metadata
language:
- en
license: apache-2.0
tags:
- t5-small
- text2text-generation
- natural language generation
- conversational system
- task-oriented dialog
datasets:
- ConvLab/sgd
- ConvLab/tm1
- ConvLab/tm2
- ConvLab/tm3
- ConvLab/multiwoz21
metrics:
- Slot Error Rate
- sacrebleu
model-index:
- name: t5-small-nlg-multiwoz21_sgd_tm1_tm2_tm3
results:
- task:
type: text2text-generation
name: natural language generation
dataset:
type: ConvLab/multiwoz21
name: MultiWOZ 2.1
split: test
revision: 5f55375edbfe0270c20bcf770751ad982c0e6614
metrics:
- type: Slot Error Rate
value: 3.2
name: SER
- type: sacrebleu
value: 35.6
name: BLEU
- task:
type: text2text-generation
name: natural language generation
dataset:
type: ConvLab/sgd
name: SGD
split: test
revision: 6e8c79b888b21cc658cf9c0ce128d263241cf70f
metrics:
- type: Slot Error Rate
value: 8.3
name: SER
- type: sacrebleu
value: 29.9
name: BLEU
- task:
type: text2text-generation
name: natural language generation
dataset:
type: ConvLab/tm1, ConvLab/tm2, ConvLab/tm3
name: TM1+TM2+TM3
split: test
metrics:
- type: Slot Error Rate
value: 2
name: SER
- type: sacrebleu
value: 51.3
name: BLEU
widget:
- text: >-
[inform][restaurant]([area][centre],[food][Indian],[choice][nine]);[request][restaurant]([price
range][])
system:
example_title: MultiWOZ 2.1
- text: >-
sgd: [confirm][Restaurants_2]([number_of_seats][2],[restaurant_name][P.f.
Chang's],[location][Corte Madera],[time][12 pm],[date][March 8th])
system:
example_title: Schema-Guided Dialog
- text: |-
tm1: [inform][pizza_ordering]([name.store][Domino's])
system:
example_title: Taskmaster-1
- text: >-
tm2: [inform][restaurant-search]([name.restaurant][Via 313, the Violet
Crown Social Club],[price_range][$8 per slice])
system:
example_title: Taskmaster-2
- text: |-
tm3: [inform][movie]([name.movie][Star Wars],[name.movie][The Grudge])
system:
example_title: Taskmaster-3
inference:
parameters:
max_length: 100
t5-small-nlg-multiwoz21_sgd_tm1_tm2_tm3
This model is a fine-tuned version of t5-small on MultiWOZ 2.1, Schema-Guided Dialog, 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: 8
- total_train_batch_size: 512
- 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