# Problems with Dataset.from\_dict() and Feature types

**URL:** <https://discuss.huggingface.co/t/problems-with-dataset-from-dict-and-feature-types/9630>\
**Category:** 🤗Datasets\
**Created:** [August 31, 2021, 11:27am UTC](https://discuss.huggingface.co/t/problems-with-dataset-from-dict-and-feature-types/9630 "2021-08-31T11:27:51Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![zhcr](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/zhcr/32/2397_2.png) [@zhcr](https://discuss.huggingface.co/u/zhcr)\
**Post date:** [August 31, 2021, 11:27am UTC](https://discuss.huggingface.co/t/problems-with-dataset-from-dict-and-feature-types/9630/1 "2021-08-31T11:27:51Z")

</div>

I am building the training pipeline for a Distilbert and am trying to define the Feature types for a Dataset that is loaded from a dictionary.

This dictionary is actually the input\_ids, labels and attention\_mask fields that the tokenizer returns, and I can’t seem to achieve the correct data assignation, specially the label feature. Here is an example:

```auto
data = {
    'input_ids': [[101, 1, 2, 3, 102], [101, 11, 12, 13, 102]],
    'labels': [[-100, 'O', 'B-NOUN', 'L-NOUN', -100], [-100, 'U-NOUN', 'O', 'O', -100]],
    'attention_mask': [[0, 1, 1], [1, 0, 0]]
}

dt = Dataset.from_dict(data)

```

The problem comes because:

1. They are Sequences of values, and I am not sure how to build a Sequence[ClassLabel] feature and hence, the above code crashes on .from\_dict(data) as it expect int not str.
2. If I map the label names previously to the .from\_dict(data) (such as the code after those lines) it is able to generate the dataset without problem, but later on creates a strange behavior in the training and prediction time for the model. Here Distilbert just builds its own tag2id mapping such as {0: ‘LABEL\_0’, 1: ‘LABEL\_1’, 2:‘LABEL\_2’} that I am not sure how to fix (and this generates problems whith the aggregation strategy on prediction time.

Example of mapped data such as explained in point 2:

```auto
data = {
    'input_ids': [[101, 1, 2, 3, 102], [101, 11, 12, 13, 102]],
    'labels': [[-100, 1, 2, 3, -100], [-100, 4, 1, 1, -100]],
    'attention_mask': [[0, 1, 1], [1, 0, 0]]
}

dt = Dataset.from_dict(data)

```

Any idea how to correctly generate the dataset and avoid that problem with Distilbert?

UPDATE: Decided to try another road and it is basically building first the dataset and then tokenizing, but the problem I got right now is that using the function align\_labels\_with\_mapping crashes when the feature is a Sequence, and casting the feature also was not able to cast the values inside the Sequence object.

What am I doing wrong?

---

<div class="post-metadata">

**Author:** ![lhoestq](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/lhoestq/32/52888_2.png) [@lhoestq](https://discuss.huggingface.co/u/lhoestq)\
**Post date:** [September 6, 2021, 1:21pm UTC](https://discuss.huggingface.co/t/problems-with-dataset-from-dict-and-feature-types/9630/2 "2021-09-06T13:21:19Z")

</div>

Hi ! You can define a Sequence of ClassLabel with

```python
Sequence(ClassLabel(names=...))

```

but in this case you need your data to be integers, not strings.  
You can convert your data to integers with

```python
data = features.encode_batch(data)

```

and then get your dataset:

```auto
dt = Dataset.from_dict(data, features=features)

```

Regarding the label2id of DistilBERT, you can define your own by specifying the `id2label` parameter of the model config when you load your model with `.from_pretrained(..., id2label=...)`

> [@zhcr](#):
>
> the function align\_labels\_with\_mapping crashes when the feature is a Sequence, and casting the feature also was not able to cast the values inside the Sequence object.

Could you open an issue on GitHub so that we can take a look into what went wrong ?  
If you have a code snippet that reproduces the error that would be of great help as well.