This makes it easier to exclude multi-line messages using single line
grepping. If your screen is wide enough this should not be a big
problem.
Example of what it looks like:
```
[2023-07-10 20:11:30,529] torch._dynamo.convert_frame.__guards: [DEBUG] GUARDS:
[2023-07-10 20:11:30,529] torch._dynamo.convert_frame.__guards: [DEBUG] hasattr(L['x'], '_dynamo_dynamic_indices') == False
[2023-07-10 20:11:30,529] torch._dynamo.convert_frame.__guards: [DEBUG] ___is_grad_enabled()
[2023-07-10 20:11:30,529] torch._dynamo.convert_frame.__guards: [DEBUG] not ___are_deterministic_algorithms_enabled()
[2023-07-10 20:11:30,529] torch._dynamo.convert_frame.__guards: [DEBUG] utils_device.CURRENT_DEVICE == None
```
Signed-off-by: Edward Z. Yang <ezyang@meta.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/104932
Approved by: https://github.com/mlazos, https://github.com/albanD
Prototype for the feature request:
>When working on a codebase that is unfamiliar to you, it can be helpful to single step through all of the code to see what is getting executed, what conditional branches are taken, and where indirect function jumps go. Model x-ray uses dynamo to give you a single step log of every source code line that does something relevant (i.e., a Tensor operation)
Dynamo logs to the ~`starts_line`~ `trace_source` logging artifact at the start of tracing new bytecode with a new line. It logs the line of source code associated with that bytecode.
~~Dynamo logs to the `graph_source` logging when a FX GraphModule is constructed. For each node in the graph, it logs the location of the original source code associated with that node.~~
Development notes: https://docs.google.com/document/d/1LjFeHzCgDDt535QUq5HydcQs56d7jWl5RvW8TLZN19g/edit?usp=sharing
Since the draft, we removed the `graph_source` logging artifact since printing the code of `GraphModule`s already displays the original source.
Sample:
```python
import torch
from functorch.experimental.control_flow import cond
def true_fn(x):
return x * 2
def false_fn(x):
return x * 3
def f_cond(pred, x):
return cond(pred, true_fn, false_fn, [x])
def f_outer(pred, x):
y = f_cond(pred, x)
if x.sum() > 0:
x = x * 2
else:
x = x * 3
return x, y
opt_f_cond = torch.compile(f_outer, backend="eager")
opt_f_cond(torch.tensor(True), torch.randn(3, 3))
```
Logs:
```shell
$ TORCH_LOGS="trace_source" python playground8.py
TRACE starts_line f_outer playground8.py:54
def f_outer(pred, x):
TRACE starts_line f_outer playground8.py:55
y = f_cond(pred, x)
TRACE starts_line f_cond playground8.py:51 (inline depth: 1)
def f_cond(pred, x):
TRACE starts_line f_cond playground8.py:52 (inline depth: 1)
return cond(pred, true_fn, false_fn, [x])
TRACE starts_line true_fn playground8.py:45 (inline depth: 2)
def true_fn(x):
TRACE starts_line true_fn playground8.py:46 (inline depth: 2)
return x * 2
TRACE starts_line false_fn playground8.py:48 (inline depth: 2)
def false_fn(x):
TRACE starts_line false_fn playground8.py:49 (inline depth: 2)
return x * 3
TRACE starts_line f_outer playground8.py:56
if x.sum() > 0:
TRACE starts_line <resume in f_outer> playground8.py:56
if x.sum() > 0:
TRACE starts_line <resume in f_outer> playground8.py:57
x = x * 2
TRACE starts_line <resume in f_outer> playground8.py:60
return x, y
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/104013
Approved by: https://github.com/ezyang
When investigating failures in https://github.com/pytorch/pytorch/pull/100017 I realized that we were reentering FakeTensorMode even though there was already one on the stack. Although we have attempted assert for these cases in the past, e.g., as in https://github.com/pytorch/pytorch/pull/97186 it seems that the existing protections were insufficient.
In this particular case, the reapplication of FakeTensorMode was due to an interaction with NotImplemented multiple dispatch handling. If proxy tensor mode detects an unrecognized tensor type (this includes FakeTensor, if it is not tracked with a proxy), it will return NotImplemented to give this tensor a chance to unpack itself into proxyable operation. However, this is never the right thing for FakeTensor, where no unpacking is possible. However, today, FakeTensor attempts to reapply the FakeTensorMode, resulting in FakeTensorMode being twice on the stack.
This PR does a number of things:
* It adds an assert in `FakeTensorMode.__torch_dispatch__` that you must not already have this mode on the stack, this is ALWAYS an error
* It modifies `FakeTensor.__torch_dispatch__` to return `NotImplemented` if the mode is already active. This prevents us from readding the mode on the stack
* It adds a new logging artifact `not_implemented` which you can use to get debug logs about all of the times a `__torch_dispatch__` handler returned NotImplemented and why it did so. Your subclass has to manually opt into this logging, but I inserted the necessary logs for ProxyTensorMode and FakeTensor(Mode)
* `with fake_mode` now no-ops if the fake mode is already on the stack, which is what users want anyway
* I am BREAKING pre-autograd tracing, because it is currently doing something weird with the original C++ mode stack. Brian is going to follow up with a fix next week.
Signed-off-by: Edward Z. Yang <ezyang@meta.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/102091
Approved by: https://github.com/thiagocrepaldi, https://github.com/eellison, https://github.com/wanchaol, https://github.com/bdhirsh
Removes a check which would sometimes allow `off_by_default` artifacts to be logged if logged at a higher level.
This change will only allow artifact messages to be displayed if the artifact is enabled, regardless of level.
closes#99144
Pull Request resolved: https://github.com/pytorch/pytorch/pull/99260
Approved by: https://github.com/lezcano
Summary:
Adds NNC-like logging that is configured through an env var `TORCH_COMPILE_LOGS`
Examples:
`TORCH_LOGS="dynamo,guards" python script.py` - prints dynamo logs at level INFO with guards of all functions that are compiled
`TORCH_LOGS="+dynamo,guards,graph" python script.py` - prints dynamo logs at level DEBUG with guards and graphs (in tabular) format of all graphs that are compiled
[More examples with full output](https://gist.github.com/mlazos/b17f474457308ce15e88c91721ac1cce)
Implementation:
The implementation parses the log settings from the environment, finds any components (aot, dynamo, inductor) or other loggable objects (guards, graph, etc.) and generates a log_state object. This object contains all of the enabled artifacts, and a qualified log name -> level mapping. _init_logs then adds handlers to the highest level logs (the registered logs), and sets any artifact loggers to level DEBUG if the artifact is enabled.
Note: set_logs is an alternative for manipulating the log_state, but if the environment contains TORCH_LOGS, the environment settings will be prioritized.
Adding a new log:
To add a new log, a dev should add their log name to torch._logging._registrations (there are examples there already).
Adding a new artifact:
To add a new artifact, a dev should add their artifact name to torch._logging._registrations as well.
Additionally, wherever the artifact is logged, `torch._logging.getArtifactLogger(__name__, <artifact_name>)` should be used instead of the standard logging implementation.
[design doc](https://docs.google.com/document/d/1ZRfTWKa8eaPq1AxaiHrq4ASTPouzzlPiuquSBEJYwS8/edit#)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/94858
Approved by: https://github.com/ezyang
