This PR has a lot of "draw the rest of the fucking owl" energy. Here's how to break it down.
1. **torch/_inductor/graph.py** - We start by tightening unbacked symbol invariants. Specifically, as we lower FX nodes, we check whether or not every unbacked_binding recorded on the FX node meta, actually ends up getting bound (according to get_unbacked_symbol_defs) in all the buffers generated by the lowering. Hopefully this invariant is self evident. This leads to a lot of failures.
2. **torch/_inductor/ir.py** - Problem 1: There is softness in how Inductor computes defs of unbacked symbols in IR node. Previously, we tried to infer it by looking at the output sizes/strides/etc and see if new unbacked symbols popped up that we hadn't seen in the inputs. I don't know exactly what was buggy about the old code, but sometimes we would fail to notice an unbacked symbol had been bound, or rebind an unbacked symbol multiple times. Fortunately, thanks to the earlier PRs in our stack, we now have a nice list of unbacked symbol bindings from FX, so we now just store it directly on ExternKernel and use it directly to report defs. This has to be done twice: once for FallbackKernel (e.g., nonzero) and once for DynamicScalar (e.g., item) (see also **torch/_inductor/lowering.py**, **torch/_inductor/codegen/wrapper.py** and **torch/_inductor/codegen/cpp_wrapper_cpu.py** for the lowering and codegen changes for item)
* **process_kernel** - Sidequest! It turns out that Inductor lowering can reallocate unbacked symbols. This happens specifically when we repropagate fake tensors through the operator in `process_kernel`. This repropagation process is necessary because Inductor may have changed the strides of input tensors, and it must now recompute the strides so that it can continue to appropriately plan the rest of the lowering process. This is fine: we just make sure we do the rebind unbacked + compute_unbacked_bindings dance we've been doing previously in the PR stack. But instead of putting unbacked_bindings on a new FX node, they go straight into our unbacked_bindings on the Inductor IR node.
* **codegen_unbacked_symbol_defs** - Sidequest! FallbackKernel lowering is done in two steps. First, you emit the FallbackKernel buffer. Then, you emit MultiOutput buffers which actually give access to the individual outputs of FallbackKernel, which may have been multi-output. There is a design decision here: does the FallbackKernel bind the unbacked symbols, or the MultiOutput buffer? Historically, we put the binding on MultiOutput buffer, because it's more convenient: the FallbackKernel buffer is fake, in fact, it doesn't even get a name in C++ codegen. But it's kind of inconsistent with the keypath model that we've been tracking unbacked bindings with: if you have a multi-output node, you'd expect a keypath like `[0].size()[0]` representing the first output's first dimension size. That suggests that it's the FallbackKernel that should define the things. So that was my first implementation. Unfortunately, the C++ codegen is too cursed and I could not understand how to make it work in that case. So now we just unsoundly assume you cannot have multi-output data dependent output, and do the codegen in MultiOutput. There are some comments explaining exactly what we are improperly assuming.
3. **_rename_unbacked_to** in **torch/fx/experimental/symbolic_shapes.py** - Previously, when we renamed unbacked symbols, we clobbered any facts we previously knew about them. So for example, if we had a replacement `u0 -> s0` but then we renamed u0 to u1, we would now setup the replacement `u0 -> u1`, clobbering the old replacement. This apparently didn't matter in earlier PRs in the stack, but with Inductor now on the ball, there were some tests that indicated this was a problem. The solution is easy: if u0 had a preexisting replacement, reapply it to u1. However...
* **torch/_functorch/_aot_autograd/collect_metadata_analysis.py** - When we run forward analysis, this triggers fake tensor repropagation and fresh allocations. Previously, we just cleared out the pending symbols when finished the analysis. But with the change above, this would also migrate replacements to the new symbols... which are now dead. So now we explicitly suppress generation of these symbols with `ignore_fresh_unbacked_symbols` so that no rebinding happens at all.
* **torch/_dynamo/eval_frame.py** - same deal; I just searched for all sites we called clear() on pending
4. The last step is fixing the long tail of extra problems that show up, now that unbacked_bindings are load bearing into Inductor
* **torch/_dynamo/eval_frame.py** - Some of the exports are making copies of nodes without repropagating fake tensors, so in this case, it is important to also copy the `unbacked_bindings` (apparently this didn't matter before without the Inductor changes)
* **torch/_export/pass_base.py** - I discover that this is doing fake tensor repropagation via a test suite failure. Do the same playbook as AOTAutograd: PropagateUnbackedSymInts too! Actually, they also have implemented their own tracer as well, so do the same playbook as proxy_tensor: record unbacked_bindings on the newly traced nodes. UGH code duplication.
* **torch/_subclasses/fake_tensor.py**, **torch/_subclasses/fake_impls.py** (with call site updates at **torch/_functorch/_aot_autograd/traced_function_transforms.py** and **torch/fx/passes/fake_tensor_prop.py**) - What's this new epoch thing? I noticed that sometimes I would be retracing, call nonzero() on a fake tensor, and not allocate a new unbacked symbol. This is actually bad, because if I don't get a new unbacked symbol, I don't know there's a binding site, and `unbacked_bindings` is now missing a binding. The reason for this is memoization: if I reuse the exact same fake tensor on my retrace, it will already have an unbacked symint memoized on it and we will short circuit allocation. Well, that's no good. So I associate the memos with a fake tensor epoch, and every time you start a new fake tensor propagation from scratch, you bump the epoch so that I clear all the memos.
* **torch/_inductor/scheduler.py** - I notice in unit tests that V.current_node is not always set when we call process_kernel. So I save it into the IR node and restore it when we are running `get_estimated_runtime`.
* **torch/fx/experimental/symbolic_shapes.py** - A few things
* **rebind_unbacked** (re **_tensor_version**). Ordinarily, when you have an unbacked SymInt, you persistently hvae it all the way to the end of the program. `_tensor_version` violates this: this generates an unbacked SymInt (for reasons I don't quite understand?) and then gets rid of it later. This triggered an assert violation. I think this op is kind of misusing unbacked SymInt, but I didn't know how to refactor it, so it gets a special case.
* **rebind_unbacked** (re **Simplify SymBool binding**). Ugh, SymBool, what a pain in the butt. I have an assert that you can only rebind unbacked symbol to another unbacked symbol. This assert fails when a boolean is involved, because the result of running keypath on the result is not `u1`, it's `sympy.Piecewise(... sympy.Eq(u1, 1) ...)`. This is actually just `u1`, but Sympy doesn't know it because it doesn't know that `u1` value range is `[0, 1]`. So we manually implement the simplification needed to get the assert to pass.
* **compute_unbacked_bindings** (re **This is pretty fragile**). There is a really funny disaster involving memoization and Inductor process kernel. Ordinarily when I retrace, if there was a memo hit in the old trace, there will be a memo hit in the new trace. However, Inductor process kernel breaks this, because it recreates fake tensor inputs to the operator call from scratch (since they might have different strides), and obviously these tensor inputs don't have the memo from the old one. I tried a little bit to try to manually transplant the memo to the new fake tensor but it seemed hopeless, so I just let the fresh symbol ride, allocating a new unbacked symbol. However, in one of our tests, we rely on knowing that the first nonzero call is equal to the second (memoized) nonzero call. The equality test looked pretty easy to discharge, so I just went ahead and added a deferred runtime assert to this effect and it worked.
Signed-off-by: Edward Z. Yang <ezyang@meta.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/124394
Approved by: https://github.com/jansel
ghstack dependencies: #124310, #124314, #124316
This is a partial revert of https://github.com/pytorch/pytorch/pull/124059
Like in #124297, profiling has revealed that testing equality on *every* output is kind of expensive. So we only test equality when we know there is an unbacked binding. This is the same playbook as the previous PR, just on FakeTensorProp instead of PropagateUnbackedSymInts. Note that we also need to populate `unbacked_bindings` in proxy_tensor.py, since we're generating an entirely new graph in that case.
We now have enough propagation that we're able to trigger a bug related to divisibility replacement. In https://github.com/pytorch/pytorch/pull/113165 we allowed to replace `u0` with `u1 * c` for some constant c, when we have determined that u0 is divisible by c. However, where does the binding for u1 come from? What we will have in practice is that there is some node that is supposed to have bound u1, but which actually is getting a `u1 * c` in its output. So, to get u1, we must divide out c. Fortunately, under the divisibility condition, this is always possible (but remember, we must test divisibility at runtime!)
Because we have tightened up asserts, it is now an error to allocate unbacked SymInts and then fail to track them under unbacked_bindings. In torch/_dynamo/eval_frame.py and torch/_functorch/_aot_autograd/collect_metadata_analysis.py there are examples of benign cases where we repropagated fake tensors but then immediately threw away the results. In these cases, it's not appropriate to rebind, since we're still using the old FX graph that has all of the old symbols. So we just manually clear it. It is possible that other cases will need to be updated, so this PR is "risky" from the perspective of hitting fbcode.
Signed-off-by: Edward Z. Yang <ezyang@meta.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/124310
Approved by: https://github.com/lezcano
I'm going to setup some extra behavior when we set example value, so
I need a convenient place to interpose. I cannot easily do it on
meta itself because its a generic dict with no interposition point.
Signed-off-by: Edward Z. Yang <ezyang@meta.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/124176
Approved by: https://github.com/oulgen
ghstack dependencies: #124105, #124059
Fixes#114844
In the linked issue we have
```
compiled_module = torch.compile(module)
compiled_module.x = ...
compiled_module(...) # Mutates self.x
```
Where since the module mutates `self.x` you would expect `compiled_module.x`
to be updated but actually `compiled_module.x = ...` sets an attribute "x"
on the `OptimizedModule` object while the forward method of the module mutates
`module.x`.
This gives the expected behavior by forwarding `compiled_module.__setattr__`
down to `module.__setattr__`. There is already a corresponding `__getattr__`
so now `compiled_module.x` becomes an alias for `module.x`.
Co-authored-by: Edward Z. Yang <ezyang@meta.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/122098
Approved by: https://github.com/ezyang, https://github.com/lezcano
`dynamo.explain()` was updated to return a structure but the docs weren't updated to match.
- Update the docs to use the new API
- Remove some dead code left when `explain` was updated.
- Drive-by: Fix some `nopython` uses that I noticed
- Drive-by: I noticed an ignored error coming from CleanupHook on shutdown - make it check the global before setting it.
Fixes#122573
Pull Request resolved: https://github.com/pytorch/pytorch/pull/122745
Approved by: https://github.com/jansel
Fixes#114844
In the linked issue we have
```
compiled_module = torch.compile(module)
compiled_module.x = ...
compiled_module(...) # Mutates self.x
```
Where since the module mutates `self.x` you would expect `compiled_module.x`
to be updated but actually `compiled_module.x = ...` sets an attribute "x"
on the `OptimizedModule` object while the forward method of the module mutates
`module.x`.
This gives the expected behavior by forwarding `compiled_module.__setattr__`
down to `module.__setattr__`. There is already a corresponding `__getattr__`
so now `compiled_module.x` becomes an alias for `module.x`.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/122098
Approved by: https://github.com/ezyang, https://github.com/lezcano
Fixes#114844
In the linked issue we have
```
compiled_module = torch.compile(module)
compiled_module.x = ...
compiled_module(...) # Mutates self.x
```
Where since the module mutates `self.x` you would expect `compiled_module.x`
to be updated but actually `compiled_module.x = ...` sets an attribute "x"
on the `OptimizedModule` object while the forward method of the module mutates
`module.x`.
This gives the expected behavior by forwarding `compiled_module.__setattr__`
down to `module.__setattr__`. There is already a corresponding `__getattr__`
so now `compiled_module.x` becomes an alias for `module.x`.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/122098
Approved by: https://github.com/ezyang, https://github.com/lezcano
Summary:
When we convert `dynamic_shapes` to `constraints` and pass them to `_dynamo.export`, we shouldn't give a deprecation warning. Such conversion happens when calling `torch.export.export`, e.g. But it can also happen when calling `capture_pre_autograd_graph` (which itself has this deprecation warning when `constraints` are passed directly as well).
Since `_log_export_usage` is an indicator of a top-level call (it is `True` by default but set to `False`, or at least passed through, by callers), we can (ab)use it to indicate when to give this deprecation warning.
Test Plan: none
Differential Revision: D54350172
Pull Request resolved: https://github.com/pytorch/pytorch/pull/120896
Approved by: https://github.com/BoyuanFeng, https://github.com/zhxchen17
Summary:
as title.
The following APIs are logged:
- capture_preautograd_graph
- torch._export.aot_compile
- external usage of _export_to_torch_ir (AOTInductor, Pippy)
- constraints API
- public use of torch._dynamo.export
Test Plan: CI
Differential Revision: D53735599
Pull Request resolved: https://github.com/pytorch/pytorch/pull/119848
Approved by: https://github.com/suo
This is going to fix a legacy issue like:
```
torch._dynamo.export(torch.ops.aten.scaled_dot_product_attention, ...)(*inputs,)
```
This is not supported any more, now the top level ```torch.export``` only support ```nn.Module```, but there are still some tests using the internal APIs and caused the ```trace_rules.check``` assertion error. This PR is going to mitigate such cases.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/119528
Approved by: https://github.com/ydwu4
Attempt #2 for https://github.com/pytorch/pytorch/pull/117875 to fix https://github.com/pytorch/pytorch/issues/112090.
Summary of changes:
- ~Changed CacheEntry linked list into a doubly-linked list structure to support deletion.~ (done by C++ refactor)
- Added CacheEntry and ExtraState borrowed references to GuardFn so that GuardFn can tell ExtraState to delete CacheEntry when the GuardFn is invalidated.
- ~Added ExtraState raw reference to CacheEntry so that we can get ExtraState to correctly point to the first CacheEntry if it gets deleted.~ (done by C++ refactor)
- CacheEntry destructor needs to reset GuardFn refs to ExtraState/CacheEntry in order to prevent use-after-free.
- code_context values that are nn.GraphModules need to be weakrefs in order to prevent circular references.
- Added tests that check for memory leaks and cache deletion operations.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/119107
Approved by: https://github.com/jansel
With this change, we now have the invariant that eval_frame only
contains "hot" functions that are called at runtime, as opposed to
cold functions which are only called at compile time.
Signed-off-by: Edward Z. Yang <ezyang@meta.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/119253
Approved by: https://github.com/yanboliang
ghstack dependencies: #119251
Fixes https://github.com/pytorch/pytorch/issues/119238
Here's what it looks like now:
```
$ TORCH_LOGS=+torch._dynamo.convert_frame python a.py
[2024-02-05 18:52:07,248] [0/0] torch._dynamo.convert_frame: [DEBUG] torchdynamo start compiling f /data/users/ezyang/b/pytorch/a.py:3, stack (elided 5 frames):
[2024-02-05 18:52:07,248] [0/0] torch._dynamo.convert_frame: [DEBUG] File "/data/users/ezyang/b/pytorch/a.py", line 7, in <module>
[2024-02-05 18:52:07,248] [0/0] torch._dynamo.convert_frame: [DEBUG] f(torch.randn(2))
[2024-02-05 18:52:07,248] [0/0] torch._dynamo.convert_frame: [DEBUG] File "/data/users/ezyang/b/pytorch/torch/_dynamo/eval_frame.py", line 453, in _fn
[2024-02-05 18:52:07,248] [0/0] torch._dynamo.convert_frame: [DEBUG] return fn(*args, **kwargs)
[2024-02-05 18:52:07,248] [0/0] torch._dynamo.convert_frame: [DEBUG]
$ cat a.py
import torch
@torch.compile
def f(x):
return x * 2
f(torch.randn(2))
```
The eval_frame frame is intentionally present, since what happens is you run the torch.compile wrapper, and then you actually hit the user frame to be compiled.
Signed-off-by: Edward Z. Yang <ezyang@meta.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/119251
Approved by: https://github.com/yanboliang, https://github.com/mlazos
This PR fixes several bugs, listed in priority:
1. `load_state_dict` with a nontensor step was incorrect for capturable and fused implementations since we don't create the tensors on the right device in `__setstate__`. This has been fixed.
2. The most recently added capturable implementations forgot the check that all tensors should be on CUDA for eager. We've now added those checks
3. The most recent change in Adamax only adds capturable for foreach but will silently be incorrect for forloop/single-tensor. I've added erroring and modified testing with many many many skips for that. Honestly my preference after this PR has only been further cemented that we should just do the single tensor and multi tensor capturable implementations together in the future. @mlazos
4. The conditional for adding cuda-supported configs for the optimizer infos was incorrect! So we hadn't been testing capturable! This also stands rectified and was the trigger for this PR in the first place.
5. In a similar way, the conditional for `_get_optim_inputs_including_global_cliquey_kwargs` was incorrect sometimes as well. This has also been corrected.
The following is not a bug, but is just something to make life simpler by not needing to handle Nones: `optim_input_funcs` must now mandatorily take in a `device`, which could be a string or a torch.device.
Details for posterity:
4. Running the test_foreach_matches_forloop test and printing the configs that get printed yields capturable getting included, which is correct.
```
(pytorch-3.10) [janeyx@devgpu023.odn1 ~/local/pytorch (5d50138f)]$ python test/test_optim.py -k test_foreach_matches_forloop_AdamW_cuda
/home/janeyx/.conda/envs/pytorch-3.10/lib/python3.10/site-packages/transformers/utils/generic.py:441: UserWarning: torch.utils._pytree._register_pytree_node is deprecated. Please use torch.utils._pytree.register_pytree_node instead.
_torch_pytree._register_pytree_node(
/home/janeyx/.conda/envs/pytorch-3.10/lib/python3.10/site-packages/scipy/__init__.py:146: UserWarning: A NumPy version >=1.17.3 and <1.25.0 is required for this version of SciPy (detected version 1.26.0
warnings.warn(f"A NumPy version >={np_minversion} and <{np_maxversion}"
params=None, kwargs={}, desc=default
params=None, kwargs={'lr': 0.01}, desc=non-default lr
params=None, kwargs={'weight_decay': 0.1}, desc=nonzero weight_decay
params=None, kwargs={'weight_decay': 0.1, 'maximize': True}, desc=maximize
params=None, kwargs={'weight_decay': 0.1, 'amsgrad': True}, desc=amsgrad
params=None, kwargs={'capturable': True}, desc=capturable
params=None, kwargs={'weight_decay': 0.1, 'amsgrad': True, 'capturable': True}, desc=capturable, amsgrad
params=None, kwargs={'lr': tensor(0.0010), 'amsgrad': True, 'capturable': True}, desc=Tensor lr with capturable and amsgrad
.
----------------------------------------------------------------------
Ran 1 test in 19.229s
OK
```
5. Running the test_optimizer_can_be_printed test (which calls `_get_optim_inputs_including_global_cliquey_kwargs`) and printing what gets run is also now correct.
```
/home/janeyx/.conda/envs/pytorch-3.10/lib/python3.10/site-packages/scipy/__init__.py:146: UserWarning: A NumPy version >=1.17.3 and <1.25.0 is required for this version of SciPy (detected version 1.26.0
warnings.warn(f"A NumPy version >={np_minversion} and <{np_maxversion}"
params=None, kwargs={'differentiable': False}, desc=default
params=None, kwargs={'differentiable': True}, desc=default & differentiable
params=None, kwargs={'lr': 0.01, 'differentiable': False}, desc=non-default lr
params=None, kwargs={'lr': 0.01, 'differentiable': True}, desc=non-default lr & differentiable
params=None, kwargs={'weight_decay': 0.1, 'differentiable': False}, desc=nonzero weight_decay
params=None, kwargs={'weight_decay': 0.1, 'differentiable': True}, desc=nonzero weight_decay & differentiable
params=None, kwargs={'weight_decay': 0.1, 'maximize': True, 'differentiable': False}, desc=maximize
params=None, kwargs={'weight_decay': 0.1, 'maximize': True, 'differentiable': True}, desc=maximize & differentiable
params=None, kwargs={'weight_decay': 0.1, 'amsgrad': True, 'differentiable': False}, desc=amsgrad
params=None, kwargs={'weight_decay': 0.1, 'amsgrad': True, 'differentiable': True}, desc=amsgrad & differentiable
.params=None, kwargs={'foreach': False, 'differentiable': False, 'fused': False}, desc=default
params=None, kwargs={'foreach': True, 'differentiable': False, 'fused': False}, desc=default & foreach
params=None, kwargs={'foreach': False, 'differentiable': True, 'fused': False}, desc=default & differentiable
params=None, kwargs={'foreach': False, 'differentiable': False, 'fused': True}, desc=default & fused
params=None, kwargs={'lr': 0.01, 'foreach': False, 'differentiable': False, 'fused': False}, desc=non-default lr
params=None, kwargs={'lr': 0.01, 'foreach': True, 'differentiable': False, 'fused': False}, desc=non-default lr & foreach
params=None, kwargs={'lr': 0.01, 'foreach': False, 'differentiable': True, 'fused': False}, desc=non-default lr & differentiable
params=None, kwargs={'lr': 0.01, 'foreach': False, 'differentiable': False, 'fused': True}, desc=non-default lr & fused
params=None, kwargs={'weight_decay': 0.1, 'foreach': False, 'differentiable': False, 'fused': False}, desc=nonzero weight_decay
params=None, kwargs={'weight_decay': 0.1, 'foreach': True, 'differentiable': False, 'fused': False}, desc=nonzero weight_decay & foreach
params=None, kwargs={'weight_decay': 0.1, 'foreach': False, 'differentiable': True, 'fused': False}, desc=nonzero weight_decay & differentiable
params=None, kwargs={'weight_decay': 0.1, 'foreach': False, 'differentiable': False, 'fused': True}, desc=nonzero weight_decay & fused
params=None, kwargs={'weight_decay': 0.1, 'maximize': True, 'foreach': False, 'differentiable': False, 'fused': False}, desc=maximize
params=None, kwargs={'weight_decay': 0.1, 'maximize': True, 'foreach': True, 'differentiable': False, 'fused': False}, desc=maximize & foreach
params=None, kwargs={'weight_decay': 0.1, 'maximize': True, 'foreach': False, 'differentiable': True, 'fused': False}, desc=maximize & differentiable
params=None, kwargs={'weight_decay': 0.1, 'maximize': True, 'foreach': False, 'differentiable': False, 'fused': True}, desc=maximize & fused
params=None, kwargs={'weight_decay': 0.1, 'amsgrad': True, 'foreach': False, 'differentiable': False, 'fused': False}, desc=amsgrad
params=None, kwargs={'weight_decay': 0.1, 'amsgrad': True, 'foreach': True, 'differentiable': False, 'fused': False}, desc=amsgrad & foreach
params=None, kwargs={'weight_decay': 0.1, 'amsgrad': True, 'foreach': False, 'differentiable': True, 'fused': False}, desc=amsgrad & differentiable
params=None, kwargs={'weight_decay': 0.1, 'amsgrad': True, 'foreach': False, 'differentiable': False, 'fused': True}, desc=amsgrad & fused
params=None, kwargs={'capturable': True, 'foreach': False, 'differentiable': False, 'fused': False}, desc=capturable
params=None, kwargs={'capturable': True, 'foreach': True, 'differentiable': False, 'fused': False}, desc=capturable & foreach
params=None, kwargs={'capturable': True, 'foreach': False, 'differentiable': True, 'fused': False}, desc=capturable & differentiable
params=None, kwargs={'capturable': True, 'foreach': False, 'differentiable': False, 'fused': True}, desc=capturable & fused
params=None, kwargs={'weight_decay': 0.1, 'amsgrad': True, 'capturable': True, 'foreach': False, 'differentiable': False, 'fused': False}, desc=capturable, amsgrad
params=None, kwargs={'weight_decay': 0.1, 'amsgrad': True, 'capturable': True, 'foreach': True, 'differentiable': False, 'fused': False}, desc=capturable, amsgrad & foreach
params=None, kwargs={'weight_decay': 0.1, 'amsgrad': True, 'capturable': True, 'foreach': False, 'differentiable': True, 'fused': False}, desc=capturable, amsgrad & differentiable
params=None, kwargs={'weight_decay': 0.1, 'amsgrad': True, 'capturable': True, 'foreach': False, 'differentiable': False, 'fused': True}, desc=capturable, amsgrad & fused
params=None, kwargs={'lr': tensor(0.0010), 'amsgrad': True, 'capturable': True, 'foreach': False, 'differentiable': False, 'fused': False}, desc=Tensor lr with capturable and amsgrad
params=None, kwargs={'lr': tensor(0.0010), 'amsgrad': True, 'capturable': True, 'foreach': True, 'differentiable': False, 'fused': False}, desc=Tensor lr with capturable and amsgrad & foreach
params=None, kwargs={'lr': tensor(0.0010), 'amsgrad': True, 'capturable': True, 'foreach': False, 'differentiable': True, 'fused': False}, desc=Tensor lr with capturable and amsgrad & differentiable
params=None, kwargs={'lr': tensor(0.0010), 'amsgrad': True, 'capturable': True, 'foreach': False, 'differentiable': False, 'fused': True}, desc=Tensor lr with capturable and amsgrad & fused
.
----------------------------------------------------------------------
Ran 2 tests in 11.112s
OK
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/118326
Approved by: https://github.com/mlazos
This PR fixes several bugs, listed in priority:
1. `load_state_dict` with a nontensor step was incorrect for capturable and fused implementations since we don't create the tensors on the right device in `__setstate__`. This has been fixed.
2. The most recently added capturable implementations forgot the check that all tensors should be on CUDA for eager. We've now added those checks
3. The most recent change in Adamax only adds capturable for foreach but will silently be incorrect for forloop/single-tensor. I've added erroring and modified testing with many many many skips for that. Honestly my preference after this PR has only been further cemented that we should just do the single tensor and multi tensor capturable implementations together in the future. @mlazos
4. The conditional for adding cuda-supported configs for the optimizer infos was incorrect! So we hadn't been testing capturable! This also stands rectified and was the trigger for this PR in the first place.
5. In a similar way, the conditional for `_get_optim_inputs_including_global_cliquey_kwargs` was incorrect sometimes as well. This has also been corrected.
The following is not a bug, but is just something to make life simpler by not needing to handle Nones: `optim_input_funcs` must now mandatorily take in a `device`, which could be a string or a torch.device.
Details for posterity:
4. Running the test_foreach_matches_forloop test and printing the configs that get printed yields capturable getting included, which is correct.
```
(pytorch-3.10) [janeyx@devgpu023.odn1 ~/local/pytorch (5d50138f)]$ python test/test_optim.py -k test_foreach_matches_forloop_AdamW_cuda
/home/janeyx/.conda/envs/pytorch-3.10/lib/python3.10/site-packages/transformers/utils/generic.py:441: UserWarning: torch.utils._pytree._register_pytree_node is deprecated. Please use torch.utils._pytree.register_pytree_node instead.
_torch_pytree._register_pytree_node(
/home/janeyx/.conda/envs/pytorch-3.10/lib/python3.10/site-packages/scipy/__init__.py:146: UserWarning: A NumPy version >=1.17.3 and <1.25.0 is required for this version of SciPy (detected version 1.26.0
warnings.warn(f"A NumPy version >={np_minversion} and <{np_maxversion}"
params=None, kwargs={}, desc=default
params=None, kwargs={'lr': 0.01}, desc=non-default lr
params=None, kwargs={'weight_decay': 0.1}, desc=nonzero weight_decay
params=None, kwargs={'weight_decay': 0.1, 'maximize': True}, desc=maximize
params=None, kwargs={'weight_decay': 0.1, 'amsgrad': True}, desc=amsgrad
params=None, kwargs={'capturable': True}, desc=capturable
params=None, kwargs={'weight_decay': 0.1, 'amsgrad': True, 'capturable': True}, desc=capturable, amsgrad
params=None, kwargs={'lr': tensor(0.0010), 'amsgrad': True, 'capturable': True}, desc=Tensor lr with capturable and amsgrad
.
----------------------------------------------------------------------
Ran 1 test in 19.229s
OK
```
5. Running the test_optimizer_can_be_printed test (which calls `_get_optim_inputs_including_global_cliquey_kwargs`) and printing what gets run is also now correct.
```
/home/janeyx/.conda/envs/pytorch-3.10/lib/python3.10/site-packages/scipy/__init__.py:146: UserWarning: A NumPy version >=1.17.3 and <1.25.0 is required for this version of SciPy (detected version 1.26.0
warnings.warn(f"A NumPy version >={np_minversion} and <{np_maxversion}"
params=None, kwargs={'differentiable': False}, desc=default
params=None, kwargs={'differentiable': True}, desc=default & differentiable
params=None, kwargs={'lr': 0.01, 'differentiable': False}, desc=non-default lr
params=None, kwargs={'lr': 0.01, 'differentiable': True}, desc=non-default lr & differentiable
params=None, kwargs={'weight_decay': 0.1, 'differentiable': False}, desc=nonzero weight_decay
params=None, kwargs={'weight_decay': 0.1, 'differentiable': True}, desc=nonzero weight_decay & differentiable
params=None, kwargs={'weight_decay': 0.1, 'maximize': True, 'differentiable': False}, desc=maximize
params=None, kwargs={'weight_decay': 0.1, 'maximize': True, 'differentiable': True}, desc=maximize & differentiable
params=None, kwargs={'weight_decay': 0.1, 'amsgrad': True, 'differentiable': False}, desc=amsgrad
params=None, kwargs={'weight_decay': 0.1, 'amsgrad': True, 'differentiable': True}, desc=amsgrad & differentiable
.params=None, kwargs={'foreach': False, 'differentiable': False, 'fused': False}, desc=default
params=None, kwargs={'foreach': True, 'differentiable': False, 'fused': False}, desc=default & foreach
params=None, kwargs={'foreach': False, 'differentiable': True, 'fused': False}, desc=default & differentiable
params=None, kwargs={'foreach': False, 'differentiable': False, 'fused': True}, desc=default & fused
params=None, kwargs={'lr': 0.01, 'foreach': False, 'differentiable': False, 'fused': False}, desc=non-default lr
params=None, kwargs={'lr': 0.01, 'foreach': True, 'differentiable': False, 'fused': False}, desc=non-default lr & foreach
params=None, kwargs={'lr': 0.01, 'foreach': False, 'differentiable': True, 'fused': False}, desc=non-default lr & differentiable
params=None, kwargs={'lr': 0.01, 'foreach': False, 'differentiable': False, 'fused': True}, desc=non-default lr & fused
params=None, kwargs={'weight_decay': 0.1, 'foreach': False, 'differentiable': False, 'fused': False}, desc=nonzero weight_decay
params=None, kwargs={'weight_decay': 0.1, 'foreach': True, 'differentiable': False, 'fused': False}, desc=nonzero weight_decay & foreach
params=None, kwargs={'weight_decay': 0.1, 'foreach': False, 'differentiable': True, 'fused': False}, desc=nonzero weight_decay & differentiable
params=None, kwargs={'weight_decay': 0.1, 'foreach': False, 'differentiable': False, 'fused': True}, desc=nonzero weight_decay & fused
params=None, kwargs={'weight_decay': 0.1, 'maximize': True, 'foreach': False, 'differentiable': False, 'fused': False}, desc=maximize
params=None, kwargs={'weight_decay': 0.1, 'maximize': True, 'foreach': True, 'differentiable': False, 'fused': False}, desc=maximize & foreach
params=None, kwargs={'weight_decay': 0.1, 'maximize': True, 'foreach': False, 'differentiable': True, 'fused': False}, desc=maximize & differentiable
params=None, kwargs={'weight_decay': 0.1, 'maximize': True, 'foreach': False, 'differentiable': False, 'fused': True}, desc=maximize & fused
params=None, kwargs={'weight_decay': 0.1, 'amsgrad': True, 'foreach': False, 'differentiable': False, 'fused': False}, desc=amsgrad
params=None, kwargs={'weight_decay': 0.1, 'amsgrad': True, 'foreach': True, 'differentiable': False, 'fused': False}, desc=amsgrad & foreach
params=None, kwargs={'weight_decay': 0.1, 'amsgrad': True, 'foreach': False, 'differentiable': True, 'fused': False}, desc=amsgrad & differentiable
params=None, kwargs={'weight_decay': 0.1, 'amsgrad': True, 'foreach': False, 'differentiable': False, 'fused': True}, desc=amsgrad & fused
params=None, kwargs={'capturable': True, 'foreach': False, 'differentiable': False, 'fused': False}, desc=capturable
params=None, kwargs={'capturable': True, 'foreach': True, 'differentiable': False, 'fused': False}, desc=capturable & foreach
params=None, kwargs={'capturable': True, 'foreach': False, 'differentiable': True, 'fused': False}, desc=capturable & differentiable
params=None, kwargs={'capturable': True, 'foreach': False, 'differentiable': False, 'fused': True}, desc=capturable & fused
params=None, kwargs={'weight_decay': 0.1, 'amsgrad': True, 'capturable': True, 'foreach': False, 'differentiable': False, 'fused': False}, desc=capturable, amsgrad
params=None, kwargs={'weight_decay': 0.1, 'amsgrad': True, 'capturable': True, 'foreach': True, 'differentiable': False, 'fused': False}, desc=capturable, amsgrad & foreach
params=None, kwargs={'weight_decay': 0.1, 'amsgrad': True, 'capturable': True, 'foreach': False, 'differentiable': True, 'fused': False}, desc=capturable, amsgrad & differentiable
params=None, kwargs={'weight_decay': 0.1, 'amsgrad': True, 'capturable': True, 'foreach': False, 'differentiable': False, 'fused': True}, desc=capturable, amsgrad & fused
params=None, kwargs={'lr': tensor(0.0010), 'amsgrad': True, 'capturable': True, 'foreach': False, 'differentiable': False, 'fused': False}, desc=Tensor lr with capturable and amsgrad
params=None, kwargs={'lr': tensor(0.0010), 'amsgrad': True, 'capturable': True, 'foreach': True, 'differentiable': False, 'fused': False}, desc=Tensor lr with capturable and amsgrad & foreach
params=None, kwargs={'lr': tensor(0.0010), 'amsgrad': True, 'capturable': True, 'foreach': False, 'differentiable': True, 'fused': False}, desc=Tensor lr with capturable and amsgrad & differentiable
params=None, kwargs={'lr': tensor(0.0010), 'amsgrad': True, 'capturable': True, 'foreach': False, 'differentiable': False, 'fused': True}, desc=Tensor lr with capturable and amsgrad & fused
.
----------------------------------------------------------------------
Ran 2 tests in 11.112s
OK
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/118326
Approved by: https://github.com/mlazos
Summary: Exposes `dynamic_shapes` api at multiple levels so it's easier to replace the old API `dynamic_dim()` with the new API `Dim()`.
Test Plan: CI
Differential Revision: D53246409
Pull Request resolved: https://github.com/pytorch/pytorch/pull/118695
Approved by: https://github.com/ydwu4
Fixes https://github.com/pytorch/pytorch/issues/118129
Suppressions automatically added with
```
import re
with open("error_file.txt", "r") as f:
errors = f.readlines()
error_lines = {}
for error in errors:
match = re.match(r"(.*):(\d+):\d+: error:.*\[(.*)\]", error)
if match:
file_path, line_number, error_type = match.groups()
if file_path not in error_lines:
error_lines[file_path] = {}
error_lines[file_path][int(line_number)] = error_type
for file_path, lines in error_lines.items():
with open(file_path, "r") as f:
code = f.readlines()
for line_number, error_type in sorted(lines.items(), key=lambda x: x[0], reverse=True):
code[line_number - 1] = code[line_number - 1].rstrip() + f" # type: ignore[{error_type}]\n"
with open(file_path, "w") as f:
f.writelines(code)
```
Signed-off-by: Edward Z. Yang <ezyang@meta.com>
Co-authored-by: Catherine Lee <csl@fb.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/118533
Approved by: https://github.com/Skylion007, https://github.com/zou3519
Fixes https://github.com/pytorch/pytorch/issues/118129
Suppressions automatically added with
```
import re
with open("error_file.txt", "r") as f:
errors = f.readlines()
error_lines = {}
for error in errors:
match = re.match(r"(.*):(\d+):\d+: error:.*\[(.*)\]", error)
if match:
file_path, line_number, error_type = match.groups()
if file_path not in error_lines:
error_lines[file_path] = {}
error_lines[file_path][int(line_number)] = error_type
for file_path, lines in error_lines.items():
with open(file_path, "r") as f:
code = f.readlines()
for line_number, error_type in sorted(lines.items(), key=lambda x: x[0], reverse=True):
code[line_number - 1] = code[line_number - 1].rstrip() + f" # type: ignore[{error_type}]\n"
with open(file_path, "w") as f:
f.writelines(code)
```
Signed-off-by: Edward Z. Yang <ezyang@meta.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/118533
Approved by: https://github.com/Skylion007, https://github.com/zou3519
The original motivation for MYPYINDUCTOR was a faster type checking configuration that only checked a subset of files. With the removal of `follow_imports = ignore`, we are now able to use dmypy to do fast incremental typechecking, eliminating the need for this.
Perhaps erroneously, when I tee'ed up this PR I elected to delete the `follow_imports = skip` designations in the mypy-inductor.ini. This lead to a number of extra type error suppressions that I manually edited. You will need to review.
Signed-off-by: Edward Z. Yang <ezyang@meta.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/118432
Approved by: https://github.com/Skylion007
ghstack dependencies: #118414, #118418
I feel it's easier to open a new PR rather than iterating on the previous PR (https://github.com/pytorch/pytorch/pull/105257 ) since this is more like a rewrite.
In this PR, instead of changing GraphModule directly which can easily causes BC issue, I create a LazyGraphModule class as Zachary & Jason suggested in comments from the previous PR.
The difference between LazyGraphModule and GraphModule is mainly about how re-compile for the graph module happens. In GraphModule the recompilation happens 'eagerly': constructing a GraphModule will cause the recompilation. While in LazyGraphModule, we just mark the module as needing recompilation. The real recompilation only happens when absolutely required (e.g. call forward method, access the code property etc.). In a lot of cases in torch.compile, the real recompilation eventually is not triggered at all. This can save a few seconds of compilation time.
By default, GraphModule rather than LazyGraphModule is used. `use_lazy_graph_module(True)` context manager can be used to pick LazyGraphModule instead. This has been applied to the torch.compile stack.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/117911
Approved by: https://github.com/jansel
Added support for constant outputs. We will just embed the constant directly into the output, like `return (x, 1)`.
Also adds support for None input/outputs. For None inputs we address it the same way we do to constants, which is that a placeholder with no users will be inserted into the graph, and the None will be embedded into whatever operator is using the None. For None outputs, we will also address the same way we do constants, which is that we embed it into the output, like `return (x, None)`.
Differential Revision: D52881070
Pull Request resolved: https://github.com/pytorch/pytorch/pull/117894
Approved by: https://github.com/zhxchen17
1. I'd like to remove the patching that avoids the profiler hook, but it adds an additional graph break due to nested wrappers. #117767 if interested, see (internal only) paste for [before](P996529232) and [after](P997507449) this PR.
```
I've locally run perf benchmarks for yolov3: Before the speedup is 4.183x, and after it is 4.208x.
I've also run it for resnet50: before, speedup is 3.706x and now it is 3.924x.
```
2. @mlazos I now unwrap twice in the dynamo and inductor tests. This feels like we're testing deficiently--should we add tests to test that tracing through the profiler hook and the use_grad hook are functioning according to expectations (I know there's at least one graph break in one).
3. There's a strange memory thing going on...what is happening? This has been resolved with @voznesenskym's [change](https://github.com/pytorch/pytorch/pull/116169). (for details see below)
<details>
This PR will fail the test_static_address_finalizer test due to a mysterious thing that is happening (idk what, but maybe the dynamo cache or a frame _expecting_ the patching to have been done).
There is no Python refcycle, as the backrefs for `p_ref()` look like:
(so 5 backrefs but none of them python)
And the refs:
</details>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/115772
Approved by: https://github.com/jansel, https://github.com/mlazos
