Fix caching allocator of out-of-tree device is destructed before the … (

pytorch#126677)

…destruction of tensors cached by autocast

## Root Cause
For out-of-tree device extension it is loaded after torch (different .so), so the global variable `cached_casts` may be constructed before caching allocator and then destructed in reversed order when exit.

## Fix
Lazily initialize `cached_casts` to correct the order.

## How to Reproduce && Test
Modify the testcase `TestAutocastGPU.test_cast_cache_is_global` in test/test_autocast.py  to run on your out-of-tree device. You will see following failure in the end of test.
```bash
----------------------------------------------------------------------
Ran 1 test in 4.812s

OK
free: 0x30080ff44000400
terminate called after throwing an instance of 'c10::Error'
  what():  invalid device pointer: 0x30080ff44000400
Exception raised from free at /projs/framework/betterman/code/pytorch_new/catch/torch_mlu/csrc/framework/core/caching_allocator.cpp:1609 (most recent call first):
frame #0: <unknown function> + 0x118fe1 (0x7ffaef4d3fe1 in /projs/framework/betterman/code/pytorch_new/torch/lib/libc10.so)
frame #1: <unknown function> + 0x11b1c4 (0x7ffaef4d61c4 in /projs/framework/betterman/code/pytorch_new/torch/lib/libc10.so)
frame #2: <unknown function> + 0x117677 (0x7ffaef4d2677 in /projs/framework/betterman/code/pytorch_new/torch/lib/libc10.so)
frame #3: <unknown function> + 0x11a2bf (0x7ffaef4d52bf in /projs/framework/betterman/code/pytorch_new/torch/lib/libc10.so)
frame #4: <unknown function> + 0x11a186 (0x7ffaef4d5186 in /projs/framework/betterman/code/pytorch_new/torch/lib/libc10.so)
frame #5: <unknown function> + 0x119fde (0x7ffaef4d4fde in /projs/framework/betterman/code/pytorch_new/torch/lib/libc10.so)
frame #6: <unknown function> + 0x119d2e (0x7ffaef4d4d2e in /projs/framework/betterman/code/pytorch_new/torch/lib/libc10.so)
frame #7: <unknown function> + 0x119be0 (0x7ffaef4d4be0 in /projs/framework/betterman/code/pytorch_new/torch/lib/libc10.so)
frame #8: <unknown function> + 0x119977 (0x7ffaef4d4977 in /projs/framework/betterman/code/pytorch_new/torch/lib/libc10.so)
frame #9: <unknown function> + 0x119313 (0x7ffaef4d4313 in /projs/framework/betterman/code/pytorch_new/torch/lib/libc10.so)
frame #10: <unknown function> + 0x118b4c (0x7ffaef4d3b4c in /projs/framework/betterman/code/pytorch_new/torch/lib/libc10.so)
frame #11: c10::Error::Error(c10::SourceLocation, std::string) + 0x34 (0x7ffaef4d27c4 in /projs/framework/betterman/code/pytorch_new/torch/lib/libc10.so)
frame #12: c10::detail::torchCheckFail(char const*, char const*, unsigned int, std::string const&) + 0x7f (0x7ffaef4d04ed in /projs/framework/betterman/code/pytorch_new/torch/lib/libc10.so)
frame #13: torch_mlu::MLUCachingAllocator::Native::NativeCachingAllocator::free(void*) + 0xe6 (0x7ff9a8eeb112 in /projs/framework/betterman/code/pytorch_new/catch/torch_mlu/csrc/lib/libtorch_mlu.so)
frame #14: torch_mlu::MLUCachingAllocator::Native::local_raw_delete(void*) + 0x3b (0x7ff9a8ed9480 in /projs/framework/betterman/code/pytorch_new/catch/torch_mlu/csrc/lib/libtorch_mlu.so)                                                                                                                         frame #15: std::unique_ptr<void, void (*)(void*)>::~unique_ptr() + 0x50 (0x7ffb0a5ea322 in /projs/framework/betterman/code/pytorch_new/torch/lib/libtorch_python.so)
frame #16: <unknown function> + 0x1269890 (0x7ffb0a5e4890 in /projs/framework/betterman/code/pytorch_new/torch/lib/libtorch_python.so)
frame #17: <unknown function> + 0x1269928 (0x7ffb0a5e4928 in /projs/framework/betterman/code/pytorch_new/torch/lib/libtorch_python.so)
frame #18: <unknown function> + 0x127572c (0x7ffb0a5f072c in /projs/framework/betterman/code/pytorch_new/torch/lib/libtorch_python.so)
frame #19: <unknown function> + 0x1275758 (0x7ffb0a5f0758 in /projs/framework/betterman/code/pytorch_new/torch/lib/libtorch_python.so)
frame #20: <unknown function> + 0xb9bc7 (0x7ffaef474bc7 in /projs/framework/betterman/code/pytorch_new/torch/lib/libc10.so)
frame #21: <unknown function> + 0xb97bc (0x7ffaef4747bc in /projs/framework/betterman/code/pytorch_new/torch/lib/libc10.so)
frame #22: <unknown function> + 0xdbc50 (0x7ffaef496c50 in /projs/framework/betterman/code/pytorch_new/torch/lib/libc10.so)
frame #23: c10::TensorImpl::~TensorImpl() + 0x82 (0x7ffaef49157e in /projs/framework/betterman/code/pytorch_new/torch/lib/libc10.so)
frame #24: c10::TensorImpl::~TensorImpl() + 0x1c (0x7ffaef4915aa in /projs/framework/betterman/code/pytorch_new/torch/lib/libc10.so)
frame #25: <unknown function> + 0x2f596d9 (0x7ffaf24fc6d9 in /projs/framework/betterman/code/pytorch_new/torch/lib/libtorch_cpu.so)
frame #26: <unknown function> + 0x2f589c2 (0x7ffaf24fb9c2 in /projs/framework/betterman/code/pytorch_new/torch/lib/libtorch_cpu.so)
frame #27: <unknown function> + 0x2f57b92 (0x7ffaf24fab92 in /projs/framework/betterman/code/pytorch_new/torch/lib/libtorch_cpu.so)
frame #28: <unknown function> + 0x2f5c228 (0x7ffaf24ff228 in /projs/framework/betterman/code/pytorch_new/torch/lib/libtorch_cpu.so)
frame #29: <unknown function> + 0x30f3f70 (0x7ffaf2696f70 in /projs/framework/betterman/code/pytorch_new/torch/lib/libtorch_cpu.so)
frame #30: <unknown function> + 0x30f3f90 (0x7ffaf2696f90 in /projs/framework/betterman/code/pytorch_new/torch/lib/libtorch_cpu.so)
frame #31: <unknown function> + 0x30f5004 (0x7ffaf2698004 in /projs/framework/betterman/code/pytorch_new/torch/lib/libtorch_cpu.so)                                                                                                                                                                                frame #32: <unknown function> + 0x30f5024 (0x7ffaf2698024 in /projs/framework/betterman/code/pytorch_new/torch/lib/libtorch_cpu.so)
frame #33: <unknown function> + 0x31207f0 (0x7ffaf26c37f0 in /projs/framework/betterman/code/pytorch_new/torch/lib/libtorch_cpu.so)
frame #34: <unknown function> + 0x3120814 (0x7ffaf26c3814 in /projs/framework/betterman/code/pytorch_new/torch/lib/libtorch_cpu.so)
frame #35: <unknown function> + 0x30f51e8 (0x7ffaf26981e8 in /projs/framework/betterman/code/pytorch_new/torch/lib/libtorch_cpu.so)
frame #36: <unknown function> + 0x30f5148 (0x7ffaf2698148 in /projs/framework/betterman/code/pytorch_new/torch/lib/libtorch_cpu.so)
frame #37: <unknown function> + 0x316ecea (0x7ffaf2711cea in /projs/framework/betterman/code/pytorch_new/torch/lib/libtorch_cpu.so)
frame #38: <unknown function> + 0x468a7 (0x7ffb0c9ed8a7 in /lib/x86_64-linux-gnu/libc.so.6)
frame #39: on_exit + 0 (0x7ffb0c9eda60 in /lib/x86_64-linux-gnu/libc.so.6)
<omitting python frames>
frame #47: __libc_start_main + 0xf3 (0x7ffb0c9cb083 in /lib/x86_64-linux-gnu/libc.so.6)

Aborted (core dumped)

```

Pull Request resolved: pytorch#126677
Approved by: https://github.com/ezyang

aten/src/ATen/autocast_mode.cpp

@@ -35,7 +35,11 @@ namespace {
 // directly against incoming TensorImpl*s.
 using weakref_type = c10::weak_intrusive_ptr<TensorImpl, UndefinedTensorImpl>;
 using val_type = std::tuple<weakref_type, Tensor>;
-ska::flat_hash_map<TensorImpl*, val_type> cached_casts;
+
+static ska::flat_hash_map<TensorImpl*, val_type>& get_cached_casts() {
+  static ska::flat_hash_map<TensorImpl*, val_type> cached_casts;
+  return cached_casts;
+}
 std::mutex cached_casts_mutex;
 
 
@@ -82,7 +86,7 @@ thread_local bool cache_enabled = true;
 
 void clear_cache() {
   const std::lock_guard<std::mutex> lock(cached_casts_mutex);
-  cached_casts.clear();
+  get_cached_casts().clear();
 }
 
 int increment_nesting() {
@@ -124,12 +128,12 @@ Tensor cached_cast(at::ScalarType to_type, const Tensor& arg, DeviceType device_
 
     if (can_try_cache) {
       const std::lock_guard<std::mutex> lock(cached_casts_mutex);
-      auto it = cached_casts.find(arg.unsafeGetTensorImpl());
-      if (it != cached_casts.end()) {
+      auto it = get_cached_casts().find(arg.unsafeGetTensorImpl());
+      if (it != get_cached_casts().end()) {
         return std::get<1>(it->second);
       } else {
         auto casted_arg = arg.to(to_type);
-        cached_casts.emplace(arg.unsafeGetTensorImpl(), val_type{weakref_type(arg.getIntrusivePtr()), casted_arg});
+        get_cached_casts().emplace(arg.unsafeGetTensorImpl(), val_type{weakref_type(arg.getIntrusivePtr()), casted_arg});
         return casted_arg;
       }
     } else {