zhuqi's picture
Create README.md
7b1959d
metadata
language:
  - en
license: apache-2.0
tags:
  - t5-small
  - text2text-generation
  - natural language understanding
  - conversational system
  - task-oriented dialog
datasets:
  - ConvLab/tm1
metrics:
  - Dialog acts Accuracy
  - Dialog acts F1
model-index:
  - name: t5-small-nlu-tm1-context3
    results:
      - task:
          type: text2text-generation
          name: natural language understanding
        dataset:
          type: ConvLab/tm1
          name: Taskmaster-1
          split: test
          revision: 187bd9f5e786d80f64b3d372386e330ae36d8488
        metrics:
          - type: Dialog acts Accuracy
            value: 76.2
            name: Accuracy
          - type: Dialog acts F1
            value: 56.2
            name: F1
widget:
  - text: |-
      user: Hi there, could you please help me with an order of Pizza?
      system: Sure, where would you like to order you pizza from?
      user: I would like to order a pizza from Domino's.
  - text: |-
      system: What kind of pizza are do you want to order?
      user: I want to order a large pizza with chicken and pepperoni please.
      system: From which Domino's location would you like to order?
      user: I would like to order from the Domino's closest to my house.
inference:
  parameters:
    max_length: 100

t5-small-nlu-tm1-context3

This model is a fine-tuned version of t5-small on Taskmaster-1 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 4.18.0
  • Pytorch 1.10.2+cu102
  • Datasets 1.18.3
  • Tokenizers 0.11.0