Skip to content

components mmdetection_image_objectdetection_instancesegmentation_finetune

github-actions[bot] edited this page Oct 22, 2023 · 21 revisions

Image Object Detection and Instance Segmentation MMDetection Model Finetune

mmdetection_image_objectdetection_instancesegmentation_finetune

Overview

Description: Component to finetune MMDetection models for image object detection and instance segmentation.

Version: 0.0.8

View in Studio: https://ml.azure.com/registries/azureml/components/mmdetection_image_objectdetection_instancesegmentation_finetune/version/0.0.8

Inputs

component input: model path

Name Description Type Default Optional Enum
model_path Output folder of model selector containing model metadata like config, checkpoints, tokenizer config. uri_folder False

component input: training mltable

Name Description Type Default Optional Enum
training_data Path to the mltable of the training dataset. mltable False

optional component input: validation mltable

Name Description Type Default Optional Enum
validation_data Path to the mltable of the validation dataset. mltable True
image_min_size Minimum image size after augmentation that is input to the network. If left empty, it would either be overwritten by image_scale in model config or would be chosen based on the task type and model selected. The image will be rescaled as large as possible within the range [image_min_size, image_max_size]. The image size will be constraint so that the max edge is no longer than image_max_size and short edge is no longer than image_min_size. integer True
image_max_size Maximum image size after augmentation that is input to the network. If left empty, it would either be overwritten by image_scale in model config or would be chosen based on the task type and model selected. The image will be rescaled as large as possible within the range [image_min_size, image_max_size]. The image size will be constraint so that the max edge is no longer than image_max_size and short edge is no longer than image_min_size. integer True
task_name Which task the model is solving. string ['image-object-detection', 'image-instance-segmentation']

primary metric

Name Description Type Default Optional Enum
metric_for_best_model Specify the metric to use to compare two different models. If left empty, will be chosen automatically based on the task type and model selected. string True ['mean_average_precision', 'precision', 'recall']

Augmentation parameters

Name Description Type Default Optional Enum
apply_augmentations If set to true, will enable data augmentations for training. boolean True True
number_of_workers Number of subprocesses to use for data loading (PyTorch only). 0 means that the data will be loaded in the main process. integer 8 True

Deepspeed Parameters

Name Description Type Default Optional Enum
apply_deepspeed If set to true, will enable deepspeed for training. Please note deepspeed is not yet supported for MMDetection, will be enabled in future. boolean False True

optional component input: deepspeed config

Name Description Type Default Optional Enum
deepspeed_config Deepspeed config to be used for finetuning. Please note deepspeed is not yet supported for MMDetection, will be enabled in future. uri_file True
apply_ort If set to true, will use the ONNXRunTime training. Please note ONNXRunTime is not yet supported for MMDetection, will be enabled in future. boolean False True

Training parameters

Name Description Type Default Optional Enum
number_of_epochs Number of training epochs. If left empty, will be chosen automatically based on the task type and model selected. integer True
max_steps If set to a positive number, the total number of training steps to perform. Overrides 'number_of_epochs'. In case of using a finite iterable dataset the training may stop before reaching the set number of steps when all data is exhausted. If left empty, will be chosen automatically based on the task type and model selected. integer True
training_batch_size Train batch size. If left empty, will be chosen automatically based on the task type and model selected. integer True
validation_batch_size Validation batch size. If left empty, will be chosen automatically based on the task type and model selected. integer True
auto_find_batch_size Flag to enable auto finding of batch size. If the provided 'per_device_train_batch_size' goes into Out Of Memory (OOM) enabling auto_find_batch_size will find the correct batch size by iteratively reducing 'per_device_train_batch_size' by a factor of 2 till the OOM is fixed. boolean False True

learning rate and learning rate scheduler

Name Description Type Default Optional Enum
learning_rate Start learning rate. Defaults to linear scheduler. If left empty, will be chosen automatically based on the task type and model selected. number True
learning_rate_scheduler The scheduler type to use. If left empty, will be chosen automatically based on the task type and model selected. string True ['warmup_linear', 'warmup_cosine', 'warmup_cosine_with_restarts', 'warmup_polynomial', 'constant', 'warmup_constant']
warmup_steps Number of steps used for a linear warmup from 0 to learning_rate. If left empty, will be chosen automatically based on the task type and model selected. integer True

optimizer

Name Description Type Default Optional Enum
optimizer optimizer to be used while training. If left empty, will be chosen automatically based on the task type and model selected. string True ['adamw_hf', 'adamw', 'sgd', 'adafactor', 'adagrad']
weight_decay The weight decay to apply (if not zero) to all layers except all bias and LayerNorm weights in Adam, AdamW & SGD optimizer. If left empty, will be chosen automatically based on the task type and model selected. number True
extra_optim_args Optional additional arguments that are supplied to SGD Optimizer. The arguments should be semi-colon separated key value pairs and should be enclosed in double quotes. For example, "momentum=0.5; nesterov=True" for sgd. Please make sure to use a valid parameter names for the chosen optimizer. For exact parameter names, please refer https://pytorch.org/docs/1.13/generated/torch.optim.SGD.html#torch.optim.SGD for SGD. Parameters supplied in extra_optim_args will take precedence over the parameter supplied via other arguments such as weight_decay. If weight_decay is provided via "weight_decay" parameter and via extra_optim_args both, values specified in extra_optim_args will be used. string True

gradient accumulation

Name Description Type Default Optional Enum
gradient_accumulation_step Number of update steps to accumulate the gradients for, before performing a backward/update pass. If left empty, will be chosen automatically based on the task type and model selected. integer True

mixed precision training

Name Description Type Default Optional Enum
precision Apply mixed precision training. This can reduce memory footprint by performing operations in half-precision. string 32 True ['32', '16']

metric thresholds

Name Description Type Default Optional Enum
iou_threshold IOU threshold used during inference in non-maximum suppression post processing. number True
box_score_threshold During inference, only return proposals with a score greater than box_score_threshold. \ The score is the multiplication of the objectness score and classification probability. number True

random seed

Name Description Type Default Optional Enum
random_seed Random seed that will be set at the beginning of training. integer 42 True

evaluation strategy parameters

Name Description Type Default Optional Enum
evaluation_strategy The evaluation strategy to adopt during training. Please note that the save_strategy and evaluation_strategy should match. string epoch True ['epoch', 'steps']
evaluation_steps Number of update steps between two evals if evaluation_strategy='steps'. Please note that the saving steps should be a multiple of the evaluation steps. integer 500 True

logging strategy parameters

Name Description Type Default Optional Enum
logging_strategy The logging strategy to adopt during training. string epoch True ['epoch', 'steps']
logging_steps Number of update steps between two logs if logging_strategy='steps'. integer 500 True

Save strategy

Name Description Type Default Optional Enum
save_strategy The checkpoint save strategy to adopt during training. Please note that the save_strategy and evaluation_strategy should match. string epoch True ['epoch', 'steps']
save_steps Number of updates steps before two checkpoint saves if save_strategy="steps". Please note that the saving steps should be a multiple of the evaluation steps. integer 500 True

model checkpointing limit

Name Description Type Default Optional Enum
save_total_limit If a value is passed, will limit the total amount of checkpoints. Deletes the older checkpoints in output_dir. If the value is -1 saves all checkpoints". integer -1 True

Early Stopping Parameters

Name Description Type Default Optional Enum
early_stopping Enable early stopping. boolean False True
early_stopping_patience Stop training when the specified metric worsens for early_stopping_patience evaluation calls. integer 1 True

Grad Norm

Name Description Type Default Optional Enum
max_grad_norm Maximum gradient norm (for gradient clipping). If left empty, will be chosen automatically based on the task type and model selected. number True

resume from the input model

Name Description Type Default Optional Enum
resume_from_checkpoint Loads optimizer, Scheduler and Trainer state for finetuning if true. boolean False True
save_as_mlflow_model Save as mlflow model with pyfunc as flavour. boolean True True

Outputs

Name Description Type
mlflow_model_folder Output dir to save the finetune model as mlflow model. mlflow_model
pytorch_model_folder Output dir to save the finetune model as torch model. custom_model

Environment

azureml://registries/azureml/environments/acft-mmdetection-image-gpu/versions/3

Clone this wiki locally