORRTLLM_ENGINE),
],
)
def test_inferred_output_format(self, monkeypatch, args, expected_format):
Expand Down Expand Up @@ -909,8 +920,8 @@ def test_get_extra_inputs_as_dict(self, extra_inputs_list, expected_dict):
"profile",
"--model",
"test_model",
"--endpoint-type",
"kserve",
"--service-kind",
"triton",
"--server-metrics-url",
test_triton_metrics_url,
],
Expand All @@ -923,8 +934,8 @@ def test_get_extra_inputs_as_dict(self, extra_inputs_list, expected_dict):
"profile",
"--model",
"test_model",
"--endpoint-type",
"kserve",
"--service-kind",
"triton",
],
None,
),
Expand Down
Loading

0 comments on commit 7a9e0bd

Please sign in to comment.