(beta) Using TORCH_LOGS python API with torch.compile

Created On: Jan 24, 2024 | Last Updated: Jan 31, 2024 | Last Verified: Nov 05, 2024

Author: Michael Lazos

import logging

This tutorial introduces the TORCH_LOGS environment variable, as well as the Python API, and demonstrates how to apply it to observe the phases of torch.compile.

Note

This tutorial requires PyTorch 2.2.0 or later.

Setup

In this example, we’ll set up a simple Python function which performs an elementwise add and observe the compilation process with TORCH_LOGS Python API.

Note

There is also an environment variable TORCH_LOGS, which can be used to change logging settings at the command line. The equivalent environment variable setting is shown for each example.

import torch

# exit cleanly if we are on a device that doesn't support torch.compile
if torch.cuda.get_device_capability() < (7, 0):
    print("Skipping because torch.compile is not supported on this device.")
else:
    @torch.compile()
    def fn(x, y):
        z = x + y
        return z + 2


    inputs = (torch.ones(2, 2, device="cuda"), torch.zeros(2, 2, device="cuda"))


# print separator and reset dynamo
# between each example
    def separator(name):
        print(f"==================={name}=========================")
        torch._dynamo.reset()


    separator("Dynamo Tracing")
# View dynamo tracing
# TORCH_LOGS="+dynamo"
    torch._logging.set_logs(dynamo=logging.DEBUG)
    fn(*inputs)

    separator("Traced Graph")
# View traced graph
# TORCH_LOGS="graph"
    torch._logging.set_logs(graph=True)
    fn(*inputs)

    separator("Fusion Decisions")
# View fusion decisions
# TORCH_LOGS="fusion"
    torch._logging.set_logs(fusion=True)
    fn(*inputs)

    separator("Output Code")
# View output code generated by inductor
# TORCH_LOGS="output_code"
    torch._logging.set_logs(output_code=True)
    fn(*inputs)

    separator("")

Skipping because torch.compile is not supported on this device.

Conclusion

In this tutorial we introduced the TORCH_LOGS environment variable and python API by experimenting with a small number of the available logging options. To view descriptions of all available options, run any python script which imports torch and set TORCH_LOGS to “help”.

Alternatively, you can view the torch._logging documentation to see descriptions of all available logging options.

For more information on torch.compile, see the torch.compile tutorial.