Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,41 @@
|
|
1 |
---
|
2 |
license: apache-2.0
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: apache-2.0
|
3 |
---
|
4 |
+
|
5 |
+
[Optimum Habana](https://github.com/huggingface/optimum-habana) is the interface between the Transformers library and Habana's Gaudi processor (HPU). It provides a set of tools enabling easy and fast model loading and fine-tuning on single- and multi-HPU settings for different downstream tasks.
|
6 |
+
Learn more about how to take advantage of the power of Habana HPUs to train Transformers models at [hf.co/Habana](https://huggingface.co/Habana).
|
7 |
+
|
8 |
+
|
9 |
+
# RoBERTa Base model HPU configuration
|
10 |
+
|
11 |
+
This model contains just the `GaudiConfig` file for running the [roberta-base](https://huggingface.co/roberta-base) model on Habana's Gaudi processors (HPU).
|
12 |
+
|
13 |
+
**This model contains no model weights, only a GaudiConfig.**
|
14 |
+
|
15 |
+
This enables to specify:
|
16 |
+
- `use_habana_mixed_precision`: whether to use Habana Mixed Precision (HMP)
|
17 |
+
- `hmp_opt_level`: optimization level for HMP, see [here](https://docs.habana.ai/en/latest/PyTorch/PyTorch_User_Guide/PT_Mixed_Precision.html#configuration-options) for a detailed explanation
|
18 |
+
- `hmp_bf16_ops`: list of operators that should run in bf16
|
19 |
+
- `hmp_fp32_ops`: list of operators that should run in fp32
|
20 |
+
- `hmp_is_verbose`: verbosity
|
21 |
+
- `use_fused_adam`: whether to use Habana's custom AdamW implementation
|
22 |
+
- `use_fused_clip_norm`: whether to use Habana's fused gradient norm clipping operator
|
23 |
+
|
24 |
+
|
25 |
+
## Usage
|
26 |
+
|
27 |
+
The model is instantiated the same way as in the Transformers library.
|
28 |
+
The only difference is that a Gaudi configuration associated to this model has to be loaded and provided to the trainer.
|
29 |
+
|
30 |
+
```
|
31 |
+
from transformers import RobertaTokenizer, RobertaModel
|
32 |
+
from optimum.habana import GaudiTrainer, GaudiTrainingArguments, GaudiConfig
|
33 |
+
|
34 |
+
tokenizer = RobertaTokenizer.from_pretrained('roberta-base')
|
35 |
+
model = RobertaModel.from_pretrained('roberta-base')
|
36 |
+
gaudi_config = GaudiConfig.from_pretrained("Habana/roberta-base")
|
37 |
+
args = GaudiTrainingArguments(output_dir=path_to_my_output_dir, use_habana=True, use_lazy_mode=True)
|
38 |
+
|
39 |
+
trainer = GaudiTrainer(model=model, gaudi_config=gaudi_config, args=args, tokenizer=tokenizer)
|
40 |
+
trainer.train()
|
41 |
+
```
|