REST Resource: projects.locations.tuningJobs

Resource: TuningJob

Represents a TuningJob that runs with Google owned models.

Fields
name string

Output only. Identifier. Resource name of a TuningJob. Format: projects/{project}/locations/{location}/tuningJobs/{tuningJob}

tunedModelDisplayName string

Optional. The display name of the TunedModel. The name can be up to 128 characters long and can consist of any UTF-8 characters. For continuous tuning, tunedModelDisplayName will by default use the same display name as the pre-tuned model. If a new display name is provided, the tuning job will create a new model instead of a new version.

description string

Optional. The description of the TuningJob.

customBaseModel string

Optional. The user-provided path to custom model weights. Set this field to tune a custom model. The path must be a Cloud Storage directory that contains the model weights in .safetensors format along with associated model metadata files. If this field is set, the baseModel field must still be set to indicate which base model the custom model is derived from. This feature is only available for open source models.

state enum (JobState)

Output only. The detailed state of the job.

createTime string (Timestamp format)

Output only. time when the TuningJob was created.

Uses RFC 3339, where generated output will always be Z-normalized and use 0, 3, 6 or 9 fractional digits. Offsets other than "Z" are also accepted. Examples: "2014-10-02T15:01:23Z", "2014-10-02T15:01:23.045123456Z" or "2014-10-02T15:01:23+05:30".

startTime string (Timestamp format)

Output only. time when the TuningJob for the first time entered the JOB_STATE_RUNNING state.

Uses RFC 3339, where generated output will always be Z-normalized and use 0, 3, 6 or 9 fractional digits. Offsets other than "Z" are also accepted. Examples: "2014-10-02T15:01:23Z", "2014-10-02T15:01:23.045123456Z" or "2014-10-02T15:01:23+05:30".

endTime string (Timestamp format)

Output only. time when the TuningJob entered any of the following JobStates: JOB_STATE_SUCCEEDED, JOB_STATE_FAILED, JOB_STATE_CANCELLED, JOB_STATE_EXPIRED.

Uses RFC 3339, where generated output will always be Z-normalized and use 0, 3, 6 or 9 fractional digits. Offsets other than "Z" are also accepted. Examples: "2014-10-02T15:01:23Z", "2014-10-02T15:01:23.045123456Z" or "2014-10-02T15:01:23+05:30".

updateTime string (Timestamp format)

Output only. time when the TuningJob was most recently updated.

Uses RFC 3339, where generated output will always be Z-normalized and use 0, 3, 6 or 9 fractional digits. Offsets other than "Z" are also accepted. Examples: "2014-10-02T15:01:23Z", "2014-10-02T15:01:23.045123456Z" or "2014-10-02T15:01:23+05:30".

error object (Status)

Output only. Only populated when job's state is JOB_STATE_FAILED or JOB_STATE_CANCELLED.

labels map (key: string, value: string)

Optional. The labels with user-defined metadata to organize TuningJob and generated resources such as Model and Endpoint.

label keys and values can be no longer than 64 characters (Unicode codepoints), can only contain lowercase letters, numeric characters, underscores and dashes. International characters are allowed.

See https://goo.gl/xmQnxf for more information and examples of labels.

experiment string

Output only. The Experiment associated with this TuningJob.

tunedModel object (TunedModel)

Output only. The tuned model resources associated with this TuningJob.

tuningDataStats object (TuningDataStats)

Output only. The tuning data statistics associated with this TuningJob.

pipelineJob
(deprecated) string

Output only. The resource name of the PipelineJob associated with the TuningJob. Format: projects/{project}/locations/{location}/pipelineJobs/{pipelineJob}.

encryptionSpec object (EncryptionSpec)

Customer-managed encryption key options for a TuningJob. If this is set, then all resources created by the TuningJob will be encrypted with the provided encryption key.

serviceAccount string

The service account that the tuningJob workload runs as. If not specified, the Vertex AI Secure Fine-Tuned service Agent in the project will be used. See https://cloud.google.com/iam/docs/service-agents#vertex-ai-secure-fine-tuning-service-agent

Users starting the pipeline must have the iam.serviceAccounts.actAs permission on this service account.

outputUri string

Optional. Cloud Storage path to the directory where tuning job outputs are written to. This field is only available and required for open source models.

evaluateDatasetRuns[] object (EvaluateDatasetRun)

Output only. Evaluation runs for the Tuning Job.

satisfiesPzs boolean

Output only. reserved for future use.

satisfiesPzi boolean

Output only. reserved for future use.

source_model Union type
source_model can be only one of the following:
baseModel string

The base model that is being tuned. See Supported models.

tuning_spec Union type
tuning_spec can be only one of the following:
supervisedTuningSpec object (SupervisedTuningSpec)

Tuning Spec for Supervised Fine Tuning.

distillationSpec object (DistillationSpec)

Tuning Spec for Distillation.

partnerModelTuningSpec object (PartnerModelTuningSpec)

Tuning Spec for open sourced and third party Partner models.

veoTuningSpec object (VeoTuningSpec)

Tuning Spec for Veo Tuning.

JSON representation 
{
  "name": string,
  "tunedModelDisplayName": string,
  "description": string,
  "customBaseModel": string,
  "state": enum (JobState),
  "createTime": string,
  "startTime": string,
  "endTime": string,
  "updateTime": string,
  "error": {
    object (Status)
  },
  "labels": {
    string: string,
    ...
  },
  "experiment": string,
  "tunedModel": {
    object (TunedModel)
  },
  "tuningDataStats": {
    object (TuningDataStats)
  },
  "pipelineJob": string,
  "encryptionSpec": {
    object (EncryptionSpec)
  },
  "serviceAccount": string,
  "outputUri": string,
  "evaluateDatasetRuns": [
    {
      object (EvaluateDatasetRun)
    }
  ],
  "satisfiesPzs": boolean,
  "satisfiesPzi": boolean,

  // source_model
  "baseModel": string
  // Union type

  // tuning_spec
  "supervisedTuningSpec": {
    object (SupervisedTuningSpec)
  },
  "distillationSpec": {
    object (DistillationSpec)
  },
  "partnerModelTuningSpec": {
    object (PartnerModelTuningSpec)
  },
  "veoTuningSpec": {
    object (VeoTuningSpec)
  }
  // Union type
}

SupervisedTuningSpec

Tuning Spec for Supervised Tuning for first party models.

Fields
trainingDatasetUri string

Required. Training dataset used for tuning. The dataset can be specified as either a Cloud Storage path to a JSONL file or as the resource name of a Vertex Multimodal Dataset.

validationDatasetUri string

Optional. Validation dataset used for tuning. The dataset can be specified as either a Cloud Storage path to a JSONL file or as the resource name of a Vertex Multimodal Dataset.

hyperParameters object (SupervisedHyperParameters)

Optional. Hyperparameters for SFT.

exportLastCheckpointOnly boolean

Optional. If set to true, disable intermediate checkpoints for SFT and only the last checkpoint will be exported. Otherwise, enable intermediate checkpoints for SFT. Default is false.

evaluationConfig object (EvaluationConfig)

Optional. Evaluation Config for Tuning Job.

tuningMode enum (TuningMode)

Tuning mode.

JSON representation 
{
  "trainingDatasetUri": string,
  "validationDatasetUri": string,
  "hyperParameters": {
    object (SupervisedHyperParameters)
  },
  "exportLastCheckpointOnly": boolean,
  "evaluationConfig": {
    object (EvaluationConfig)
  },
  "tuningMode": enum (TuningMode)
}

SupervisedHyperParameters

Hyperparameters for SFT.

Fields
epochCount string (int64 format)

Optional. Number of complete passes the model makes over the entire training dataset during training.

learningRateMultiplier number

Optional. Multiplier for adjusting the default learning rate. Mutually exclusive with learningRate. This feature is only available for 1P models.

learningRate number

Optional. Learning rate for tuning. Mutually exclusive with learningRateMultiplier. This feature is only available for open source models.

adapterSize enum (AdapterSize)

Optional. Adapter size for tuning.

batchSize string (int64 format)

Optional. Batch size for tuning. This feature is only available for open source models.

JSON representation 
{
  "epochCount": string,
  "learningRateMultiplier": number,
  "learningRate": number,
  "adapterSize": enum (AdapterSize),
  "batchSize": string
}

AdapterSize

Supported adapter sizes for tuning.

Enums
ADAPTER_SIZE_UNSPECIFIED Adapter size is unspecified.
ADAPTER_SIZE_ONE Adapter size 1.
ADAPTER_SIZE_TWO Adapter size 2.
ADAPTER_SIZE_FOUR Adapter size 4.
ADAPTER_SIZE_EIGHT Adapter size 8.
ADAPTER_SIZE_SIXTEEN Adapter size 16.
ADAPTER_SIZE_THIRTY_TWO Adapter size 32.

EvaluationConfig

Evaluation Config for Tuning Job.

Fields
metrics[] object (Metric)

Required. The metrics used for evaluation.

outputConfig object (OutputConfig)

Required. Config for evaluation output.

autoraterConfig object (AutoraterConfig)

Optional. Autorater config for evaluation.

JSON representation 
{
  "metrics": [
    {
      object (Metric)
    }
  ],
  "outputConfig": {
    object (OutputConfig)
  },
  "autoraterConfig": {
    object (AutoraterConfig)
  }
}

Metric

The metric used for running evaluations.

Fields
aggregationMetrics[] enum (AggregationMetric)

Optional. The aggregation metrics to use.

metric_spec Union type
The spec for the metric. It would be either a pre-defined metric, or a inline metric spec. metric_spec can be only one of the following:
pointwiseMetricSpec object (PointwiseMetricSpec)

Spec for pointwise metric.

pairwiseMetricSpec object (PairwiseMetricSpec)

Spec for pairwise metric.

exactMatchSpec object (ExactMatchSpec)

Spec for exact match metric.

bleuSpec object (BleuSpec)

Spec for bleu metric.

rougeSpec object (RougeSpec)

Spec for rouge metric.

JSON representation 
{
  "aggregationMetrics": [
    enum (AggregationMetric)
  ],

  // metric_spec
  "pointwiseMetricSpec": {
    object (PointwiseMetricSpec)
  },
  "pairwiseMetricSpec": {
    object (PairwiseMetricSpec)
  },
  "exactMatchSpec": {
    object (ExactMatchSpec)
  },
  "bleuSpec": {
    object (BleuSpec)
  },
  "rougeSpec": {
    object (RougeSpec)
  }
  // Union type
}

PointwiseMetricSpec

Spec for pointwise metric.

Fields
customOutputFormatConfig object (CustomOutputFormatConfig)

Optional. CustomOutputFormatConfig allows customization of metric output. By default, metrics return a score and explanation. When this config is set, the default output is replaced with either: - The raw output string. - A parsed output based on a user-defined schema. If a custom format is chosen, the score and explanation fields in the corresponding metric result will be empty.

metricPromptTemplate string

Required. Metric prompt template for pointwise metric.

systemInstruction string

Optional. System instructions for pointwise metric.

JSON representation 
{
  "customOutputFormatConfig": {
    object (CustomOutputFormatConfig)
  },
  "metricPromptTemplate": string,
  "systemInstruction": string
}

CustomOutputFormatConfig

Spec for custom output format configuration.

Fields
custom_output_format_config Union type
Custom output format configuration. custom_output_format_config can be only one of the following:
returnRawOutput boolean

Optional. Whether to return raw output.

JSON representation 
{

  // custom_output_format_config
  "returnRawOutput": boolean
  // Union type
}

PairwiseMetricSpec

Spec for pairwise metric.

Fields
candidateResponseFieldName string

Optional. The field name of the candidate response.

baselineResponseFieldName string

Optional. The field name of the baseline response.

customOutputFormatConfig object (CustomOutputFormatConfig)

Optional. CustomOutputFormatConfig allows customization of metric output. When this config is set, the default output is replaced with the raw output string. If a custom format is chosen, the pairwiseChoice and explanation fields in the corresponding metric result will be empty.

metricPromptTemplate string

Required. Metric prompt template for pairwise metric.

systemInstruction string

Optional. System instructions for pairwise metric.

JSON representation 
{
  "candidateResponseFieldName": string,
  "baselineResponseFieldName": string,
  "customOutputFormatConfig": {
    object (CustomOutputFormatConfig)
  },
  "metricPromptTemplate": string,
  "systemInstruction": string
}

ExactMatchSpec

This type has no fields.

Spec for exact match metric - returns 1 if prediction and reference exactly matches, otherwise 0.

BleuSpec

Spec for bleu score metric - calculates the precision of n-grams in the prediction as compared to reference - returns a score ranging between 0 to 1.

Fields
useEffectiveOrder boolean

Optional. Whether to useEffectiveOrder to compute bleu score.

JSON representation 
{
  "useEffectiveOrder": boolean
}

RougeSpec

Spec for rouge score metric - calculates the recall of n-grams in prediction as compared to reference - returns a score ranging between 0 and 1.

Fields
rougeType string

Optional. Supported rouge types are rougen[1-9], rougeL, and rougeLsum.

useStemmer boolean

Optional. Whether to use stemmer to compute rouge score.

splitSummaries boolean

Optional. Whether to split summaries while using rougeLsum.

JSON representation 
{
  "rougeType": string,
  "useStemmer": boolean,
  "splitSummaries": boolean
}

AggregationMetric

The aggregation metrics supported by EvaluationService.EvaluateDataset.

Enums
AGGREGATION_METRIC_UNSPECIFIED Unspecified aggregation metric.
AVERAGE Average aggregation metric. Not supported for Pairwise metric.
MODE Mode aggregation metric.
STANDARD_DEVIATION Standard deviation aggregation metric. Not supported for pairwise metric.
VARIANCE Variance aggregation metric. Not supported for pairwise metric.
MINIMUM Minimum aggregation metric. Not supported for pairwise metric.
MAXIMUM Maximum aggregation metric. Not supported for pairwise metric.
MEDIAN Median aggregation metric. Not supported for pairwise metric.
PERCENTILE_P90 90th percentile aggregation metric. Not supported for pairwise metric.
PERCENTILE_P95 95th percentile aggregation metric. Not supported for pairwise metric.
PERCENTILE_P99 99th percentile aggregation metric. Not supported for pairwise metric.

OutputConfig

Config for evaluation output.

Fields
destination Union type
The destination for evaluation output. destination can be only one of the following:
gcsDestination object (GcsDestination)

Cloud storage destination for evaluation output.

JSON representation 
{

  // destination
  "gcsDestination": {
    object (GcsDestination)
  }
  // Union type
}

AutoraterConfig

The configs for autorater. This is applicable to both EvaluateInstances and EvaluateDataset.

Fields
autoraterModel string

Optional. The fully qualified name of the publisher model or tuned autorater endpoint to use.

Publisher model format: projects/{project}/locations/{location}/publishers/*/models/*

Tuned model endpoint format: projects/{project}/locations/{location}/endpoints/{endpoint}

generationConfig object (GenerationConfig)

Optional. Configuration options for model generation and outputs.

samplingCount integer

Optional. Number of samples for each instance in the dataset. If not specified, the default is 4. Minimum value is 1, maximum value is 32.

flipEnabled boolean

Optional. Default is true. Whether to flip the candidate and baseline responses. This is only applicable to the pairwise metric. If enabled, also provide PairwiseMetricSpec.candidate_response_field_name and PairwiseMetricSpec.baseline_response_field_name. When rendering PairwiseMetricSpec.metric_prompt_template, the candidate and baseline fields will be flipped for half of the samples to reduce bias.

JSON representation 
{
  "autoraterModel": string,
  "generationConfig": {
    object (GenerationConfig)
  },
  "samplingCount": integer,
  "flipEnabled": boolean
}

GenerationConfig

Configuration for content generation.

This message contains all the parameters that control how the model generates content. It allows you to influence the randomness, length, and structure of the output.

Fields
stopSequences[] string

Optional. A list of character sequences that will stop the model from generating further tokens. If a stop sequence is generated, the output will end at that point. This is useful for controlling the length and structure of the output. For example, you can use ["\n", "###"] to stop generation at a new line or a specific marker.

responseMimeType string

Optional. The IANA standard MIME type of the response. The model will generate output that conforms to this MIME type. Supported values include 'text/plain' (default) and 'application/json'. The model needs to be prompted to output the appropriate response type, otherwise the behavior is undefined. This is a preview feature.

responseModalities[] enum (Modality)

Optional. The modalities of the response. The model will generate a response that includes all the specified modalities. For example, if this is set to [TEXT, IMAGE], the response will include both text and an image.

thinkingConfig object (ThinkingConfig)

Optional. Configuration for thinking features. An error will be returned if this field is set for models that don't support thinking.

modelConfig
(deprecated) object (ModelConfig)

Optional. Config for model selection.

temperature number

Optional. Controls the randomness of the output. A higher temperature results in more creative and diverse responses, while a lower temperature makes the output more predictable and focused. The valid range is (0.0, 2.0].

topP number

Optional. Specifies the nucleus sampling threshold. The model considers only the smallest set of tokens whose cumulative probability is at least topP. This helps generate more diverse and less repetitive responses. For example, a topP of 0.9 means the model considers tokens until the cumulative probability of the tokens to select from reaches 0.9. It's recommended to adjust either temperature or topP, but not both.

topK number

Optional. Specifies the top-k sampling threshold. The model considers only the top k most probable tokens for the next token. This can be useful for generating more coherent and less random text. For example, a topK of 40 means the model will choose the next word from the 40 most likely words.

candidateCount integer

Optional. The number of candidate responses to generate.

A higher candidateCount can provide more options to choose from, but it also consumes more resources. This can be useful for generating a variety of responses and selecting the best one.

maxOutputTokens integer

Optional. The maximum number of tokens to generate in the response.

A token is approximately four characters. The default value varies by model. This parameter can be used to control the length of the generated text and prevent overly long responses.

responseLogprobs boolean

Optional. If set to true, the log probabilities of the output tokens are returned.

log probabilities are the logarithm of the probability of a token appearing in the output. A higher log probability means the token is more likely to be generated. This can be useful for analyzing the model's confidence in its own output and for debugging.

logprobs integer

Optional. The number of top log probabilities to return for each token.

This can be used to see which other tokens were considered likely candidates for a given position. A higher value will return more options, but it will also increase the size of the response.

presencePenalty number

Optional. Penalizes tokens that have already appeared in the generated text. A positive value encourages the model to generate more diverse and less repetitive text. Valid values can range from [-2.0, 2.0].

frequencyPenalty number

Optional. Penalizes tokens based on their frequency in the generated text. A positive value helps to reduce the repetition of words and phrases. Valid values can range from [-2.0, 2.0].

seed integer

Optional. A seed for the random number generator.

By setting a seed, you can make the model's output mostly deterministic. For a given prompt and parameters (like temperature, topP, etc.), the model will produce the same response every time. However, it's not a guaranteed absolute deterministic behavior. This is different from parameters like temperature, which control the level of randomness. seed ensures that the "random" choices the model makes are the same on every run, making it essential for testing and ensuring reproducible results.

responseSchema object (Schema)

Optional. Lets you to specify a schema for the model's response, ensuring that the output conforms to a particular structure. This is useful for generating structured data such as JSON. The schema is a subset of the OpenAPI 3.0 schema object object.

When this field is set, you must also set the responseMimeType to application/json.

responseJsonSchema value (Value format)

Optional. When this field is set, responseSchema must be omitted and responseMimeType must be set to application/json.

routingConfig object (RoutingConfig)

Optional. Routing configuration.

audioTimestamp boolean

Optional. If enabled, audio timestamps will be included in the request to the model. This can be useful for synchronizing audio with other modalities in the response.

mediaResolution enum (MediaResolution)

Optional. The token resolution at which input media content is sampled. This is used to control the trade-off between the quality of the response and the number of tokens used to represent the media. A higher resolution allows the model to perceive more detail, which can lead to a more nuanced response, but it will also use more tokens. This does not affect the image dimensions sent to the model.

speechConfig object (SpeechConfig)

Optional. The speech generation config.

enableAffectiveDialog boolean

Optional. If enabled, the model will detect emotions and adapt its responses accordingly. For example, if the model detects that the user is frustrated, it may provide a more empathetic response.

JSON representation 
{
  "stopSequences": [
    string
  ],
  "responseMimeType": string,
  "responseModalities": [
    enum (Modality)
  ],
  "thinkingConfig": {
    object (ThinkingConfig)
  },
  "modelConfig": {
    object (ModelConfig)
  },
  "temperature": number,
  "topP": number,
  "topK": number,
  "candidateCount": integer,
  "maxOutputTokens": integer,
  "responseLogprobs": boolean,
  "logprobs": integer,
  "presencePenalty": number,
  "frequencyPenalty": number,
  "seed": integer,
  "responseSchema": {
    object (Schema)
  },
  "responseJsonSchema": value,
  "routingConfig": {
    object (RoutingConfig)
  },
  "audioTimestamp": boolean,
  "mediaResolution": enum (MediaResolution),
  "speechConfig": {
    object (SpeechConfig)
  },
  "enableAffectiveDialog": boolean
}

RoutingConfig

The configuration for routing the request to a specific model. This can be used to control which model is used for the generation, either automatically or by specifying a model name.

Fields
routing_config Union type
The routing mode for the request. routing_config can be only one of the following:
autoMode object (AutoRoutingMode)

In this mode, the model is selected automatically based on the content of the request.

manualMode object (ManualRoutingMode)

In this mode, the model is specified manually.

JSON representation 
{

  // routing_config
  "autoMode": {
    object (AutoRoutingMode)
  },
  "manualMode": {
    object (ManualRoutingMode)
  }
  // Union type
}

AutoRoutingMode

The configuration for automated routing.

When automated routing is specified, the routing will be determined by the pretrained routing model and customer provided model routing preference.

Fields
modelRoutingPreference enum (ModelRoutingPreference)

The model routing preference.

JSON representation 
{
  "modelRoutingPreference": enum (ModelRoutingPreference)
}

ModelRoutingPreference

The model routing preference.

Enums
UNKNOWN Unspecified model routing preference.
PRIORITIZE_QUALITY The model will be selected to prioritize the quality of the response.
BALANCED The model will be selected to balance quality and cost.
PRIORITIZE_COST The model will be selected to prioritize the cost of the request.

ManualRoutingMode

The configuration for manual routing.

When manual routing is specified, the model will be selected based on the model name provided.

Fields
modelName string

The name of the model to use. Only public LLM models are accepted.

JSON representation 
{
  "modelName": string
}

Modality

The modalities of the response.

Enums
MODALITY_UNSPECIFIED Unspecified modality. Will be processed as text.
TEXT Text modality.
IMAGE Image modality.
AUDIO Audio modality.

MediaResolution

Media resolution for the input media.

Enums
MEDIA_RESOLUTION_UNSPECIFIED Media resolution has not been set.
MEDIA_RESOLUTION_LOW Media resolution set to low (64 tokens).
MEDIA_RESOLUTION_MEDIUM Media resolution set to medium (256 tokens).
MEDIA_RESOLUTION_HIGH Media resolution set to high (zoomed reframing with 256 tokens).

SpeechConfig

Configuration for speech generation.

Fields
voiceConfig object (VoiceConfig)

The configuration for the voice to use.

languageCode string

Optional. The language code (ISO 639-1) for the speech synthesis.

JSON representation 
{
  "voiceConfig": {
    object (VoiceConfig)
  },
  "languageCode": string
}

VoiceConfig

Configuration for a voice.

Fields
voice_config Union type
The configuration for the speaker to use. voice_config can be only one of the following:
prebuiltVoiceConfig object (PrebuiltVoiceConfig)

The configuration for a prebuilt voice.

replicatedVoiceConfig object (ReplicatedVoiceConfig)

Optional. The configuration for a replicated voice. This enables users to replicate a voice from an audio sample.

JSON representation 
{

  // voice_config
  "prebuiltVoiceConfig": {
    object (PrebuiltVoiceConfig)
  },
  "replicatedVoiceConfig": {
    object (ReplicatedVoiceConfig)
  }
  // Union type
}

PrebuiltVoiceConfig

Configuration for a prebuilt voice.

Fields
voiceName string

The name of the prebuilt voice to use.

JSON representation 
{
  "voiceName": string
}

ReplicatedVoiceConfig

The configuration for the replicated voice to use.

Fields
mimeType string

Optional. The mimetype of the voice sample. Currently only mimeType=audio/pcm is supported, which is raw mono 16-bit signed little-endian pcm data, with 24k sampling rate.

voiceSampleAudio string (bytes format)

Optional. The sample of the custom voice.

A base64-encoded string.

JSON representation 
{
  "mimeType": string,
  "voiceSampleAudio": string
}

ThinkingConfig

Configuration for the model's thinking features.

"Thinking" is a process where the model breaks down a complex task into smaller, manageable steps. This allows the model to reason about the task, plan its approach, and execute the plan to generate a high-quality response.

Fields
includeThoughts boolean

Optional. If true, the model will include its thoughts in the response. "Thoughts" are the intermediate steps the model takes to arrive at the final response. They can provide insights into the model's reasoning process and help with debugging. If this is true, thoughts are returned only when available.

thinkingBudget integer

Optional. The token budget for the model's thinking process. The model will make a best effort to stay within this budget. This can be used to control the trade-off between response quality and latency.

JSON representation 
{
  "includeThoughts": boolean,
  "thinkingBudget": integer
}

ModelConfig

Config for model selection.

Fields
featureSelectionPreference enum (FeatureSelectionPreference)

Required. feature selection preference.

JSON representation 
{
  "featureSelectionPreference": enum (FeatureSelectionPreference)
}

FeatureSelectionPreference

Options for feature selection preference.

Enums
FEATURE_SELECTION_PREFERENCE_UNSPECIFIED Unspecified feature selection preference.
PRIORITIZE_QUALITY Prefer higher quality over lower cost.
BALANCED Balanced feature selection preference.
PRIORITIZE_COST Prefer lower cost over higher quality.

TuningMode

Supported tuning modes.

Enums
TUNING_MODE_UNSPECIFIED Tuning mode is unspecified.
TUNING_MODE_FULL Full fine-tuning mode.
TUNING_MODE_PEFT_ADAPTER PEFT adapter tuning mode.

DistillationSpec

Tuning Spec for Distillation.

Fields
trainingDatasetUri
(deprecated) string

Deprecated. Cloud Storage path to file containing training dataset for tuning. The dataset must be formatted as a JSONL file.

hyperParameters object (DistillationHyperParameters)

Optional. Hyperparameters for Distillation.

studentModel
(deprecated) string

The student model that is being tuned, e.g., "google/gemma-2b-1.1-it". Deprecated. Use baseModel instead.

pipelineRootDirectory
(deprecated) string

Deprecated. A path in a Cloud Storage bucket, which will be treated as the root output directory of the distillation pipeline. It is used by the system to generate the paths of output artifacts.

teacher_model Union type
The teacher model that is being distilled from. See Supported models. teacher_model can be only one of the following:
baseTeacherModel string

The base teacher model that is being distilled. See Supported models.

tunedTeacherModelSource string

The resource name of the Tuned teacher model. Format: projects/{project}/locations/{location}/models/{model}.

validationDatasetUri string

Optional. Cloud Storage path to file containing validation dataset for tuning. The dataset must be formatted as a JSONL file.

JSON representation 
{
  "trainingDatasetUri": string,
  "hyperParameters": {
    object (DistillationHyperParameters)
  },
  "studentModel": string,
  "pipelineRootDirectory": string,

  // teacher_model
  "baseTeacherModel": string,
  "tunedTeacherModelSource": string
  // Union type
  "validationDatasetUri": string
}

DistillationHyperParameters

Hyperparameters for Distillation.

Fields
adapterSize enum (AdapterSize)

Optional. Adapter size for distillation.

epochCount string (int64 format)

Optional. Number of complete passes the model makes over the entire training dataset during training.

learningRateMultiplier number

Optional. Multiplier for adjusting the default learning rate.

JSON representation 
{
  "adapterSize": enum (AdapterSize),
  "epochCount": string,
  "learningRateMultiplier": number
}

PartnerModelTuningSpec

Tuning spec for Partner models.

Fields
trainingDatasetUri string

Required. Cloud Storage path to file containing training dataset for tuning. The dataset must be formatted as a JSONL file.

validationDatasetUri string

Optional. Cloud Storage path to file containing validation dataset for tuning. The dataset must be formatted as a JSONL file.

hyperParameters map (key: string, value: value (Value format))

Hyperparameters for tuning. The accepted hyperParameters and their valid range of values will differ depending on the base model.

JSON representation 
{
  "trainingDatasetUri": string,
  "validationDatasetUri": string,
  "hyperParameters": {
    string: value,
    ...
  }
}

VeoTuningSpec

Tuning Spec for Veo Model Tuning.

Fields
trainingDatasetUri string

Required. Training dataset used for tuning. The dataset can be specified as either a Cloud Storage path to a JSONL file or as the resource name of a Vertex Multimodal Dataset.

validationDatasetUri string

Optional. Validation dataset used for tuning. The dataset can be specified as either a Cloud Storage path to a JSONL file or as the resource name of a Vertex Multimodal Dataset.

hyperParameters object (VeoHyperParameters)

Optional. Hyperparameters for Veo.

JSON representation 
{
  "trainingDatasetUri": string,
  "validationDatasetUri": string,
  "hyperParameters": {
    object (VeoHyperParameters)
  }
}

VeoHyperParameters

Hyperparameters for Veo.

Fields
epochCount string (int64 format)

Optional. Number of complete passes the model makes over the entire training dataset during training.

learningRateMultiplier number

Optional. Multiplier for adjusting the default learning rate.

tuningTask enum (TuningTask)

Optional. The tuning task. Either I2V or T2V.

JSON representation 
{
  "epochCount": string,
  "learningRateMultiplier": number,
  "tuningTask": enum (TuningTask)
}

TuningTask

An enum defining the tuning task used for Veo.

Enums
TUNING_TASK_UNSPECIFIED Default value. This value is unused.
TUNING_TASK_I2V Tuning task for image to video.
TUNING_TASK_T2V Tuning task for text to video.
TUNING_TASK_R2V Tuning task for reference to video.

TunedModel

The Model Registry Model and Online Prediction Endpoint associated with this TuningJob.

Fields
model string

Output only. The resource name of the TunedModel. Format:

projects/{project}/locations/{location}/models/{model}@{versionId}

When tuning from a base model, the version id will be 1.

For continuous tuning, if the provided tunedModelDisplayName is set and different from parent model's display name, the tuned model will have a new parent model with version 1. Otherwise the version id will be incremented by 1 from the last version id in the parent model. E.g.,

projects/{project}/locations/{location}/models/{model}@{last_version_id + 1}

endpoint string

Output only. A resource name of an Endpoint. Format: projects/{project}/locations/{location}/endpoints/{endpoint}.

checkpoints[] object (TunedModelCheckpoint)

Output only. The checkpoints associated with this TunedModel. This field is only populated for tuning jobs that enable intermediate checkpoints.

JSON representation 
{
  "model": string,
  "endpoint": string,
  "checkpoints": [
    {
      object (TunedModelCheckpoint)
    }
  ]
}

TunedModelCheckpoint

TunedModelCheckpoint for the Tuned Model of a Tuning Job.

Fields
checkpointId string

The id of the checkpoint.

epoch string (int64 format)

The epoch of the checkpoint.

step string (int64 format)

The step of the checkpoint.

endpoint string

The Endpoint resource name that the checkpoint is deployed to. Format: projects/{project}/locations/{location}/endpoints/{endpoint}.

JSON representation 
{
  "checkpointId": string,
  "epoch": string,
  "step": string,
  "endpoint": string
}

TuningDataStats

The tuning data statistic values for TuningJob.

Fields
tuning_data_stats Union type
tuning_data_stats can be only one of the following:
supervisedTuningDataStats object (SupervisedTuningDataStats)

The SFT Tuning data stats.

distillationDataStats object (DistillationDataStats)

Output only. Statistics for distillation prompt dataset. These statistics do not include the responses sampled from the teacher model.

JSON representation 
{

  // tuning_data_stats
  "supervisedTuningDataStats": {
    object (SupervisedTuningDataStats)
  },
  "distillationDataStats": {
    object (DistillationDataStats)
  }
  // Union type
}

SupervisedTuningDataStats

Tuning data statistics for Supervised Tuning.

Fields
tuningDatasetExampleCount string (int64 format)

Output only. Number of examples in the tuning dataset.

totalTuningCharacterCount string (int64 format)

Output only. Number of tuning characters in the tuning dataset.

totalBillableCharacterCount
(deprecated) string (int64 format)

Output only. Number of billable characters in the tuning dataset.

totalBillableTokenCount string (int64 format)

Output only. Number of billable tokens in the tuning dataset.

tuningStepCount string (int64 format)

Output only. Number of tuning steps for this Tuning Job.

userInputTokenDistribution object (SupervisedTuningDatasetDistribution)

Output only. Dataset distributions for the user input tokens.

userOutputTokenDistribution object (SupervisedTuningDatasetDistribution)

Output only. Dataset distributions for the user output tokens.

userMessagePerExampleDistribution object (SupervisedTuningDatasetDistribution)

Output only. Dataset distributions for the messages per example.

userDatasetExamples[] object (Content)

Output only. Sample user messages in the training dataset uri.

totalTruncatedExampleCount string (int64 format)

Output only. The number of examples in the dataset that have been dropped. An example can be dropped for reasons including: too many tokens, contains an invalid image, contains too many images, etc.

truncatedExampleIndices[] string (int64 format)

Output only. A partial sample of the indices (starting from 1) of the dropped examples.

droppedExampleReasons[] string

Output only. For each index in truncatedExampleIndices, the user-facing reason why the example was dropped.

JSON representation 
{
  "tuningDatasetExampleCount": string,
  "totalTuningCharacterCount": string,
  "totalBillableCharacterCount": string,
  "totalBillableTokenCount": string,
  "tuningStepCount": string,
  "userInputTokenDistribution": {
    object (SupervisedTuningDatasetDistribution)
  },
  "userOutputTokenDistribution": {
    object (SupervisedTuningDatasetDistribution)
  },
  "userMessagePerExampleDistribution": {
    object (SupervisedTuningDatasetDistribution)
  },
  "userDatasetExamples": [
    {
      object (Content)
    }
  ],
  "totalTruncatedExampleCount": string,
  "truncatedExampleIndices": [
    string
  ],
  "droppedExampleReasons": [
    string
  ]
}

SupervisedTuningDatasetDistribution

Dataset distribution for Supervised Tuning.

Fields
sum string (int64 format)

Output only. Sum of a given population of values.

billableSum string (int64 format)

Output only. Sum of a given population of values that are billable.

min number

Output only. The minimum of the population values.

max number

Output only. The maximum of the population values.

mean number

Output only. The arithmetic mean of the values in the population.

median number

Output only. The median of the values in the population.

p5 number

Output only. The 5th percentile of the values in the population.

p95 number

Output only. The 95th percentile of the values in the population.

buckets[] object (DatasetBucket)

Output only. Defines the histogram bucket.

JSON representation 
{
  "sum": string,
  "billableSum": string,
  "min": number,
  "max": number,
  "mean": number,
  "median": number,
  "p5": number,
  "p95": number,
  "buckets": [
    {
      object (DatasetBucket)
    }
  ]
}

DatasetBucket

Dataset bucket used to create a histogram for the distribution given a population of values.

Fields
count number

Output only. Number of values in the bucket.

left number

Output only. left bound of the bucket.

right number

Output only. Right bound of the bucket.

JSON representation 
{
  "count": number,
  "left": number,
  "right": number
}

DistillationDataStats

Statistics for distillation prompt dataset. These statistics do not include the responses sampled from the teacher model.

Fields
trainingDatasetStats object (DatasetStats)

Output only. Statistics computed for the training dataset.

JSON representation 
{
  "trainingDatasetStats": {
    object (DatasetStats)
  }
}

DatasetStats

Statistics computed over a tuning dataset.

Fields
tuningDatasetExampleCount string (int64 format)

Output only. Number of examples in the tuning dataset.

totalTuningCharacterCount string (int64 format)

Output only. Number of tuning characters in the tuning dataset.

totalBillableCharacterCount string (int64 format)

Output only. Number of billable characters in the tuning dataset.

tuningStepCount string (int64 format)

Output only. Number of tuning steps for this Tuning Job.

userInputTokenDistribution object (DatasetDistribution)

Output only. Dataset distributions for the user input tokens.

userMessagePerExampleDistribution object (DatasetDistribution)

Output only. Dataset distributions for the messages per example.

userDatasetExamples[] object (Content)

Output only. Sample user messages in the training dataset uri.

droppedExampleIndices[] string (int64 format)

Output only. A partial sample of the indices (starting from 1) of the dropped examples.

droppedExampleReasons[] string

Output only. For each index in droppedExampleIndices, the user-facing reason why the example was dropped.

userOutputTokenDistribution object (DatasetDistribution)

Output only. Dataset distributions for the user output tokens.

JSON representation 
{
  "tuningDatasetExampleCount": string,
  "totalTuningCharacterCount": string,
  "totalBillableCharacterCount": string,
  "tuningStepCount": string,
  "userInputTokenDistribution": {
    object (DatasetDistribution)
  },
  "userMessagePerExampleDistribution": {
    object (DatasetDistribution)
  },
  "userDatasetExamples": [
    {
      object (Content)
    }
  ],
  "droppedExampleIndices": [
    string
  ],
  "droppedExampleReasons": [
    string
  ],
  "userOutputTokenDistribution": {
    object (DatasetDistribution)
  }
}

DatasetDistribution

Distribution computed over a tuning dataset.

Fields
sum number

Output only. Sum of a given population of values.

min number

Output only. The minimum of the population values.

max number

Output only. The maximum of the population values.

mean number

Output only. The arithmetic mean of the values in the population.

median number

Output only. The median of the values in the population.

p5 number

Output only. The 5th percentile of the values in the population.

p95 number

Output only. The 95th percentile of the values in the population.

buckets[] object (DistributionBucket)

Output only. Defines the histogram bucket.

JSON representation 
{
  "sum": number,
  "min": number,
  "max": number,
  "mean": number,
  "median": number,
  "p5": number,
  "p95": number,
  "buckets": [
    {
      object (DistributionBucket)
    }
  ]
}

DistributionBucket

Dataset bucket used to create a histogram for the distribution given a population of values.

Fields
count string (int64 format)

Output only. Number of values in the bucket.

left number

Output only. left bound of the bucket.

right number

Output only. Right bound of the bucket.

JSON representation 
{
  "count": string,
  "left": number,
  "right": number
}

EvaluateDatasetRun

Evaluate Dataset Run result for Tuning Job.

Fields
operationName string

Output only. The operation id of the evaluation run. Format: projects/{project}/locations/{location}/operations/{operationId}.

checkpointId string

Output only. The checkpoint id used in the evaluation run. Only populated when evaluating checkpoints.

evaluateDatasetResponse object (EvaluateDatasetResponse)

Output only. Results for EvaluationService.EvaluateDataset.

error object (Status)

Output only. The error of the evaluation run if any.

JSON representation 
{
  "operationName": string,
  "checkpointId": string,
  "evaluateDatasetResponse": {
    object (EvaluateDatasetResponse)
  },
  "error": {
    object (Status)
  }
}

EvaluateDatasetResponse

Response in LRO for EvaluationService.EvaluateDataset.

Fields
aggregationOutput object (AggregationOutput)

Output only. Aggregation statistics derived from results of EvaluationService.EvaluateDataset.

outputInfo object (OutputInfo)

Output only. Output info for EvaluationService.EvaluateDataset.

JSON representation 
{
  "aggregationOutput": {
    object (AggregationOutput)
  },
  "outputInfo": {
    object (OutputInfo)
  }
}

AggregationOutput

The aggregation result for the entire dataset and all metrics.

Fields
dataset object (EvaluationDataset)

The dataset used for evaluation & aggregation.

aggregationResults[] object (AggregationResult)

One AggregationResult per metric.

JSON representation 
{
  "dataset": {
    object (EvaluationDataset)
  },
  "aggregationResults": [
    {
      object (AggregationResult)
    }
  ]
}

EvaluationDataset

The dataset used for evaluation.

Fields
source Union type
The source of the dataset. source can be only one of the following:
gcsSource object (GcsSource)

Cloud storage source holds the dataset. Currently only one Cloud Storage file path is supported.

bigquerySource object (BigQuerySource)

BigQuery source holds the dataset.

JSON representation 
{

  // source
  "gcsSource": {
    object (GcsSource)
  },
  "bigquerySource": {
    object (BigQuerySource)
  }
  // Union type
}

AggregationResult

The aggregation result for a single metric.

Fields
aggregationMetric enum (AggregationMetric)

Aggregation metric.

aggregation_result Union type
The aggregation result. aggregation_result can be only one of the following:
pointwiseMetricResult object (PointwiseMetricResult)

result for pointwise metric.

pairwiseMetricResult object (PairwiseMetricResult)

result for pairwise metric.

exactMatchMetricValue object (ExactMatchMetricValue)

Results for exact match metric.

bleuMetricValue object (BleuMetricValue)

Results for bleu metric.

rougeMetricValue object (RougeMetricValue)

Results for rouge metric.

JSON representation 
{
  "aggregationMetric": enum (AggregationMetric),

  // aggregation_result
  "pointwiseMetricResult": {
    object (PointwiseMetricResult)
  },
  "pairwiseMetricResult": {
    object (PairwiseMetricResult)
  },
  "exactMatchMetricValue": {
    object (ExactMatchMetricValue)
  },
  "bleuMetricValue": {
    object (BleuMetricValue)
  },
  "rougeMetricValue": {
    object (RougeMetricValue)
  }
  // Union type
}

PointwiseMetricResult

Spec for pointwise metric result.

Fields
explanation string

Output only. Explanation for pointwise metric score.

customOutput object (CustomOutput)

Output only. Spec for custom output.

score number

Output only. Pointwise metric score.

JSON representation 
{
  "explanation": string,
  "customOutput": {
    object (CustomOutput)
  },
  "score": number
}

CustomOutput

Spec for custom output.

Fields
custom_output Union type
Custom output. custom_output can be only one of the following:
rawOutputs object (RawOutput)

Output only. List of raw output strings.

JSON representation 
{

  // custom_output
  "rawOutputs": {
    object (RawOutput)
  }
  // Union type
}

RawOutput

Raw output.

Fields
rawOutput[] string

Output only. Raw output string.

JSON representation 
{
  "rawOutput": [
    string
  ]
}

PairwiseMetricResult

Spec for pairwise metric result.

Fields
pairwiseChoice enum (PairwiseChoice)

Output only. Pairwise metric choice.

explanation string

Output only. Explanation for pairwise metric score.

customOutput object (CustomOutput)

Output only. Spec for custom output.

JSON representation 
{
  "pairwiseChoice": enum (PairwiseChoice),
  "explanation": string,
  "customOutput": {
    object (CustomOutput)
  }
}

PairwiseChoice

Pairwise prediction autorater preference.

Enums
PAIRWISE_CHOICE_UNSPECIFIED Unspecified prediction choice.
BASELINE baseline prediction wins
CANDIDATE Candidate prediction wins
TIE Winner cannot be determined

ExactMatchMetricValue

Exact match metric value for an instance.

Fields
score number

Output only. Exact match score.

JSON representation 
{
  "score": number
}

BleuMetricValue

Bleu metric value for an instance.

Fields
score number

Output only. Bleu score.

JSON representation 
{
  "score": number
}

RougeMetricValue

Rouge metric value for an instance.

Fields
score number

Output only. Rouge score.

JSON representation 
{
  "score": number
}

OutputInfo

Describes the info for output of EvaluationService.EvaluateDataset.

Fields
output_location Union type
The output location into which evaluation output is written. output_location can be only one of the following:
gcsOutputDirectory string

Output only. The full path of the Cloud Storage directory created, into which the evaluation results and aggregation results are written.

JSON representation 
{

  // output_location
  "gcsOutputDirectory": string
  // Union type
}

Methods

cancel

 Cancels a TuningJob.

create

 Creates a TuningJob.

get

 Gets a TuningJob.

list

 Lists TuningJobs in a Location.

rebaseTunedModel

 Rebase a TunedModel.