How to interpret logit score from Hugging face binary classification model and convert it to probability sore

I am downloading the model microsoft/Multilingual-MiniLM-L12-H384 at main microsoft/Multilingual-MiniLM-L12-H384 and then using it. I am loading model using BertForSequenceClassification

Transformer Version: ‘4.11.3’

I have written the below code:

def compute_metrics(eval_pred):
    logits, labels = eval_pred

    predictions = np.argmax(logits, axis=-1)
    acc = np.sum(predictions == labels) / predictions.shape[0]
    return {"accuracy" : acc}

model = tr.BertForSequenceClassification.from_pretrained("/home/pc/minilm_model",num_labels=2)


training_args = tr.TrainingArguments(
    output_dir='/home/pc/proj/results2',          # output directory
    num_train_epochs=10,              # total number of training epochs
    per_device_train_batch_size=16,  # batch size per device during training
    per_device_eval_batch_size=32,   # batch size for evaluation
    warmup_steps=1000,                # number of warmup steps for learning rate scheduler
    weight_decay=0.01,               # strength of weight decay
    logging_dir='./logs',            # directory for storing logs

trainer = tr.Trainer(
    model=model,                         # the instantiated 🤗 Transformers model to be trained
    args=training_args,                  # training arguments, defined above
    train_dataset=train_data,         # training dataset
    eval_dataset=val_data,             # evaluation dataset

The folder is empty after I train the model.

Is it okay to pass classes=2 for binary classification?

The model last layer is simple linear connection which gives logits value. How to get its interpretation and probability score out of it? Does logit score is directly proportional to probability.?

model = tr.BertForSequenceClassification.from_pretrained("/home/pchhapolika/minilm_model",num_labels=2)