assets/training/finetune_acft_hf_nlp/components/validation/common/spec.yaml

$schema: https://azuremlschemas.azureedge.net/latest/commandComponent.schema.json
name: ft_nlp_common_validation
version: 0.0.73
type: command

is_deterministic: true

display_name: Common Validation Component
description: Component to validate the finetune job against Validation Service

environment: azureml://registries/azureml/environments/acft-hf-nlp-gpu/versions/87

code: ../../../src/validation

inputs:
  mlflow_model_path:
    type: mlflow_model
    optional: true
    description: MLflow model asset path. Special characters like \ and ' are invalid in the parameter value.
    mode: rw_mount

  compute_finetune:
    type: string
    optional: true
    default: serverless
    description: compute to be used for finetune component eg. provide 'FT-Cluster' if your compute is named 'FT-Cluster'. Special characters like \ and ' are invalid in the parameter value. If compute cluster name is provided, instance_type field will be ignored and the respective cluster will be used

  compute_model_import:
    type: string
    optional: true
    default: serverless
    description: compute to be used for model import component eg. provide 'FT-Cluster' if your compute is named 'FT-Cluster'. Special characters like \ and ' are invalid in the parameter value. If compute cluster name is provided, instance_type field will be ignored and the respective cluster will be used

  compute_data_preprocess:
    type: string
    optional: true
    default: serverless
    description: compute to be used for data preprocess component eg. provide 'FT-Cluster' if your compute is named 'FT-Cluster'. Special characters like \ and ' are invalid in the parameter value. If compute cluster name is provided, instance_type field will be ignored and the respective cluster will be used

  task_name:
    type: string
    optional: true
    default: SingleLabelClassification
    description: Finetuning task type

  num_nodes_finetune:
    type: integer
    min: 1
    default: 1
    optional: true
    description: number of nodes to be used for finetuning (used for distributed training)

  number_of_gpu_to_use_finetuning:
    type: integer
    optional: true
    default: 1
    description: Number of GPUs to use for finetuning

  train_file_path:
    type: uri_file
    optional: true
    description: Path to the registered training data asset. The supported data formats are `jsonl`, `json`, `csv`, `tsv` and `parquet`. Special characters like \ and ' are invalid in the parameter value.
    mode: rw_mount

  validation_file_path:
    type: uri_file
    optional: true
    description: Path to the registered validation data asset. The supported data formats are `jsonl`, `json`, `csv`, `tsv` and `parquet`. Special characters like \ and ' are invalid in the parameter value.
    mode: rw_mount

  test_file_path:
    type: uri_file
    optional: true
    description: Path to the registered test data asset. The supported data formats are `jsonl`, `json`, `csv`, `tsv` and `parquet`. Special characters like \ and ' are invalid in the parameter value.
    mode: rw_mount

  train_mltable_path:
    type: mltable
    optional: true
    description: Path to the registered training data asset in `mltable` format. Special characters like \ and ' are invalid in the parameter value.

  validation_mltable_path:
    type: mltable
    optional: true
    description: Path to the registered validation data asset in `mltable` format. Special characters like \ and ' are invalid in the parameter value.

  test_mltable_path:
    type: mltable
    optional: true
    description: Path to the registered test data asset in `mltable` format. Special characters like \ and ' are invalid in the parameter value.

  user_column_names:
    type: string
    optional: true
    description: Comma separated list of column names to be used for training

  # Validation parameters
  system_properties:
    type: string
    optional: true
    description: Validation parameters propagated from pipeline.

  num_train_epochs:
    type: integer
    optional: true
    description: Number of training epochs

  max_steps:
    type: integer
    optional: true
    description: Maximum number of training steps

  per_device_train_batch_size:
    type: integer
    optional: true
    description: Batch size per GPU/CPU for training

  per_device_eval_batch_size:
    type: integer
    optional: true
    description: Batch size per GPU/CPU for evaluation

  auto_find_batch_size:
    type: string
    enum:
    - "true"
    - "false"
    default: "false"
    optional: true
    description: If set to true, will enable auto_find_batch_size for training

  learning_rate:
    type: number
    optional: true
    description: Learning rate for optimizer

  adam_beta1:
    type: number
    optional: true
    description: Beta1 hyperparameter for the Adam optimizer

  adam_beta2:
    type: number
    optional: true
    description: Beta2 hyperparameter for the Adam optimizer

  adam_epsilon:
    type: number
    optional: true
    description: Epsilon hyperparameter for the Adam optimizer

  apply_deepspeed:
    type: string
    enum:
    - "true"
    - "false"
    default: "false"
    optional: true
    description: If set to true, will enable deepspeed for training

  precision:
    type: string
    enum:
    - "32"
    - "16"
    default: "32"
    optional: true
    description: Apply mixed precision training. This can reduce memory footprint by performing operations in half-precision.

  apply_lora:
    type: string
    enum:
    - "true"
    - "false"
    default: "false"
    optional: true
    description: If "true" enables lora.

  apply_ort:
    type: string
    enum:
    - "true"
    - "false"
    default: "false"
    optional: true
    description: If set to true, will use the ONNXRunTime training

  deepspeed_stage:
    type: string
    optional: true
    default: "2"
    enum:
    - "2"
    - "3"
    description: This parameter configures which DEFAULT deepspeed config to be used - stage2 or stage3. The default choice is stage2. Note that, this parameter is ONLY applicable when user doesn't pass any config information via deepspeed port.

  ignore_mismatched_sizes:
    type: string
    enum:
    - "true"
    - "false"
    default: "true"
    optional: true
    description: Not setting this flag will raise an error if some of the weights from the checkpoint do not have the same size as the weights of the model.

  max_seq_length:
    type: integer
    optional: true
    default: -1
    description: Controls the maximum length to use when pad_to_max_length parameter is set to `true`. Default is -1 which means the padding is done up to the model's max length. Else will be padded to `max_seq_length`.

outputs:
  validation_info:
    type: uri_file
    description: Validation status.
    mode: rw_mount

command: >-
  python validation.py $[[--system_properties '${{inputs.system_properties}}']] --validation-info '${{outputs.validation_info}}'