zeroshot.py

from transformers.modeling_outputs import SequenceClassifierOutput
from transformers.modeling_utils import PreTrainedModel
from transformers.configuration_utils import PretrainedConfig
import torch
from transformers import ZeroShotClassificationPipeline


class CustomConfig(PretrainedConfig):
    model_type = "test-zeroshot"

    def __init__(self, **kwargs):
        super().__init__(**kwargs)


class CustomModel(PreTrainedModel):
    config_class = CustomConfig

    def __init__(self, config: CustomConfig):
        super().__init__(config)
        self.config = config
        self.embeddings = torch.nn.Embedding(num_embeddings=1, embedding_dim=1)

    def forward(self, **kwargs) -> SequenceClassifierOutput:

        return SequenceClassifierOutput(logits=torch.tensor([[1, 2, 3]]))


from transformers.pipelines import PIPELINE_REGISTRY

from transformers import AutoModelForSequenceClassification, TFAutoModelForSequenceClassification

if __name__ == "__main__":
    from transformers import pipeline
    classifier = pipeline("zero-shot-classification",
                          model="cl-tohoku/bert-base-japanese")
    from transformers import AutoConfig, AutoModel, AutoModelForImageClassification

    CustomConfig.register_for_auto_class()
    CustomModel.register_for_auto_class("AutoModel")

    p = ZeroShotClassificationPipeline(model=CustomModel(CustomConfig()),
                                       tokenizer=classifier.tokenizer)
    from huggingface_hub import Repository

    repo = Repository("zero-shot-classification",
                      clone_from="paulhindemith/zero-shot-classification")
    p.save_pretrained("zero-shot-classification")
    repo.push_to_hub()