zhuqi's picture
Create README.md
f9f8935
metadata
language:
  - en
license: apache-2.0
tags:
  - t5-small
  - text2text-generation
  - natural language understanding
  - conversational system
  - task-oriented dialog
datasets:
  - ConvLab/multiwoz21
metrics:
  - Dialog acts Accuracy
  - Dialog acts F1
model-index:
  - name: t5-small-nlu-all-multiwoz21-context3
    results:
      - task:
          type: text2text-generation
          name: natural language understanding
        dataset:
          type: ConvLab/multiwoz21
          name: MultiWOZ 2.1
          split: test
          revision: 5f55375edbfe0270c20bcf770751ad982c0e6614
        metrics:
          - type: Dialog acts Accuracy
            value: 73.6
            name: Accuracy
          - type: Dialog acts F1
            value: 86.9
            name: F1
widget:
  - text: >-
      user: I would like a taxi from Saint John's college to Pizza Hut Fen
      Ditton.

      system: What time do you want to leave and what time do you want to arrive
      by?

      user: I want to leave after 17:15.
  - text: >-
      user: I want to find a moderately priced restaurant. 

      system: I have many options available for you! Is there a certain area or
      cuisine that interests you?

      user: Yes I would like the restaurant to be located in the center of the
      attractions. 

      system: There are 21 restaurants available in the centre of town. How
      about a specific type of cuisine?
inference:
  parameters:
    max_length: 100

t5-small-nlu-all-multiwoz21-context3

This model is a fine-tuned version of t5-small on MultiWOZ 2.1 both user and system utterances with context window size == 3.

Refer to ConvLab-3 for model description and usage.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 0.001
  • train_batch_size: 128
  • eval_batch_size: 64
  • seed: 42
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 256
  • optimizer: Adafactor
  • lr_scheduler_type: linear
  • num_epochs: 10.0

Framework versions

  • Transformers 4.20.1
  • Pytorch 1.11.0+cu102
  • Datasets 2.3.2
  • Tokenizers 0.12.1