Commit ·
6957d8f
1
Parent(s): 8afc850
Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language:
|
| 3 |
+
- ms
|
| 4 |
+
---
|
| 5 |
+
|
| 6 |
+
# Mosaic format for filtered starcoder 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-starcoder.ipynb
|
| 11 |
+
2. using tokenizer https://huggingface.co/malaysia-ai/bpe-tokenizer
|
| 12 |
+
3. 4096 context length.
|
| 13 |
+
|
| 14 |
+
## how-to
|
| 15 |
+
|
| 16 |
+
1. git clone,
|
| 17 |
+
|
| 18 |
+
```bash
|
| 19 |
+
git lfs clone https://huggingface.co/datasets/malaysia-ai/mosaic-starcoder-filtered
|
| 20 |
+
```
|
| 21 |
+
|
| 22 |
+
2. load it,
|
| 23 |
+
|
| 24 |
+
```python
|
| 25 |
+
from streaming import LocalDataset
|
| 26 |
+
import numpy as np
|
| 27 |
+
from streaming.base.format.mds.encodings import Encoding, _encodings
|
| 28 |
+
|
| 29 |
+
class UInt16(Encoding):
|
| 30 |
+
def encode(self, obj) -> bytes:
|
| 31 |
+
return obj.tobytes()
|
| 32 |
+
|
| 33 |
+
def decode(self, data: bytes):
|
| 34 |
+
return np.frombuffer(data, np.uint16)
|
| 35 |
+
|
| 36 |
+
_encodings['uint16'] = UInt16
|
| 37 |
+
|
| 38 |
+
dataset = LocalDataset('mosaic-starcoder-filtered')
|
| 39 |
+
len(dataset)
|
| 40 |
+
```
|