SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

update QuickTour docs to reflect model output object (#11462)

Browse Source 
* update docs to reflect model output object

* run make style`
This commit is contained in:
Hamel Husain
2021-04-26 19:18:37 -07:00
committed by GitHub
parent 741d48f5c7
commit 88ac60f7b5
2 changed files with 12 additions and 13 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
docs/source/main_classes/output.rst
View File

@@ -13,8 +13,8 @@
Model outputs Model outputs
----------------------------------------------------------------------------------------------------------------------- -----------------------------------------------------------------------------------------------------------------------
PyTorch models have outputs that are instances of subclasses of :class:`~transformers.file_utils.ModelOutput`. Those All models have outputs that are instances of subclasses of :class:`~transformers.file_utils.ModelOutput`. Those are
are data structures containing all the information returned by the model, but that can also be used as tuples or data structures containing all the information returned by the model, but that can also be used as tuples or
dictionaries. dictionaries.
Let's see of this looks on an example: Let's see of this looks on an example:

21
docs/source/quicktour.rst
View File

@@ -238,23 +238,22 @@ keys directly to tensors, for a PyTorch model, you need to unpack the dictionary
>>> ## TENSORFLOW CODE >>> ## TENSORFLOW CODE
>>> tf_outputs = tf_model(tf_batch) >>> tf_outputs = tf_model(tf_batch)
In 🤗 Transformers, all outputs are tuples (with only one element potentially). Here, we get a tuple with just the final In 🤗 Transformers, all outputs are objects that contain the model's final activations along with other metadata. These
activations of the model. objects are described in greater detail :doc:`here <main_classes/output>`. For now, let's inspect the output ourselves:
.. code-block:: .. code-block::
>>> ## PYTORCH CODE >>> ## PYTORCH CODE
>>> print(pt_outputs) >>> print(pt_outputs)
(tensor([[-4.0833, 4.3364], SequenceClassifierOutput(loss=None, logits=tensor([[-4.0833, 4.3364],
[ 0.0818, -0.0418]], grad_fn=<AddmmBackward>),) [ 0.0818, -0.0418]], grad_fn=<AddmmBackward>), hidden_states=None, attentions=None)
>>> ## TENSORFLOW CODE >>> ## TENSORFLOW CODE
>>> print(tf_outputs) >>> print(tf_outputs)
(<tf.Tensor: shape=(2, 2), dtype=float32, numpy= TFSequenceClassifierOutput(loss=None, logits=<tf.Tensor: shape=(2, 2), dtype=float32, numpy=
array([[-4.0832963 , 4.336414 ], array([[-4.0832963 , 4.3364143 ],
[ 0.08181786, -0.04179301]], dtype=float32)>,) [ 0.081807 , -0.04178282]], dtype=float32)>, hidden_states=None, attentions=None)
The model can return more than just the final activations, which is why the output is a tuple. Here we only asked for Notice how the output object has a ``logits`` attribute. You can use this to access the model's final activations.
the final activations, so we get a tuple with one element.
.. note:: .. note::
@@ -267,10 +266,10 @@ Let's apply the SoftMax activation to get predictions.
>>> ## PYTORCH CODE >>> ## PYTORCH CODE
>>> import torch.nn.functional as F >>> import torch.nn.functional as F
>>> pt_predictions = F.softmax(pt_outputs[0], dim=-1) >>> pt_predictions = F.softmax(pt_outputs.logits, dim=-1)
>>> ## TENSORFLOW CODE >>> ## TENSORFLOW CODE
>>> import tensorflow as tf >>> import tensorflow as tf
>>> tf_predictions = tf.nn.softmax(tf_outputs[0], axis=-1) >>> tf.nn.softmax(tf_outputs.logits, axis=-1)
We can see we get the numbers from before: We can see we get the numbers from before: