Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,72 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
language:
|
3 |
+
- en
|
4 |
+
license: apache-2.0
|
5 |
+
tags:
|
6 |
+
- t5-small
|
7 |
+
- text2text-generation
|
8 |
+
- natural language understanding
|
9 |
+
- conversational system
|
10 |
+
- task-oriented dialog
|
11 |
+
datasets:
|
12 |
+
- ConvLab/tm2
|
13 |
+
metrics:
|
14 |
+
- Dialog acts Accuracy
|
15 |
+
- Dialog acts F1
|
16 |
+
|
17 |
+
model-index:
|
18 |
+
- name: t5-small-nlu-tm2-context3
|
19 |
+
results:
|
20 |
+
- task:
|
21 |
+
type: text2text-generation
|
22 |
+
name: natural language understanding
|
23 |
+
dataset:
|
24 |
+
type: ConvLab/tm2
|
25 |
+
name: Taskmaster-2
|
26 |
+
split: test
|
27 |
+
revision: cdc314b156e7f7ffa81a1e7398f1f8a2e86c0095
|
28 |
+
metrics:
|
29 |
+
- type: Dialog acts Accuracy
|
30 |
+
value: 82.4
|
31 |
+
name: Accuracy
|
32 |
+
- type: Dialog acts F1
|
33 |
+
value: 74.3
|
34 |
+
name: F1
|
35 |
+
|
36 |
+
widget:
|
37 |
+
- text: "user: Hi, I'm looking for a flight. I need to visit a friend.\nsystem: Hello, how can I help you? Sure, I can help you with that. On what dates?\nuser: I'm looking to travel from March 20th to 22nd."
|
38 |
+
- text: "system: Anything else?\nuser: That should be everything.\nsystem: I found a flight for $424 on United Airlines.\nuser: Okay, is that for New York?"
|
39 |
+
|
40 |
+
inference:
|
41 |
+
parameters:
|
42 |
+
max_length: 100
|
43 |
+
|
44 |
+
---
|
45 |
+
|
46 |
+
# t5-small-nlu-tm2-context3
|
47 |
+
|
48 |
+
This model is a fine-tuned version of [t5-small](https://huggingface.co/t5-small) on [Taskmaster-2](https://huggingface.co/datasets/ConvLab/tm2) with context window size == 3.
|
49 |
+
|
50 |
+
Refer to [ConvLab-3](https://github.com/ConvLab/ConvLab-3) for model description and usage.
|
51 |
+
|
52 |
+
## Training procedure
|
53 |
+
|
54 |
+
### Training hyperparameters
|
55 |
+
|
56 |
+
The following hyperparameters were used during training:
|
57 |
+
- learning_rate: 0.001
|
58 |
+
- train_batch_size: 128
|
59 |
+
- eval_batch_size: 64
|
60 |
+
- seed: 42
|
61 |
+
- gradient_accumulation_steps: 2
|
62 |
+
- total_train_batch_size: 256
|
63 |
+
- optimizer: Adafactor
|
64 |
+
- lr_scheduler_type: linear
|
65 |
+
- num_epochs: 10.0
|
66 |
+
|
67 |
+
### Framework versions
|
68 |
+
|
69 |
+
- Transformers 4.18.0
|
70 |
+
- Pytorch 1.10.2+cu102
|
71 |
+
- Datasets 1.18.3
|
72 |
+
- Tokenizers 0.11.0
|