t5-small-dst-sgd / README.md
zhuqi's picture
Update README.md
4d11bc6
metadata
language:
  - en
license: apache-2.0
tags:
  - t5-small
  - text2text-generation
  - dialog state tracking
  - conversational system
  - task-oriented dialog
datasets:
  - ConvLab/sgd
metrics:
  - Joint Goal Accuracy
  - Slot F1
model-index:
  - name: t5-small-dst-sgd
    results:
      - task:
          type: text2text-generation
          name: dialog state tracking
        dataset:
          type: ConvLab/sgd
          name: SGD
          split: test
          revision: 6e8c79b888b21cc658cf9c0ce128d263241cf70f
        metrics:
          - type: Joint Goal Accuracy
            value: 20.1
            name: JGA
          - type: Slot F1
            value: 58.5
            name: Slot F1
widget:
  - text: >-
      user: Hi, could you get me a restaurant booking on the 8th please?

      system: Any preference on the restaurant, location and time?

      user: Could you get me a reservation at P.f. Chang's in Corte Madera at
      afternoon 12?
  - text: >-
      user: I need to book a dinner reservation for a date. Help me reserve a
      table at a restaurant.

      system: What time and location do you have in mind?

      user: Something around 8 in the night should be fine. Oh, and look in the
      San Jose area.
inference:
  parameters:
    max_length: 100

t5-small-dst-sgd

This model is a fine-tuned version of t5-small on Schema-Guided Dialog.

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

Framework versions

  • Transformers 4.20.1
  • Pytorch 1.11.0+cu113
  • Datasets 2.3.2
  • Tokenizers 0.12.1