Commit
·
e8f23cc
1
Parent(s):
b484bef
Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,39 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
language:
|
3 |
+
- ms
|
4 |
+
---
|
5 |
+
|
6 |
+
# Mosaic format for combine all dataset to train Malaysian LLM
|
7 |
+
|
8 |
+
This repository is to store dataset shards using mosaic format.
|
9 |
+
|
10 |
+
1. prepared at https://github.com/malaysia-ai/dedup-text-dataset/blob/main/pretrain-llm/combine-all.ipynb
|
11 |
+
2. using tokenizer https://huggingface.co/malaysia-ai/bpe-tokenizer
|
12 |
+
|
13 |
+
## how-to
|
14 |
+
|
15 |
+
1. git clone,
|
16 |
+
|
17 |
+
```bash
|
18 |
+
git lfs clone https://huggingface.co/datasets/malaysia-ai/mosaic-combine-all
|
19 |
+
```
|
20 |
+
|
21 |
+
2. load it,
|
22 |
+
|
23 |
+
```python
|
24 |
+
from streaming import LocalDataset
|
25 |
+
import numpy as np
|
26 |
+
from streaming.base.format.mds.encodings import Encoding, _encodings
|
27 |
+
|
28 |
+
class UInt16(Encoding):
|
29 |
+
def encode(self, obj) -> bytes:
|
30 |
+
return obj.tobytes()
|
31 |
+
|
32 |
+
def decode(self, data: bytes):
|
33 |
+
return np.frombuffer(data, np.uint16)
|
34 |
+
|
35 |
+
_encodings['uint16'] = UInt16
|
36 |
+
|
37 |
+
dataset = LocalDataset('mosaic-combine-all')
|
38 |
+
len(dataset)
|
39 |
+
```
|