metadata
language:
- en
license: apache-2.0
tags:
- t5-small
- text2text-generation
- dialog state tracking
- conversational system
- task-oriented dialog
datasets:
- ConvLab/tm1
- ConvLab/tm2
- ConvLab/tm3
metrics:
- Joint Goal Accuracy
- Slot F1
model-index:
- name: t5-small-dst-tm1_tm2_tm3
results:
- task:
type: text2text-generation
name: dialog state tracking
dataset:
type: ConvLab/tm1, ConvLab/tm2, ConvLab/tm3
name: TM1+TM2+TM3
split: test
metrics:
- type: Joint Goal Accuracy
value: 48.5
name: JGA
- type: Slot F1
value: 81.1
name: Slot F1
widget:
- text: |-
tm1: 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: >-
tm2: user: I need help finding a hotel in New Orleans.
system: Okay.
user: I need something that's around $300 a night and it's a five star
rating.
- text: |-
tm3: user: Hi, I'm hoping to see a movie tonight.
system: Great, I can assist with that. What genre of film do you prefer.
user: I usually like comedies.
inference:
parameters:
max_length: 100
t5-small-dst-tm1_tm2_tm3
This model is a fine-tuned version of t5-small on Taskmaster-1, Taskmaster-2, and Taskmaster-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: 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