buruzaemon
commited on
Merge branch 'main' of https://huggingface.co/buruzaemon/distilbert-base-uncased-finetuned-clinc
Browse files
README.md
CHANGED
@@ -34,7 +34,7 @@ It achieves the following results on the evaluation set:
|
|
34 |
|
35 |
## Model description
|
36 |
|
37 |
-
|
38 |
|
39 |
## Intended uses & limitations
|
40 |
|
@@ -42,13 +42,17 @@ More information needed
|
|
42 |
|
43 |
## Training and evaluation data
|
44 |
|
45 |
-
|
46 |
|
47 |
## Training procedure
|
48 |
|
|
|
|
|
49 |
### Training hyperparameters
|
50 |
|
51 |
The following hyperparameters were used during training:
|
|
|
|
|
52 |
- learning_rate: 2e-05
|
53 |
- train_batch_size: 48
|
54 |
- eval_batch_size: 48
|
|
|
34 |
|
35 |
## Model description
|
36 |
|
37 |
+
This is an initial example of knowledge-distillation where the student loss is all cross-entropy loss \\(L_{CE}\\) of the ground-truth labels and none of the distillation loss \\(L_{KD}\\).
|
38 |
|
39 |
## Intended uses & limitations
|
40 |
|
|
|
42 |
|
43 |
## Training and evaluation data
|
44 |
|
45 |
+
The training and evaluation data come straight from the `train` and `validation` splits in the clinc_oos dataset, respectively; and tokenized using the `distilbert-base-uncased` tokenization.
|
46 |
|
47 |
## Training procedure
|
48 |
|
49 |
+
Please see page 224 in Chapter 8: Making Transformers Efficient in Production, Natural Language Processing with Transformers, May 2022.
|
50 |
+
|
51 |
### Training hyperparameters
|
52 |
|
53 |
The following hyperparameters were used during training:
|
54 |
+
- alpha: 1.0
|
55 |
+
- temperature: 2.0
|
56 |
- learning_rate: 2e-05
|
57 |
- train_batch_size: 48
|
58 |
- eval_batch_size: 48
|