Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,37 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: cc-by-nc-sa-4.0
|
| 3 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: cc-by-nc-sa-4.0
|
| 3 |
+
---
|
| 4 |
+
|
| 5 |
+
This dataset contains the embeddings for the 100B variant of [fineweb-edu](https://huggingface.co/datasets/HuggingFaceFW/fineweb-edu), embedded with the Cohere Embed V3 model.
|
| 6 |
+
|
| 7 |
+
You can search on this dataset with just 500MB of memory using [DiskVectorIndex](https://github.com/cohere-ai/DiskVectorIndex).
|
| 8 |
+
|
| 9 |
+
# Installation & Usage
|
| 10 |
+
|
| 11 |
+
Get your free **Cohere API key** from [cohere.com](https://cohere.com). You must set this API key as an environment variable:
|
| 12 |
+
```
|
| 13 |
+
export COHERE_API_KEY=your_api_key
|
| 14 |
+
```
|
| 15 |
+
|
| 16 |
+
Install the package:
|
| 17 |
+
```
|
| 18 |
+
pip install DiskVectorIndex
|
| 19 |
+
```
|
| 20 |
+
|
| 21 |
+
You can then search via:
|
| 22 |
+
```python
|
| 23 |
+
from DiskVectorIndex import DiskVectorIndex
|
| 24 |
+
|
| 25 |
+
index = DiskVectorIndex("Cohere/fineweb-edu-100B-index")
|
| 26 |
+
|
| 27 |
+
while True:
|
| 28 |
+
query = input("\n\nEnter a question: ")
|
| 29 |
+
docs = index.search(query, top_k=3)
|
| 30 |
+
for doc in docs:
|
| 31 |
+
print(doc)
|
| 32 |
+
print("=========")
|
| 33 |
+
```
|
| 34 |
+
|
| 35 |
+
# License
|
| 36 |
+
|
| 37 |
+
Please observe the License for the [Fineweb-edu](https://huggingface.co/datasets/HuggingFaceFW/fineweb-edu). The license displayed here is just for the embeddings.
|