Skip to content

Commit

Permalink
feat: tune_model method for class TextEmbeddingModel.
Browse files Browse the repository at this point in the history 
PiperOrigin-RevId: 627266791
  • Loading branch information
@vertex-sdk-bot
vertex-sdk-bot authored and Copybara-Service committed Apr 23, 2024
1 parent 415912e commit 3eda55d
Show file tree
Hide file tree
Showing 2 changed files with 492 additions and 78 deletions.
202 changes: 201 additions & 1 deletion tests/unit/aiplatform/test_language_models.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -101,7 +101,18 @@
"prediction_schema_uri": "gs://google-cloud-aiplatform/schema/predict/prediction/text_generation_1.0.0.yaml",
},
}

_TEXT_GECKO_PUBLISHER_MODEL_DICT = {
"name": "publishers/google/models/textembedding-gecko",
"version_id": "003",
"open_source_category": "PROPRIETARY",
"launch_stage": gca_publisher_model.PublisherModel.LaunchStage.GA,
"publisher_model_template": "projects/{user-project}/locations/{location}/publishers/google/models/textembedding-gecko@003",
"predict_schemata": {
"instance_schema_uri": "gs://google-cloud-aiplatform/schema/predict/instance/text_embedding_1.0.0.yaml",
"parameters_schema_uri": "gs://google-cloud-aiplatfrom/schema/predict/params/text_embedding_1.0.0.yaml",
"prediction_schema_uri": "gs://google-cloud-aiplatform/schema/predict/prediction/text_embedding_1.0.0.yaml",
},
}
_CHAT_BISON_PUBLISHER_MODEL_DICT = {
"name": "publishers/google/models/chat-bison",
"version_id": "001",
Expand Down Expand Up @@ -528,6 +539,105 @@ def reverse_string_2(s):""",
},
)

_EMBEDING_MODEL_TUNING_PIPELINE_SPEC = {
"components": {},
"deploymentSpec": {},
"pipelineInfo": {
"description": "Pipeline definition for v1.1.x embedding tuning pipelines.",
"name": "tune-text-embedding-model",
},
"root": {
"dag": {"tasks": {}},
"inputDefinitions": {
"parameters": {
"accelerator_count": {
"defaultValue": 4,
"description": "how many accelerators to use when running the\ncontainer.",
"isOptional": True,
"parameterType": "NUMBER_INTEGER",
},
"accelerator_type": {
"defaultValue": "NVIDIA_TESLA_V100",
"description": "the accelerator type for running the trainer component.",
"isOptional": True,
"parameterType": "STRING",
},
"base_model_version_id": {
"defaultValue": "textembedding-gecko@001",
"description": "which base model to tune. This may be any stable\nnumbered version, for example `textembedding-gecko@001`.",
"isOptional": True,
"parameterType": "STRING",
},
"batch_size": {
"defaultValue": 128,
"description": "training batch size.",
"isOptional": True,
"parameterType": "NUMBER_INTEGER",
},
"corpus_path": {
"description": "the GCS path to the corpus data location.",
"parameterType": "STRING",
},
"iterations": {
"defaultValue": 1000,
"description": "the number of steps to perform fine-tuning.",
"isOptional": True,
"parameterType": "NUMBER_INTEGER",
},
"location": {
"defaultValue": "us-central1",
"description": "GCP region to run the pipeline.",
"isOptional": True,
"parameterType": "STRING",
},
"machine_type": {
"defaultValue": "n1-standard-16",
"description": "the type of the machine to run the trainer component. For\nmore details about this input config, see:\nhttps://cloud.google.com/vertex-ai/docs/training/configure-compute.",
"isOptional": True,
"parameterType": "STRING",
},
"model_display_name": {
"defaultValue": "tuned-text-embedding-model",
"description": "output model display name.",
"isOptional": True,
"parameterType": "STRING",
},
"project": {
"description": "user's project id.",
"parameterType": "STRING",
},
"queries_path": {
"description": "the GCS path to the queries location.",
"parameterType": "STRING",
},
"task_type": {
"defaultValue": "DEFAULT",
"description": "the task type expected to be used during inference. Valid\nvalues are `DEFAULT`, `RETRIEVAL_QUERY`, `RETRIEVAL_DOCUMENT`,\n`SEMANTIC_SIMILARITY`, `CLASSIFICATION`, and `CLUSTERING`.",
"isOptional": True,
"parameterType": "STRING",
},
"test_label_path": {
"defaultValue": "",
"description": "the GCS path to the test label data location.",
"isOptional": True,
"parameterType": "STRING",
},
"train_label_path": {
"description": "the GCS path to the train label data location.",
"parameterType": "STRING",
},
"validation_label_path": {
"defaultValue": "",
"description": "The GCS path to the validation label data location.",
"isOptional": True,
"parameterType": "STRING",
},
}
},
},
"schemaVersion": "2.1.0",
"sdkVersion": "kfp-2.6.0",
}
_TEST_PIPELINE_SPEC = {
"components": {},
"pipelineInfo": {"name": "evaluation-llm-text-generation-pipeline"},
Expand Down Expand Up @@ -641,6 +751,9 @@ def reverse_string_2(s):""",
}


_EMBEDING_MODEL_TUNING_PIPELINE_SPEC_JSON = json.dumps(
_EMBEDING_MODEL_TUNING_PIPELINE_SPEC,
)
_TEST_PIPELINE_SPEC_JSON = json.dumps(
_TEST_PIPELINE_SPEC,
)
Expand Down Expand Up @@ -1460,6 +1573,18 @@ def mock_request_urlopen(request: str) -> Tuple[str, mock.MagicMock]:
yield request.param, mock_urlopen


@pytest.fixture
def mock_request_urlopen_gecko(request: str) -> Tuple[str, mock.MagicMock]:
data = _EMBEDING_MODEL_TUNING_PIPELINE_SPEC
with mock.patch.object(urllib_request, "urlopen") as mock_urlopen:
mock_read_response = mock.MagicMock()
mock_decode_response = mock.MagicMock()
mock_decode_response.return_value = json.dumps(data)
mock_read_response.return_value.decode = mock_decode_response
mock_urlopen.return_value.read = mock_read_response
yield request.param, mock_urlopen


@pytest.fixture
def mock_request_urlopen_rlhf(request: str) -> Tuple[str, mock.MagicMock]:
data = _TEST_RLHF_PIPELINE_SPEC
Expand Down Expand Up @@ -1528,6 +1653,21 @@ def get_endpoint_mock():
yield get_endpoint_mock


@pytest.fixture
def mock_get_tuned_embedding_model(get_endpoint_mock):
with mock.patch.object(
_language_models._TunableEmbeddingModelMixin, "get_tuned_model"
) as mock_text_generation_model:
mock_text_generation_model.return_value._model_id = (
test_constants.ModelConstants._TEST_MODEL_RESOURCE_NAME
)
mock_text_generation_model.return_value._endpoint_name = (
test_constants.EndpointConstants._TEST_ENDPOINT_NAME
)
mock_text_generation_model.return_value._endpoint = get_endpoint_mock
yield mock_text_generation_model


@pytest.fixture
def mock_get_tuned_model(get_endpoint_mock):
with mock.patch.object(
Expand Down Expand Up @@ -2134,6 +2274,66 @@ def test_text_generation_response_repr(self):
assert "blocked" in response_repr
assert "Violent" in response_repr

@pytest.mark.parametrize(
"job_spec",
[_EMBEDING_MODEL_TUNING_PIPELINE_SPEC_JSON],
)
@pytest.mark.parametrize(
"mock_request_urlopen_gecko",
["https://us-central1-kfp.pkg.dev/proj/repo/pack/latest"],
indirect=True,
)
def test_tune_text_embedding_model(
self,
mock_pipeline_service_create,
mock_pipeline_job_get,
mock_pipeline_bucket_exists,
job_spec,
mock_load_yaml_and_json,
mock_gcs_from_string,
mock_gcs_upload,
mock_request_urlopen_gecko,
mock_get_tuned_embedding_model,
):
"""Tests tuning the text embedding model."""
aiplatform.init(
project=_TEST_PROJECT,
location=_TEST_LOCATION,
encryption_spec_key_name=_TEST_ENCRYPTION_KEY_NAME,
)
with mock.patch.object(
target=model_garden_service_client.ModelGardenServiceClient,
attribute="get_publisher_model",
return_value=gca_publisher_model.PublisherModel(
_TEXT_GECKO_PUBLISHER_MODEL_DICT
),
):
model = language_models.TextEmbeddingModel.from_pretrained(
"textembedding-gecko@003"
)
tuning_job = model.tune_model(
training_data="gs://bucket/training.tsv",
corpus_data="gs://bucket/corpus.jsonl",
queries_data="gs://bucket/queries.jsonl",
test_data="gs://bucket/test.tsv",
tuned_model_location="us-central1",
train_steps=10,
accelerator="NVIDIA_TESLA_A100",
)
call_kwargs = mock_pipeline_service_create.call_args[1]
pipeline_arguments = call_kwargs[
"pipeline_job"
].runtime_config.parameter_values
assert pipeline_arguments["iterations"] == 10
assert pipeline_arguments["accelerator_type"] == "NVIDIA_TESLA_A100"

# Testing the tuned model
tuned_model = tuning_job.get_tuned_model()
assert (
tuned_model._endpoint_name
== test_constants.EndpointConstants._TEST_ENDPOINT_NAME
)

@pytest.mark.parametrize(
"job_spec",
[_TEST_PIPELINE_SPEC_JSON, _TEST_PIPELINE_JOB],
Expand Down

0 comments on commit 3eda55d

Please sign in to comment.