jwieczorekhabana
commited on
Commit
•
4542ff3
1
Parent(s):
219eee5
Update README.md
Browse filesUpdate command and gaudi_config.json description.
README.md
CHANGED
@@ -15,6 +15,9 @@ This model only contains the `GaudiConfig` file for running the [Wav2Vec2](https
|
|
15 |
This enables to specify:
|
16 |
- `use_fused_adam`: whether to use Habana's custom AdamW implementation
|
17 |
- `use_fused_clip_norm`: whether to use Habana's fused gradient norm clipping operator
|
|
|
|
|
|
|
18 |
|
19 |
## Usage
|
20 |
|
@@ -44,7 +47,8 @@ python run_audio_classification.py \
|
|
44 |
--use_habana \
|
45 |
--use_lazy_mode \
|
46 |
--gaudi_config_name Habana/wav2vec2 \
|
47 |
-
--throughput_warmup_steps 2
|
|
|
48 |
```
|
49 |
|
50 |
Check the [documentation](https://huggingface.co/docs/optimum/habana/index) out for more advanced usage and examples.
|
|
|
15 |
This enables to specify:
|
16 |
- `use_fused_adam`: whether to use Habana's custom AdamW implementation
|
17 |
- `use_fused_clip_norm`: whether to use Habana's fused gradient norm clipping operator
|
18 |
+
- `disable_autocast`: whether to disable autocast; this parameter takes precedence over --bf16 flag and is temporary as some scripts produce nan values.
|
19 |
+
In those cases this parameter is already present in huggingface topology Habana gaudi_config.json.
|
20 |
+
|
21 |
|
22 |
## Usage
|
23 |
|
|
|
47 |
--use_habana \
|
48 |
--use_lazy_mode \
|
49 |
--gaudi_config_name Habana/wav2vec2 \
|
50 |
+
--throughput_warmup_steps 2 \
|
51 |
+
--bf16
|
52 |
```
|
53 |
|
54 |
Check the [documentation](https://huggingface.co/docs/optimum/habana/index) out for more advanced usage and examples.
|