Upload folder using huggingface_hub
Browse files- README.md +14 -0
- data-00000-of-00001.arrow +3 -0
README.md
ADDED
|
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: odc-by
|
| 3 |
+
language:
|
| 4 |
+
- he
|
| 5 |
+
- en
|
| 6 |
+
pretty_name: DictaLM2.0 Quantization Calibration Dataset
|
| 7 |
+
---
|
| 8 |
+
|
| 9 |
+
# DictaLM 2.0 Quantization Calibration Dataset
|
| 10 |
+
|
| 11 |
+
This dataset consists of documents which can be used as a calibration dataset for quantizing Hebrew-English LLMs, such as [DictaLM 2.0](https://huggingface.co/dicta-il/dictalm2.0). You can read more about the model on the [release page](https://dicta.org.il/dicta-lm) or in the [technical report](https://arxiv.org/abs/2407.07080).
|
| 12 |
+
|
| 13 |
+
The dataset consists of 1024 documents - 512 in English (FineWeb-EDU) and 512 in Hebrew (Wikipedia, OSCAR).
|
| 14 |
+
|
data-00000-of-00001.arrow
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ddc08a26ed6ec97a9652fd7161659eb2d03dd57ba95b89f13174a8c12b74d339
|
| 3 |
+
size 6302472
|