Update README.md
Browse files
README.md
CHANGED
|
@@ -33,24 +33,29 @@ pip install onnxruntime transformers
|
|
| 33 |
|
| 34 |
### Python Example
|
| 35 |
```python
|
| 36 |
-
|
| 37 |
-
from transformers import AutoTokenizer
|
| 38 |
import onnxruntime as ort
|
| 39 |
import numpy as np
|
| 40 |
|
| 41 |
-
# 1. Load
|
| 42 |
-
tokenizer =
|
| 43 |
|
| 44 |
-
# 2. Load
|
| 45 |
session = ort.InferenceSession("model.onnx")
|
| 46 |
|
| 47 |
-
# 3. Preprocess
|
| 48 |
-
text = "
|
| 49 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 50 |
|
| 51 |
-
# 4. Inference
|
| 52 |
-
outputs = session.run(None,
|
| 53 |
-
print(outputs[0])
|
| 54 |
|
| 55 |
```
|
| 56 |
|
|
|
|
| 33 |
|
| 34 |
### Python Example
|
| 35 |
```python
|
| 36 |
+
from tokenizers import Tokenizer
|
|
|
|
| 37 |
import onnxruntime as ort
|
| 38 |
import numpy as np
|
| 39 |
|
| 40 |
+
# 1. Load the lightweight tokenizer (No Transformers dependency needed)
|
| 41 |
+
tokenizer = Tokenizer.from_pretrained("broadfield-dev/bert-small-ner-pii-tuned-12261022-onnx")
|
| 42 |
|
| 43 |
+
# 2. Load the ONNX model
|
| 44 |
session = ort.InferenceSession("model.onnx")
|
| 45 |
|
| 46 |
+
# 3. Preprocess (Simple text encoding)
|
| 47 |
+
text = "Run inference on mobile!"
|
| 48 |
+
encoding = tokenizer.encode(text)
|
| 49 |
+
|
| 50 |
+
# Prepare inputs (Exact names vary by model, usually input_ids + attention_mask)
|
| 51 |
+
inputs = {{
|
| 52 |
+
"input_ids": np.array([encoding.ids], dtype=np.int64),
|
| 53 |
+
"attention_mask": np.array([encoding.attention_mask], dtype=np.int64)
|
| 54 |
+
}}
|
| 55 |
|
| 56 |
+
# 4. Run Inference
|
| 57 |
+
outputs = session.run(None, inputs)
|
| 58 |
+
print("Output logits shape:", outputs[0].shape)
|
| 59 |
|
| 60 |
```
|
| 61 |
|