cyvidia-ai-utils 0.5.0

No project description provided

ai-utils

Installation

pip install cyvidia-ai-utils

Cross validation

Huggingface Trainer API

from cyvidia_ai_utils import TransformerCrossValidationModel, cross_validate, EvaluationResult
from transformers import AutoTokenizer, AutoModelForSequenceClassification,TrainingArguments
from datasets import load_dataset, Dataset
from transformers import DataCollatorWithPadding,Trainer

folds = load_dataset("dipesh/Intent-Classification-small",split=[f"train[{k}%:{k+10}%]" for k in range(0, 100, 10)])
assert(isinstance(folds, list))


def create_trainer(model, tokenizer, train_ds: Dataset, val_ds:Dataset):
    def preprocess_function(examples):
        return tokenizer(examples["text"], truncation=True)

    tokenized_train_ds = train_ds.map(preprocess_function, batched=True)
    tokenized_val_ds = val_ds.map(preprocess_function, batched=True)
    data_collator = DataCollatorWithPadding(tokenizer=tokenizer)
    training_args = TrainingArguments(
        output_dir=f'tests/test_models/{uuid.uuid4()}',
        learning_rate=2e-5,
        per_device_train_batch_size=16,
        per_device_eval_batch_size=16,
        num_train_epochs=0.001,
        weight_decay=0.01,
        evaluation_strategy="epoch",
        save_strategy="epoch",
        load_best_model_at_end=True,
    )


    trainer = Trainer(
        model=model,
        args=training_args,
        train_dataset=tokenized_train_ds,
        eval_dataset=tokenized_val_ds,
        tokenizer=tokenizer,
        data_collator=data_collator,
    )

    return trainer



model = AutoModelForSequenceClassification.from_pretrained("prajjwal1/bert-tiny", num_labels=21)
tokenizer= AutoTokenizer.from_pretrained("prajjwal1/bert-tiny")

cross_val_model= TransformerCrossValidationModel(
    model= model,
    tokenizer= tokenizer,
    create_trainer= create_trainer
)


results= cross_validate(cross_val_model, folds, target_id_column="label", input_text_column="text")

aggragated_result= EvaluationResult.aggregate(list(results.values()))

Custom Trainer

from cyvidia_ai_utils import CrossValidationModel

class MyCustomCrossValidationModel(CrossValidationModel):
    def get_label_for_id(self, id: int)-> str:
        # Implement

    def train(self, train_ds, val_ds)-> CrossValidationModel:
        # Implement

    def predict_values(self, values)-> Dict[str,Any]:
        # Implement

results= cross_validate(MyCustomCrossValidationModel(), folds, target_id_column="label", input_text_column="text")