Add link to Neuron-optimized version
#39
by
badaoui
HF Staff
- opened
README.md
CHANGED
|
@@ -70,4 +70,16 @@ tokenizer = AutoTokenizer.from_pretrained("bert-base-chinese")
|
|
| 70 |
|
| 71 |
model = AutoModelForMaskedLM.from_pretrained("bert-base-chinese")
|
| 72 |
|
| 73 |
-
```
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 70 |
|
| 71 |
model = AutoModelForMaskedLM.from_pretrained("bert-base-chinese")
|
| 72 |
|
| 73 |
+
```
|
| 74 |
+
|
| 75 |
+
---
|
| 76 |
+
## π AWS Neuron Optimized Version Available
|
| 77 |
+
|
| 78 |
+
A Neuron-optimized version of this model is available for improved performance on AWS Inferentia/Trainium instances:
|
| 79 |
+
|
| 80 |
+
**[badaoui/google-bert-bert-base-chinese-neuron](https://huggingface.co/badaoui/google-bert-bert-base-chinese-neuron)**
|
| 81 |
+
|
| 82 |
+
The Neuron-optimized version provides:
|
| 83 |
+
- Pre-compiled artifacts for faster loading
|
| 84 |
+
- Optimized performance on AWS Neuron devices
|
| 85 |
+
- Same model capabilities with improved inference speed
|