Text Classification
Transformers
Safetensors
English
distilbert
encryption
equivariant-encryption
nesa
text-embeddings-inference
Instructions to use nesaorg/distilbert-sentiment-encrypted with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use nesaorg/distilbert-sentiment-encrypted with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-classification", model="nesaorg/distilbert-sentiment-encrypted")# Load model directly from transformers import AutoTokenizer, AutoModelForSequenceClassification tokenizer = AutoTokenizer.from_pretrained("nesaorg/distilbert-sentiment-encrypted") model = AutoModelForSequenceClassification.from_pretrained("nesaorg/distilbert-sentiment-encrypted") - Notebooks
- Google Colab
- Kaggle
Update README.md
Browse files
README.md
CHANGED
|
@@ -29,7 +29,7 @@ Developed at Nesa, Equivariant Encryption allows LLMs to run inference on encryp
|
|
| 29 |
## Limitations of the Community Version
|
| 30 |
|
| 31 |
Due to the approximations involved, the results of this model are only expected to reproduce the results of the original model about 92% of the time with a small change in the confidence scores as well.
|
| 32 |
-
This public version demonstrates that inference with our encryption scheme results in zero additional latency. We encourage the community to benchmark the encrypted version with the original version confirm this level of fidelity. We also invite people to compare our encrypted model weights against the original model.
|
| 33 |
|
| 34 |
## Running the model
|
| 35 |
|
|
|
|
| 29 |
## Limitations of the Community Version
|
| 30 |
|
| 31 |
Due to the approximations involved, the results of this model are only expected to reproduce the results of the original model about 92% of the time with a small change in the confidence scores as well.
|
| 32 |
+
This public version demonstrates that inference with our encryption scheme results in zero additional latency. We encourage the community to benchmark the encrypted version with the original version to confirm this level of fidelity. We also invite people to compare our encrypted model weights against the original model.
|
| 33 |
|
| 34 |
## Running the model
|
| 35 |
|