components oss_text_generation_finetune

OSS Text Generation Finetune

oss_text_generation_finetune

Overview

FTaaS component to finetune model for Text Generation task

Version: 0.0.15

View in Studio: https://ml.azure.com/registries/azureml/components/oss_text_generation_finetune/version/0.0.15

Inputs

Name	Description	Type	Default	Optional	Enum
task_name	Finetune task name.	string	TextGeneration	True	['TextGeneration']
mlflow_model_path	Input folder path containing mlflow model for further finetuning. Proper model/huggingface id must be passed.	mlflow_model		True
model_asset_id	Asset id of model	string		False
dataset_input	Output of data import component. The folder contains train and validation data.	uri_folder		False
text_key	key for text in an example. format your data keeping in mind that text is concatenated with ground_truth while finetuning in the form - text + groundtruth. for eg. "text"="knock knock\n", "ground_truth"="who's there"; will be treated as "knock knock\nwho's there"	string		False
ground_truth_key	key for ground_truth in an example. we take separate column for ground_truth to enable use cases like summarization, translation, question_answering, etc. which can be repurposed in form of text-generation where both text and ground_truth are needed. This separation is useful for calculating metrics. for eg. "text"="Summarize this dialog:\n{input_dialogue}\nSummary:\n", "ground_truth"="{summary of the dialogue}"	string		True
batch_size	Number of examples to batch before calling the tokenization function	integer	1000	True
pad_to_max_length	If set to True, the returned sequences will be padded according to the model's padding side and padding index, up to their max_seq_length. If no max_seq_length is specified, the padding is done up to the model's max length.	string	false	True	['true', 'false']
max_seq_length	Default is -1 which means the padding is done up to the model's max length. Else will be padded to max_seq_length.	integer	-1	True
number_of_gpu_to_use_finetuning	number of gpus to be used per node for finetuning, should be equal to number of gpu per node in the compute SKU used for finetune	integer	1	False
apply_lora	lora enabled	string	false	True	['true', 'false']
merge_lora_weights	if set to true, the lora trained weights will be merged to base model before saving	string	true	True	['true', 'false']
lora_alpha	lora attention alpha	integer	128	True
lora_r	lora dimension	integer	8	True
lora_dropout	lora dropout value	number	0.0	True
num_train_epochs	training epochs	integer	1	True
max_steps	If set to a positive number, the total number of training steps to perform. Overrides '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.	integer	-1	True
per_device_train_batch_size	Train batch size	integer	1	True
per_device_eval_batch_size	Validation batch size	integer	1	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	string	false	True	['true', 'false']
optim	Optimizer to be used while training	string	adamw_hf	True	['adamw_hf', 'adamw_torch', 'adafactor']
learning_rate	Start learning rate. Defaults to linear scheduler.	number	2e-05	True
warmup_steps	Number of steps used for a linear warmup from 0 to learning_rate	integer	0	True
weight_decay	The weight decay to apply (if not zero) to all layers except all bias and LayerNorm weights in AdamW optimizer	number	0.0	True
adam_beta1	The beta1 hyperparameter for the AdamW optimizer	number	0.9	True
adam_beta2	The beta2 hyperparameter for the AdamW optimizer	number	0.999	True
adam_epsilon	The epsilon hyperparameter for the AdamW optimizer	number	1e-08	True
gradient_accumulation_steps	Number of updates steps to accumulate the gradients for, before performing a backward/update pass	integer	1	True
eval_accumulation_steps	Number of predictions steps to accumulate before moving the tensors to the CPU, will be passed as None if set to -1	integer	-1	True
lr_scheduler_type	learning rate scheduler to use.	string	linear	True	['linear', 'cosine', 'cosine_with_restarts', 'polynomial', 'constant', 'constant_with_warmup']
precision	Apply mixed precision training. This can reduce memory footprint by performing operations in half-precision.	string	32	True	['32', '16']
seed	Random seed that will be set at the beginning of training	integer	42	True
enable_full_determinism	Ensure reproducible behavior during distributed training	string	false	True	['true', 'false']
dataloader_num_workers	Number of subprocesses to use for data loading. 0 means that the data will be loaded in the main process.	integer	0	True
ignore_mismatched_sizes	Whether or not to raise an error if some of the weights from the checkpoint do not have the same size as the weights of the model	string	false	True	['true', 'false']
max_grad_norm	Maximum gradient norm (for gradient clipping)	number	1.0	True
evaluation_strategy	The evaluation strategy to adopt during training	string	epoch	True	['epoch', 'steps']
evaluation_steps_interval	The evaluation steps in fraction of an epoch steps to adopt during training. Overwrites evaluation_steps if not 0.	number	0.0	True
eval_steps	Number of update steps between two evals if evaluation_strategy='steps'	integer	500	True
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
metric_for_best_model	Specify the metric to use to compare two different models	string	loss	True	['loss']
resume_from_checkpoint	Loads Optimizer, Scheduler and Trainer state for finetuning if true	string	false	True	['true', 'false']
save_strategy	The checkpoint save strategy to adopt during training	string	evaluation_strategy	True	['evaluation_strategy', 'epoch', 'steps']
save_steps	Number of update steps between two checkpoint saves if save_strategy='steps'	integer	100	True
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
apply_early_stopping	Enable early stopping	string	false	True	['true', 'false']
early_stopping_patience	Stop training when the specified metric worsens for early_stopping_patience evaluation calls	integer	1	True
early_stopping_threshold	Denotes how much the specified metric must improve to satisfy early stopping conditions	number	0.0	True
apply_deepspeed	If set to true, will enable deepspeed for training	string	false	True	['true', 'false']
deepspeed_stage	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.	string	2	True	['2', '3']
apply_ort	If set to true, will use the ONNXRunTime training	string	false	True	['true', 'false']
system_properties	Validation parameters propagated from pipeline.	string		True
registered_model_name	Name of the registered model	string		True

Outputs

Name	Description	Type
output_model	Output dir to save the finetuned lora weights	uri_folder
intermediate_folder	Folder to store intermediate outputs like model selector output, preprocess output, checkpoints, etc. to preserve them across Singularity preemptions.	uri_folder

Environment

azureml://registries/azureml/environments/acft-hf-nlp-gpu/versions/58