# KeyError: 'loss' while training QnA

**URL:** <https://discuss.huggingface.co/t/keyerror-loss-while-training-qna/4111>\
**Category:** Beginners\
**Created:** [March 3, 2021, 7:46am UTC](https://discuss.huggingface.co/t/keyerror-loss-while-training-qna/4111 "2021-03-03T07:46:07Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![fenil](https://avatars.discourse-cdn.com/v4/letter/f/a698b9/32.png) [@fenil](https://discuss.huggingface.co/u/fenil)\
**Post date:** [March 3, 2021, 7:46am UTC](https://discuss.huggingface.co/t/keyerror-loss-while-training-qna/4111/1 "2021-03-03T07:46:07Z")

</div>

I was finetuning `BertForQuestionAnswering` on nlp squad dateset with the following arguments

```auto
training_args = TrainingArguments(
    "test-qa-squad",
    learning_rate=2e-5,
    weight_decay=0.01,
    label_names = ["start_positions", "end_positions"],
    num_train_epochs=5,
    load_best_model_at_end=True,
    evaluation_strategy='epoch'
)

```

```auto
trainer = Trainer(
    model=model,
    args=training_args,
    train_dataset=train_dl,
    eval_dataset=train_dl
)

```

Then doing `trainer.train()` trains for some batches but then after a specific batch throws this error (one epoch isn’t complete yet)

```
KeyError Traceback (most recent call last)

<ipython-input-19-3435b262f1ae> in <module>()
----> 1 trainer.train()

3 frames

/usr/local/lib/python3.7/dist-packages/transformers/file_utils.py in __getitem__ (self, k)
   1444 if isinstance(k, str):
   1445 inner_dict = {k: v for (k, v) in self.items()}
-> 1446 return inner_dict[k]
   1447 else:
   1448 return self.to_tuple()[k]

KeyError: 'loss'

```

Is this some issue in the dataset? Any help is much appreciated

---

<div class="post-metadata">

**Author:** ![sgugger](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/sgugger/32/2291_2.png) [@sgugger](https://discuss.huggingface.co/u/sgugger)\
**Post date:** [March 3, 2021, 1:54pm UTC](https://discuss.huggingface.co/t/keyerror-loss-while-training-qna/4111/2 "2021-03-03T13:54:38Z")

</div>

You should double check your datasets has items that are dictionaries with the keys `"start_positions", "end_positions"` (that may be why the model is not returning the loss).

Also, you seem to be passing dataloaders to the `Trainer`? It takes datasets.

Lastly, for easy debug you can do the following:

```auto
for batch in trainer.get_train_dataloader():
    break
batch = {k: v.cuda() for k, v in batch.items()}
outputs = trainer.model(**batch)

```

to easily inspect what’s in your `batch` and your `outputs`.

---

<div class="post-metadata">

**Author:** ![mjc00](https://avatars.discourse-cdn.com/v4/letter/m/58956e/32.png) [@mjc00](https://discuss.huggingface.co/u/mjc00)\
**Post date:** [March 17, 2022, 7:30pm UTC](https://discuss.huggingface.co/t/keyerror-loss-while-training-qna/4111/3 "2022-03-17T19:30:03Z")

</div>

@sgugger I am running into a similar problem `KeyError: 'loss'` my dataset does have the items as dictionaries (see image)

 ![Screen Shot 2022-03-17 at 12.22.14 PM](https://us1.discourse-cdn.com/hellohellohello/original/2X/3/36b15a072e84fc092a35f09c5b5479b25e07f35c.png)

and my code is as follows:

```auto
from transformers import Trainer, TrainingArguments
batch_size = 64
logging_steps = len(dataset["train"]) // batch_size
model_name = f"{model_ckpt}-finetuned-test"
training_args = TrainingArguments(output_dir=model_name,
                                  num_train_epochs=2,
                                  learning_rate=2e-5,
                                  per_device_train_batch_size=batch_size,
                                  per_device_eval_batch_size=batch_size,
                                  weight_decay=0.01,
                                  evaluation_strategy="epoch",
                                  disable_tqdm=False,
                                  logging_steps=logging_steps,
                                  label_names = ['CategoryCode'],
                                  #push_to_hub=True, 
                                  log_level="error")

trainer = Trainer(model=model, 
                  args=training_args, 
                  compute_metrics=compute_metrics,
                  train_dataset=dataset["train"],
                  eval_dataset=dataset["vald"],
                  tokenizer=tokenizer)
trainer.train();

```

Note: I am running the above mentioned code locally Mac M1.