Modify the "Usage" section of the model card

The code snippet in the "Usage" section of the model card is not runnable per se because it requires a dataset and pre/post-processing. We now points toward the QA example of optimum-habana and give the command to run this example with ALBERT Large.

README.md

@@ -23,25 +23,26 @@ This enables to specify:
 ## Usage
 
 The model is instantiated the same way as in the Transformers library.
-The only difference is that there are a few new training arguments specific to HPUs:
-
-```
-from optimum.habana import GaudiTrainer, GaudiTrainingArguments
-from transformers import AlbertTokenizer, AlbertModel
-
-tokenizer = AlbertTokenizer.from_pretrained("albert-large-v2")
-model = AlbertModel.from_pretrained("albert-large-v2")
-args = GaudiTrainingArguments(
-    output_dir="/tmp/output_dir",
-    use_habana=True,
-    use_lazy_mode=True,
-    gaudi_config_name="Habana/albert-large-v2",
-)
-
-trainer = GaudiTrainer(
-    model=model,
-    args=args,
-    tokenizer=tokenizer,
-)
-trainer.train()
+The only difference is that there are a few new training arguments specific to HPUs.
+
+[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 ALBERT Large with the following command:
+```bash
+python run_qa.py \
+  --model_name_or_path albert-large-v2 \
+  --gaudi_config_name Habana/albert-large-v2 \
+  --dataset_name squad \
+  --do_train \
+  --do_eval \
+  --per_device_train_batch_size 32 \
+  --per_device_eval_batch_size 4 \
+  --learning_rate 5e-5 \
+  --num_train_epochs 2 \
+  --max_seq_length 384 \
+  --doc_stride 128 \
+  --output_dir /tmp/squad/ \
+  --use_habana \
+  --use_lazy_mode \
+  --throughput_warmup_steps 2
 ```
+
+Check the [documentation](https://huggingface.co/docs/optimum/habana_index) out for more advanced usage and examples.