Summarize scalars and tensors in PyTorch#

It is often useful to track various values of interest during training. This section introduces APIs for summarizing various tensors in a PyTorch model and inspecting them during or after a run.

Scalar Summaries#


It is often useful to visualize various scalar values during training. This may include scalar values such as learning rate, gradient norms, etc. For this, we provide the scalar_summary API which allows to summarize scalar model tensors. These summaries are written to Tensorboard events files and can be visualized using Tensorboard.

How to Use Scalar Summaries#

The scalar summary API can be imported as follows:

>> from modelzoo.common.pytorch.summaries import scalar_summary

In order to summarize a scalar tensor S, add the following statement to the model definition code:

scalar_summary("my_scalar_tensor", S)

During training, the value of S will be periodically written to the Tensorboard events file and can be visualized in TensorBoard.


Some common scalar tensors that are generally desirable to track during training are already implemented in Cerebras Model Zoo. They are disabled by default, but to enable them, set log_summaries: True in the runconfig section of the params file passed to a Cerebras Model Zoo run.

Tensor Summaries#


In the section above, we described how to summarize scalar values which can be visualized in TensorBoard. However, there are cases where it is desirable to summarize arbitrary tensor shapes. Since TensorBoard only supports scalar summaries, we provide a separate API, which is very similar to scalar_summary API, but for summarizing tensors of arbitrary shapes.

How to Use Tensor Summaries#

The tensor summary API can be imported as follows:

>> from modelzoo.common.pytorch.summaries import tensor_summary

In order to summarize a tensor T, add the following statement to the model definition code:

>> tensor_summary("my_custom_name", T)

Under the hood, we mark the provided tensor as an output of the graph and fetch its value at every log step (similar to losses and other scalar summaries). This value is then written out to a file and can be later retrieved through the TensorSummaryReader API (see below).

Here’s a simple example where we’d like to summarize the input features and last layer’s logits of a fully connected network:

from modelzoo.common.pytorch.summaries import tensor_summary

class FC(nn.Module):
    def forward(self, features):
        tensor_summary("features", features)
        logits = self.fc_layer(features)
        tensor_summary("last_layer_logits", logits)
        return logits

In order to retrieve the saved values of these tensors during or after a run, use the TensorSummaryReader API which supports listing all available tensor names and fetching a tensor by name for a given step. TensorSummaryReader object takes as input a single argument denoting the path to a Tensorboard events file or a directory containing Tensorboard events files. Location of tensor summaries are inferred from these events files as there is a one-to-one mapping from Tensorboard events files and tensor summary directories.

In the example above, we added summaries for features and last_layer_logits. We can then use the TensorSummaryReader API to load the summarized values of these tensors at a given step:

>>> from modelzoo.common.pytorch.summaries import TensorSummaryReader
>>> reader = TensorSummaryReader("model_dir/train")
>>> reader.names()   # Grab all tensor summary names
['features', 'last_layer_logits']
>>> reader.load("features", 2)   # Load tensor "features" from step 2
TensorDescriptor(step=2, tensor=tensor([[2, 4],
        [6, 8]]), utctime='2023-02-07T05:45:29.017264')
>>> reader.load("non_existing", 100)   # Load a non-existing tensor
WARNING:root:No tensor with name non_existing has been summarized at step 100

TensorSummaryReader.load() returns one or more TensorDescriptor objects. TensorDescriptor is a POD structure which holds:

  • step: The step at which this tensor was summarized.

  • utctime: The UTC time at which the value was saved.

  • tensor: The summarized value.


  • Currently, only tensors smaller than 2GB can be summarized. Note that this applies to each individual tensor. So while summarizing a tensor of size 4GB is not supported, summarizing two or more tensors each with a size of 2GB is supported.

  • Adding tensor summaries may change how the graph is lowered and can create a different compile. This is because marking a tensor as an output may prevent it from being pruned out in certain operation fusions. From an overall computation standpoint, however, the graphs should be identical. The only difference is how the computation is represented.