Add transformers.js example code (#6)
Browse files- Add transformers.js example code (f91abb96989460f391cea54e86c52049dd5599ff)
Co-authored-by: Joshua <Xenova@users.noreply.huggingface.co>
README.md
CHANGED
@@ -6,6 +6,7 @@ tags:
|
|
6 |
- sentence-transformers
|
7 |
- gte
|
8 |
- mteb
|
|
|
9 |
license: apache-2.0
|
10 |
language:
|
11 |
- en
|
@@ -2686,6 +2687,32 @@ embeddings = model.encode(sentences)
|
|
2686 |
print(cos_sim(embeddings[0], embeddings[1]))
|
2687 |
```
|
2688 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2689 |
## Training Details
|
2690 |
|
2691 |
### Training Data
|
|
|
6 |
- sentence-transformers
|
7 |
- gte
|
8 |
- mteb
|
9 |
+
- transformers.js
|
10 |
license: apache-2.0
|
11 |
language:
|
12 |
- en
|
|
|
2687 |
print(cos_sim(embeddings[0], embeddings[1]))
|
2688 |
```
|
2689 |
|
2690 |
+
Use with `transformers.js`:
|
2691 |
+
|
2692 |
+
```js
|
2693 |
+
// npm i @xenova/transformers
|
2694 |
+
import { pipeline, dot } from '@xenova/transformers';
|
2695 |
+
|
2696 |
+
// Create feature extraction pipeline
|
2697 |
+
const extractor = await pipeline('feature-extraction', 'Alibaba-NLP/gte-large-en-v1.5', {
|
2698 |
+
quantized: false, // Comment out this line to use the quantized version
|
2699 |
+
});
|
2700 |
+
|
2701 |
+
// Generate sentence embeddings
|
2702 |
+
const sentences = [
|
2703 |
+
"what is the capital of China?",
|
2704 |
+
"how to implement quick sort in python?",
|
2705 |
+
"Beijing",
|
2706 |
+
"sorting algorithms"
|
2707 |
+
]
|
2708 |
+
const output = await extractor(sentences, { normalize: true, pooling: 'cls' });
|
2709 |
+
|
2710 |
+
// Compute similarity scores
|
2711 |
+
const [source_embeddings, ...document_embeddings ] = output.tolist();
|
2712 |
+
const similarities = document_embeddings.map(x => 100 * dot(source_embeddings, x));
|
2713 |
+
console.log(similarities); // [41.86354093370361, 77.07076371259589, 37.02981979677899]
|
2714 |
+
```
|
2715 |
+
|
2716 |
## Training Details
|
2717 |
|
2718 |
### Training Data
|