Commit
•
ec45403
1
Parent(s):
6150653
Update README.md
Browse files
README.md
CHANGED
|
@@ -23,6 +23,18 @@ The Kaz-RoBERTa model was pretrained on the reunion of 2 datasets:
|
|
| 23 |
- [Conversational data] Preprocessed dialogs between Customer Support Team and clients of Beeline KZ (Veon Group)(https://beeline.kz/)
|
| 24 |
|
| 25 |
Together these datasets weigh 25GB of text.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 26 |
## Usage
|
| 27 |
|
| 28 |
You can use this model directly with a pipeline for masked language modeling:
|
|
|
|
| 23 |
- [Conversational data] Preprocessed dialogs between Customer Support Team and clients of Beeline KZ (Veon Group)(https://beeline.kz/)
|
| 24 |
|
| 25 |
Together these datasets weigh 25GB of text.
|
| 26 |
+
## Training procedure
|
| 27 |
+
|
| 28 |
+
### Preprocessing
|
| 29 |
+
|
| 30 |
+
The texts are tokenized using a byte version of Byte-Pair Encoding (BPE) and a vocabulary size of 52,000. The inputs of
|
| 31 |
+
the model take pieces of 512 contiguous tokens that may span over documents. The beginning of a new document is marked
|
| 32 |
+
with `<s>` and the end of one by `</s>`
|
| 33 |
+
|
| 34 |
+
### Pretraining
|
| 35 |
+
|
| 36 |
+
The model was trained on 2 V100 GPUs for 500K steps with a batch size of 128 and a sequence length of 512.
|
| 37 |
+
|
| 38 |
## Usage
|
| 39 |
|
| 40 |
You can use this model directly with a pipeline for masked language modeling:
|