YAML Metadata Warning:The pipeline tag "conversational" is not in the official list: text-classification, token-classification, table-question-answering, question-answering, zero-shot-classification, translation, summarization, feature-extraction, text-generation, fill-mask, sentence-similarity, text-to-speech, text-to-audio, automatic-speech-recognition, audio-to-audio, audio-classification, audio-text-to-text, voice-activity-detection, depth-estimation, image-classification, object-detection, image-segmentation, text-to-image, image-to-text, image-to-image, image-to-video, unconditional-image-generation, video-classification, reinforcement-learning, robotics, tabular-classification, tabular-regression, tabular-to-text, table-to-text, multiple-choice, text-ranking, text-retrieval, time-series-forecasting, text-to-video, image-text-to-text, image-text-to-image, image-text-to-video, visual-question-answering, document-question-answering, zero-shot-image-classification, graph-ml, mask-generation, zero-shot-object-detection, text-to-3d, image-to-3d, image-feature-extraction, video-text-to-text, keypoint-detection, visual-document-retrieval, any-to-any, video-to-video, other

blenderbot-1B-distill fine-tuned on the ESConv dataset and AugESC dataset.

See the original paper for details.

Usage example:

import torch
from transformers import AutoTokenizer
from transformers.models.blenderbot import BlenderbotTokenizer, BlenderbotForConditionalGeneration

def _norm(x):
 return ' '.join(x.strip().split())

tokenizer = BlenderbotTokenizer.from_pretrained('thu-coai/blenderbot-1B-augesc')
model = BlenderbotForConditionalGeneration.from_pretrained('thu-coai/blenderbot-1B-augesc')
model.eval()

utterances = [
 "I am having a lot of anxiety about quitting my current job. It is too stressful but pays well",
 "What makes your job stressful for you?",
 "I have to deal with many people in hard financial situations and it is upsetting",
 "Do you help your clients to make it to a better financial situation?",
 "I do, but often they are not going to get back to what they want. Many people are going to lose their home when safeguards are lifted",
]
input_sequence = ' '.join([' ' + e for e in utterances]) + tokenizer.eos_token # add space prefix and separate utterances with two spaces
input_ids = tokenizer.convert_tokens_to_ids(tokenizer.tokenize(input_sequence))[-128:]
input_ids = torch.LongTensor([input_ids])

model_output = model.generate(input_ids, num_beams=1, do_sample=True, top_p=0.9, num_return_sequences=5, return_dict=False)
generation = tokenizer.batch_decode(model_output, skip_special_tokens=True)
generation = [_norm(e) for e in generation]
print(generation)

utterances.append(generation[0]) # for future loop

Please kindly cite our papers if you use this model:

@inproceedings{liu-etal-2021-towards,
 title={Towards Emotional Support Dialog Systems},
 author={Liu, Siyang and 
 Zheng, Chujie and 
 Demasi, Orianna and 
 Sabour, Sahand and 
 Li, Yu and 
 Yu, Zhou and 
 Jiang, Yong and 
 Huang, Minlie},
 booktitle={ACL},
 year={2021}
}

@inproceedings{zheng-etal-2023-augesc,
 title={AugESC: Dialogue Augmentation with Large Language Models for Emotional Support Conversation},
 author={Zheng, Chujie and
 Sabour, Sahand and
 Wen, Jiaxin and
 Zhang, Zheng and
 Huang, Minlie},
 booktitle={Findings of ACL},
 year={2023}
}