MCPcopy Create free account
hub / github.com/OpenBitSys/BitDistiller / make_supervised_data_module

Function make_supervised_data_module

data/generation/single_generate.py:143–147  ·  view source on GitHub ↗

Make dataset and collator for supervised fine-tuning.

(tokenizer: transformers.PreTrainedTokenizer, dataset_name, max_sample=None)

Source from the content-addressed store, hash-verified

141 )
142
143def make_supervised_data_module(tokenizer: transformers.PreTrainedTokenizer, dataset_name, max_sample=None) -> Dict:
144 """Make dataset and collator for supervised fine-tuning."""
145 gen_dataset = SupervisedDataset(tokenizer=tokenizer, dataset_name=dataset_name, max_sample=max_sample)
146 data_collator = DataCollatorForSupervisedDataset(tokenizer=tokenizer)
147 return gen_dataset, data_collator
148
149def main(args):
150 torch.manual_seed(args.seed)

Callers 1

mainFunction · 0.70

Calls 2

SupervisedDatasetClass · 0.70

Tested by

no test coverage detected