Update README.md
Browse files
README.md
CHANGED
@@ -2,51 +2,80 @@
|
|
2 |
license: apache-2.0
|
3 |
base_model: google/mobilebert-uncased
|
4 |
tags:
|
5 |
-
-
|
|
|
|
|
|
|
6 |
metrics:
|
7 |
- f1
|
8 |
widget:
|
9 |
-
|
10 |
-
|
11 |
-
|
12 |
-
|
13 |
-
|
14 |
-
|
15 |
-
|
16 |
-
|
17 |
-
|
18 |
-
|
19 |
-
|
20 |
-
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
|
|
|
|
|
27 |
---
|
28 |
|
29 |
-
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
30 |
-
should probably proofread and complete it, then remove this comment. -->
|
31 |
|
32 |
-
# mobilebert-uncased-
|
33 |
|
34 |
-
This model is a fine-tuned version of [google/mobilebert-uncased](https://huggingface.co/google/mobilebert-uncased)
|
35 |
-
It achieves the following results on the evaluation set:
|
36 |
-
- Loss: 0.2658
|
37 |
-
- F1: 0.5395
|
38 |
|
39 |
## Model description
|
40 |
|
41 |
-
|
42 |
-
|
43 |
-
|
44 |
-
|
45 |
-
|
46 |
-
|
47 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
48 |
|
49 |
-
|
|
|
|
|
50 |
|
51 |
## Training procedure
|
52 |
|
@@ -64,25 +93,10 @@ The following hyperparameters were used during training:
|
|
64 |
- lr_scheduler_warmup_ratio: 0.03
|
65 |
- num_epochs: 10.0
|
66 |
|
67 |
-
### Training results
|
68 |
-
|
69 |
-
| Training Loss | Epoch | Step | Validation Loss | F1 |
|
70 |
-
|:-------------:|:-----:|:----:|:---------------:|:------:|
|
71 |
-
| 0.3118 | 0.99 | 123 | 0.2938 | 0.3004 |
|
72 |
-
| 0.2732 | 2.0 | 247 | 0.2774 | 0.4601 |
|
73 |
-
| 0.2705 | 3.0 | 371 | 0.2660 | 0.4703 |
|
74 |
-
| 0.2671 | 3.99 | 494 | 0.2679 | 0.4754 |
|
75 |
-
| 0.2529 | 5.0 | 618 | 0.2655 | 0.5163 |
|
76 |
-
| 0.2395 | 6.0 | 742 | 0.2637 | 0.5213 |
|
77 |
-
| 0.2469 | 6.99 | 865 | 0.2650 | 0.5235 |
|
78 |
-
| 0.2301 | 8.0 | 989 | 0.2636 | 0.5251 |
|
79 |
-
| 0.2273 | 9.0 | 1113 | 0.2648 | 0.5394 |
|
80 |
-
| 0.2169 | 9.94 | 1230 | 0.2658 | 0.5395 |
|
81 |
-
|
82 |
|
83 |
### Framework versions
|
84 |
|
85 |
- Transformers 4.35.0.dev0
|
86 |
- Pytorch 2.0.1+cpu
|
87 |
- Datasets 2.14.5
|
88 |
-
- Tokenizers 0.14.0
|
|
|
2 |
license: apache-2.0
|
3 |
base_model: google/mobilebert-uncased
|
4 |
tags:
|
5 |
+
- dataset tools
|
6 |
+
- books
|
7 |
+
- book
|
8 |
+
- genre
|
9 |
metrics:
|
10 |
- f1
|
11 |
widget:
|
12 |
+
- text: The Quantum Chip
|
13 |
+
example_title: Science Fiction & Fantasy
|
14 |
+
- text: One Dollar's Journey
|
15 |
+
example_title: Business & Finance
|
16 |
+
- text: Timmy The Talking Tree
|
17 |
+
example_title: idk fiction
|
18 |
+
- text: The Cursed Canvas
|
19 |
+
example_title: Arts & Design
|
20 |
+
- text: Hoops and Hegel
|
21 |
+
example_title: Philosophy & Religion
|
22 |
+
- text: Overview of Streams in North Dakota
|
23 |
+
example_title: Nature
|
24 |
+
- text: Advanced Topology
|
25 |
+
example_title: Non-fiction/Math
|
26 |
+
- text: Cooking Up Love
|
27 |
+
example_title: Food & Cooking
|
28 |
+
- text: Dr. Doolittle's Extraplanatary Commute
|
29 |
+
example_title: Science & Technology
|
30 |
+
language:
|
31 |
+
- en
|
32 |
---
|
33 |
|
|
|
|
|
34 |
|
35 |
+
# mobilebert-uncased-title2genre
|
36 |
|
37 |
+
This model is a fine-tuned version of [google/mobilebert-uncased](https://huggingface.co/google/mobilebert-uncased) for multi-label classification (18 labels).
|
|
|
|
|
|
|
38 |
|
39 |
## Model description
|
40 |
|
41 |
+
This classifies one or more **genre** labels in a **multi-label** setting for a given book **title**.
|
42 |
+
|
43 |
+
The 'standard' way of interpreting the predictions is that the predicted labels for a given example are **only the ones with a greater than 50% probability.**
|
44 |
+
|
45 |
+
## Details
|
46 |
+
|
47 |
+
### Labels
|
48 |
+
|
49 |
+
There are 18 labels, these are already integrated into the `config.json` and should be output by the model:
|
50 |
+
|
51 |
+
```json
|
52 |
+
"id2label": {
|
53 |
+
"0": "History & Politics",
|
54 |
+
"1": "Health & Medicine",
|
55 |
+
"2": "Mystery & Thriller",
|
56 |
+
"3": "Arts & Design",
|
57 |
+
"4": "Self-Help & Wellness",
|
58 |
+
"5": "Sports & Recreation",
|
59 |
+
"6": "Non-Fiction",
|
60 |
+
"7": "Science Fiction & Fantasy",
|
61 |
+
"8": "Countries & Geography",
|
62 |
+
"9": "Other",
|
63 |
+
"10": "Nature & Environment",
|
64 |
+
"11": "Business & Finance",
|
65 |
+
"12": "Romance",
|
66 |
+
"13": "Philosophy & Religion",
|
67 |
+
"14": "Literature & Fiction",
|
68 |
+
"15": "Science & Technology",
|
69 |
+
"16": "Children & Young Adult",
|
70 |
+
"17": "Food & Cooking"
|
71 |
+
},
|
72 |
+
```
|
73 |
+
|
74 |
+
### Eval results (validation)
|
75 |
|
76 |
+
It achieves the following results on the evaluation set:
|
77 |
+
- Loss: 0.2658
|
78 |
+
- F1: 0.5395
|
79 |
|
80 |
## Training procedure
|
81 |
|
|
|
93 |
- lr_scheduler_warmup_ratio: 0.03
|
94 |
- num_epochs: 10.0
|
95 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
96 |
|
97 |
### Framework versions
|
98 |
|
99 |
- Transformers 4.35.0.dev0
|
100 |
- Pytorch 2.0.1+cpu
|
101 |
- Datasets 2.14.5
|
102 |
+
- Tokenizers 0.14.0
|