astachowicz
commited on
Commit
•
f0d6983
1
Parent(s):
145f34d
Update README.md
Browse files
README.md
CHANGED
@@ -25,22 +25,27 @@ It is strongly recommended to train this model doing bf16 mixed-precision traini
|
|
25 |
|
26 |
[Here](https://github.com/huggingface/optimum-habana/blob/main/examples/question-answering/run_qa.py) is a question-answering example script to fine-tune a model on SQuAD. You can run it with DistilBERT with the following command:
|
27 |
```bash
|
28 |
-
python run_qa.py \
|
29 |
--model_name_or_path distilbert-base-uncased \
|
30 |
--gaudi_config_name Habana/distilbert-base-uncased \
|
31 |
--dataset_name squad \
|
32 |
--do_train \
|
33 |
--do_eval \
|
34 |
-
--per_device_train_batch_size
|
35 |
--per_device_eval_batch_size 8 \
|
36 |
-
--learning_rate
|
37 |
--num_train_epochs 3 \
|
38 |
--max_seq_length 384 \
|
39 |
--output_dir /tmp/squad/ \
|
40 |
--use_habana \
|
41 |
-
--use_lazy_mode \
|
42 |
-
--
|
43 |
-
--
|
|
|
|
|
|
|
|
|
|
|
44 |
```
|
45 |
|
46 |
Check the [documentation](https://huggingface.co/docs/optimum/habana/index) out for more advanced usage and examples.
|
|
|
25 |
|
26 |
[Here](https://github.com/huggingface/optimum-habana/blob/main/examples/question-answering/run_qa.py) is a question-answering example script to fine-tune a model on SQuAD. You can run it with DistilBERT with the following command:
|
27 |
```bash
|
28 |
+
PT_HPU_LAZY_MODE=0 PT_ENABLE_INT64_SUPPORT=1 python run_qa.py \
|
29 |
--model_name_or_path distilbert-base-uncased \
|
30 |
--gaudi_config_name Habana/distilbert-base-uncased \
|
31 |
--dataset_name squad \
|
32 |
--do_train \
|
33 |
--do_eval \
|
34 |
+
--per_device_train_batch_size 64 \
|
35 |
--per_device_eval_batch_size 8 \
|
36 |
+
--learning_rate 0.0005 \
|
37 |
--num_train_epochs 3 \
|
38 |
--max_seq_length 384 \
|
39 |
--output_dir /tmp/squad/ \
|
40 |
--use_habana \
|
41 |
+
--use_lazy_mode false \
|
42 |
+
--torch_compile \
|
43 |
+
--torch_compile_backend hpu_backend \
|
44 |
+
--throughput_warmup_steps 5\
|
45 |
+
--bf16 \
|
46 |
+
--save_strategy no \
|
47 |
+
--doc_stride 128 \
|
48 |
+
--overwrite_output_dir \
|
49 |
```
|
50 |
|
51 |
Check the [documentation](https://huggingface.co/docs/optimum/habana/index) out for more advanced usage and examples.
|