Add new dispatch keys for Fake Tensor and Deferred Module Initialization

[cbalioglu]

Thanks to @bdhirsh's work, we now have room for new dispatch keys in DispatchKey enum. This PR adds two new keys for out-of-core Fake Tensor and Deferred Module Initialization features.

[facebook-github-bot]

🔗 Helpful links

• 🧪  See artifacts and rendered test results at hud.pytorch.org/pr/76139
• 📄  Preview Python docs built from this PR
• 📄  Preview C++ docs built from this PR
• ❓Need help or want to give feedback on the CI? Visit our office hours
• ↩️  [fb-only] Re-run with SSH instructions

💊 CI failures summary and remediations

As of commit f3815bb (more details on the Dr. CI page):

Expand to see more

---

• 1/1 failures introduced in this PR

---

🕵️ 1 new failure recognized by patterns

The following CI failures do not appear to be due to upstream breakages

pull / linux-xenial-py3.7-gcc5.4 / test (backwards_compat, 1, 1, linux.2xlarge) (1/1)

Step: "Test" (full log | diagnosis details | 🔁 rerun)

2022-04-27T14:53:34.1898475Z The PR is introduc...m to confirm whether this change is wanted or not.

2022-04-27T14:53:34.1885711Z processing existing schema:  text(__torch__.torch.classes.profiling.SourceRef _0) -> (str _0)
2022-04-27T14:53:34.1886760Z processing existing schema:  count(__torch__.torch.classes.profiling.InstructionStats _0) -> (int _0)
2022-04-27T14:53:34.1888011Z processing existing schema:  duration_ns(__torch__.torch.classes.profiling.InstructionStats _0) -> (int _0)
2022-04-27T14:53:34.1889351Z processing existing schema:  source(__torch__.torch.classes.profiling.SourceStats _0) -> (__torch__.torch.classes.profiling.SourceRef _0)
2022-04-27T14:53:34.1891075Z processing existing schema:  line_map(__torch__.torch.classes.profiling.SourceStats _0) -> (Dict(int, __torch__.torch.classes.profiling.InstructionStats) _0)
2022-04-27T14:53:34.1892089Z processing existing schema:  __init__(__torch__.torch.classes.profiling._ScriptProfile _0) -> (NoneType _0)
2022-04-27T14:53:34.1893324Z processing existing schema:  enable(__torch__.torch.classes.profiling._ScriptProfile _0) -> (NoneType _0)
2022-04-27T14:53:34.1894605Z processing existing schema:  disable(__torch__.torch.classes.profiling._ScriptProfile _0) -> (NoneType _0)
2022-04-27T14:53:34.1896391Z processing existing schema:  _dump_stats(__torch__.torch.classes.profiling._ScriptProfile _0) -> (__torch__.torch.classes.profiling.SourceStats[] _0)
2022-04-27T14:53:34.1898034Z processing existing schema:  __init__(__torch__.torch.classes.dist_rpc.WorkerInfo _0, str _1, int _2) -> (NoneType _0)
2022-04-27T14:53:34.1898475Z The PR is introducing backward incompatible changes to the operator library. Please contact PyTorch team to confirm whether this change is wanted or not. 
2022-04-27T14:53:34.1898497Z 
2022-04-27T14:53:34.1898632Z Broken ops: [
2022-04-27T14:53:34.1899116Z 	aten::_thnn_fused_lstm_cell_backward_impl(Tensor? grad_hy, Tensor? grad_cy, Tensor cx, Tensor cy, Tensor workspace, bool has_bias) -> (Tensor, Tensor, Tensor)
2022-04-27T14:53:34.1899182Z ]
2022-04-27T14:53:34.2930153Z + cleanup
2022-04-27T14:53:34.2930300Z + retcode=1
2022-04-27T14:53:34.2930425Z + set +x
2022-04-27T14:53:34.2964643Z ##[error]Process completed with exit code 1.
2022-04-27T14:53:34.3002114Z ##[group]Run pytorch/pytorch/.github/actions/get-workflow-job-id@master
2022-04-27T14:53:34.3002185Z with:

---

This comment was automatically generated by Dr. CI (expand for details).

Please report bugs/suggestions to the (internal) Dr. CI Users group.

Click here to manually regenerate this comment.

[cbalioglu]

Per our discussion in today's meeting, bumping up this PR.

cc @bdhirsh

[bdhirsh]

Are you sure that you want Fake to have higher priority than ZeroTensor/Negative/Conjugate? (I think I'd expect it to have ~ the same priority as the Python key, right above backend select, or you might hit weird issues with code that uses forward mode AD / complex numbers)

[cbalioglu]

In fact it works in both orders, but has different effects (Zero tensor treated as Fake vs. Fake treated as Zero). Anyways, as you suggested, I moved it next to Python to make it more consistent with the behavior of Python subclasses.

[bdhirsh]

Left a comment, but preemptively approving as per the composability conversation earlier today

[cbalioglu]

Note: verified that the failing GitHub Action is not related to the changes included in this PR.

[cbalioglu]

@pytorchbot merge this please

[github-actions]

Hey @cbalioglu.
You've committed this PR, but it does not have both a 'release notes: ...' and 'topics: ...' label. Please add one of each to the PR. The 'release notes: ...' label should represent the part of PyTorch that this PR changes (fx, autograd, distributed, etc) and the 'topics: ...' label should represent the kind of PR it is (not user facing, new feature, bug fix, perf improvement, etc). The list of valid labels can be found here for the 'release notes: ...' and here for the 'topics: ...'.
For changes that are 'topic: not user facing' there is no need for a release notes label.