File size: 2,719 Bytes
53e90a8
af4f81b
 
d3bca73
53e90a8
af4f81b
 
 
 
 
 
 
 
 
 
d3bca73
 
 
 
 
 
 
 
 
 
 
 
 
 
af4f81b
 
4986ecd
af4f81b
 
 
 
 
d3bca73
 
53e90a8
 
 
 
 
331d23f
 
 
53e90a8
a4758a9
 
53e90a8
 
331d23f
 
53e90a8
 
 
d0dbeff
53e90a8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
---
language:
- en
license: apache-2.0
tags:
- t5
- qa
- askscience
- lfqa
- information retrieval
datasets:
- vblagoje/lfqa
metrics:
- rouge
widget:
- text: why hasn't humanity expanded to live on other planets in our solar system?
  example_title: solar system
- text: 'question: what is a probability distribution? context: I am just learning
    about statistics.'
  example_title: probability distribution
- text: 'question: What are the underlying physical processes by which exercise helps
    us lose weight? context: I started working out two weeks ago and already feel
    a lot better, and started to think about it and became deeply confused.'
  example_title: pumpen
- text: what is a neural network?
  example_title: deep learning
- text: What is the process that computers use to understand human language in deep
    learning models?
  example_title: NLP
inference:
  parameters:
    max_length: 64
    no_repeat_ngram_size: 2
    encoder_no_repeat_ngram_size: 4
    repetition_penalty: 3.51
    length_penalty: 0.8
    num_beams: 4
    early_stopping: true
base_model: google/t5-v1_1-base
---


# checkpoints

- This model is a fine-tuned version of [google/t5-v1_1-base](https://huggingface.co/google/t5-v1_1-base) on the `vblagoje/lfqa` dataset, with training duration of 2 epochs, for a (_somewhat_) apples-to-apples comparison with [t5-base](https://huggingface.co/pszemraj/t5-base-askscience) on the standard eli5 dataset.
  - This checkpoint does seem to be more coherent than t5-base on the original dataset.
- Compared to [bart on lfqa](https://huggingface.co/vblagoje/bart_lfqa), it seems to be able to respond to some questions independently of retrieval.

> NOTE: the inference API is limited to generating approx. 64 chars for runtime reasons, for longer outputs try using it in python as a transformers pipeline object.

## Intended uses & limitations

- Q&A, information retrieval
- it is probably better to use it with a [retrieval pipeline](https://github.com/deepset-ai/haystack) than alone

## Training and evaluation data

- see linked dataset. the dataset was filtered to only included the `askscience` subreddit in an attempt to focus on academic/technical queries.

## Training procedure

### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 4e-05
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- distributed_type: multi-GPU
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- num_epochs: 2

### Training results



### Framework versions

- Transformers 4.16.2
- Pytorch 1.10.0+cu113
- Datasets 1.18.3
- Tokenizers 0.11.0