test_accelerator.py

# Owner(s): ["module: tests"]

import gc
import sys
import unittest
from contextlib import nullcontext

import torch
from torch.testing._internal.common_utils import (
    NoTest,
    run_tests,
    TEST_ACCELERATOR,
    TEST_MPS,
    TEST_MULTIACCELERATOR,
    TEST_XPU,
    TestCase,
)


if not TEST_ACCELERATOR:
    print("No available accelerator detected, skipping tests", file=sys.stderr)
    TestCase = NoTest
    # Skip because failing when run on cuda build with no GPU, see #150059 for example
    sys.exit()


class TestAccelerator(TestCase):
    def test_current_accelerator(self):
        self.assertTrue(torch.accelerator.is_available())
        accelerators = ["cuda", "xpu", "mps"]
        for accelerator in accelerators:
            if torch.get_device_module(accelerator).is_available():
                self.assertEqual(
                    torch.accelerator.current_accelerator().type, accelerator
                )
                self.assertIsNone(torch.accelerator.current_accelerator().index)
                with self.assertRaisesRegex(
                    ValueError, "doesn't match the current accelerator"
                ):
                    torch.accelerator.set_device_index("cpu")

    @unittest.skipIf(not TEST_MULTIACCELERATOR, "only one accelerator detected")
    def test_generic_multi_device_behavior(self):
        orig_device = torch.accelerator.current_device_index()
        target_device = (orig_device + 1) % torch.accelerator.device_count()

        torch.accelerator.set_device_index(target_device)
        self.assertEqual(target_device, torch.accelerator.current_device_index())
        torch.accelerator.set_device_index(orig_device)
        self.assertEqual(orig_device, torch.accelerator.current_device_index())

        s1 = torch.Stream(target_device)
        torch.accelerator.set_stream(s1)
        self.assertEqual(target_device, torch.accelerator.current_device_index())
        torch.accelerator.synchronize(orig_device)
        self.assertEqual(target_device, torch.accelerator.current_device_index())

    def test_generic_stream_behavior(self):
        s1 = torch.Stream()
        s2 = torch.Stream()
        torch.accelerator.set_stream(s1)
        self.assertEqual(torch.accelerator.current_stream(), s1)
        event = torch.Event()
        a = torch.randn(1000)
        b = torch.randn(1000)
        c = a + b
        torch.accelerator.set_stream(s2)
        self.assertEqual(torch.accelerator.current_stream(), s2)
        a_acc = a.to(torch.accelerator.current_accelerator(), non_blocking=True)
        b_acc = b.to(torch.accelerator.current_accelerator(), non_blocking=True)
        torch.accelerator.set_stream(s1)
        self.assertEqual(torch.accelerator.current_stream(), s1)
        event.record(s2)
        event.synchronize()
        c_acc = a_acc + b_acc
        event.record(s2)
        torch.accelerator.synchronize()
        self.assertTrue(event.query())
        self.assertEqual(c_acc.cpu(), c)

    def test_current_stream_query(self):
        s = torch.accelerator.current_stream()
        self.assertEqual(torch.accelerator.current_stream(s.device), s)
        self.assertEqual(torch.accelerator.current_stream(s.device.index), s)
        self.assertEqual(torch.accelerator.current_stream(str(s.device)), s)
        other_device = torch.device("cpu")
        with self.assertRaisesRegex(
            ValueError, "doesn't match the current accelerator"
        ):
            torch.accelerator.current_stream(other_device)

    def test_device_context_manager(self):
        prev_device = torch.accelerator.current_device_index()
        with torch.accelerator.device_index(None):
            self.assertEqual(torch.accelerator.current_device_index(), prev_device)
        self.assertEqual(torch.accelerator.current_device_index(), prev_device)
        with torch.accelerator.device_index(0):
            self.assertEqual(torch.accelerator.current_device_index(), 0)
        self.assertEqual(torch.accelerator.current_device_index(), prev_device)

    def test_device_index_fullgraph(self):
        def fn(x):
            with torch.accelerator.device_index(x.device.index):
                x = torch.sin(x + 1)
            return x

        x = torch.randn((2, 2), device=0)
        ref = fn(x)
        opt_fn = torch.compile(backend="eager", fullgraph=True)(fn)
        res = opt_fn(x)
        self.assertEqual(ref, res)

    @unittest.skipIf(not TEST_MULTIACCELERATOR, "only one accelerator detected")
    def test_multi_device_context_manager(self):
        src_device = 0
        dst_device = 1
        torch.accelerator.set_device_index(src_device)
        with torch.accelerator.device_index(dst_device):
            self.assertEqual(torch.accelerator.current_device_index(), dst_device)
        self.assertEqual(torch.accelerator.current_device_index(), src_device)

    def test_stream_context_manager(self):
        prev_stream = torch.accelerator.current_stream()
        with torch.Stream() as s:
            self.assertEqual(torch.accelerator.current_stream(), s)
        self.assertEqual(torch.accelerator.current_stream(), prev_stream)

    def test_stream_context_manager_reentrance(self):
        prev_stream = torch.accelerator.current_stream()
        s0 = torch.Stream()
        with s0, s0:
            self.assertEqual(torch.accelerator.current_stream(), s0)
        self.assertEqual(torch.accelerator.current_stream(), prev_stream)
        s1 = torch.Stream()
        with s0:
            self.assertEqual(torch.accelerator.current_stream(), s0)
            with s1:
                self.assertEqual(torch.accelerator.current_stream(), s1)
                with s0:
                    self.assertEqual(torch.accelerator.current_stream(), s0)
        self.assertEqual(torch.accelerator.current_stream(), prev_stream)

    @unittest.skipIf(not TEST_MULTIACCELERATOR, "only one accelerator detected")
    def test_multi_device_stream_context_manager(self):
        src_device = 0
        dst_device = 1
        torch.accelerator.set_device_index(src_device)
        src_prev_stream = torch.accelerator.current_stream()
        dst_prev_stream = torch.accelerator.current_stream(dst_device)
        with torch.Stream(dst_device) as dst_stream:
            self.assertEqual(torch.accelerator.current_device_index(), dst_device)
            self.assertEqual(torch.accelerator.current_stream(), dst_stream)
            self.assertEqual(
                torch.accelerator.current_stream(src_device), src_prev_stream
            )
        self.assertEqual(torch.accelerator.current_device_index(), src_device)
        self.assertEqual(torch.accelerator.current_stream(), src_prev_stream)
        self.assertEqual(torch.accelerator.current_stream(dst_device), dst_prev_stream)

    @unittest.skipIf(TEST_MPS, "MPS doesn't support pin memory!")
    def test_pin_memory_on_non_blocking_copy(self):
        t_acc = torch.randn(100).to(torch.accelerator.current_accelerator())
        t_host = t_acc.to("cpu", non_blocking=True)
        torch.accelerator.synchronize()
        self.assertTrue(t_host.is_pinned())
        self.assertEqual(t_acc.cpu(), t_host)

    def test_generic_event_behavior(self):
        event1 = torch.Event(enable_timing=False)
        event2 = torch.Event(enable_timing=False)
        with self.assertRaisesRegex(
            ValueError,
            "Both events must be created with argument 'enable_timing=True'",
        ):
            event1.elapsed_time(event2)

        event1 = torch.Event(enable_timing=True)
        event2 = torch.Event(enable_timing=True)
        with self.assertRaisesRegex(
            ValueError,
            "Both events must be recorded before calculating elapsed time",
        ):
            event1.elapsed_time(event2)

        # check default value of enable_timing: False
        event1 = torch.Event()
        event2 = torch.Event()
        with self.assertRaisesRegex(
            ValueError,
            "Both events must be created with argument 'enable_timing=True'",
        ):
            event1.elapsed_time(event2)

    @unittest.skipIf(TEST_MPS, "MPS doesn't support torch.accelerator memory API!")
    def test_memory_stats(self):
        # Ensure that device allocator is initialized
        acc = torch.accelerator.current_accelerator()
        tmp = torch.randn(100, device=acc)
        del tmp
        gc.collect()
        self.assertTrue(torch._C._accelerator_isAllocatorInitialized())
        torch.accelerator.empty_cache()

        pool_type = ["all", "small_pool", "large_pool"]
        metric_type = ["peak", "current", "allocated", "freed"]
        stats_type = [
            "allocated_bytes",
            "reserved_bytes",
            "active_bytes",
            "requested_bytes",
        ]
        mem_stats = torch.accelerator.memory_stats()
        expected_stats = [
            f"{st}.{pt}.{mt}"
            for st in stats_type
            for pt in pool_type
            for mt in metric_type
        ]
        missing_stats = [stat for stat in expected_stats if stat not in mem_stats]
        self.assertEqual(
            len(missing_stats),
            0,
            f"Missing expected memory statistics: {missing_stats}",
        )

        prev_allocated = torch.accelerator.memory_allocated()
        prev_reserved = torch.accelerator.memory_reserved()
        prev_max_allocated = torch.accelerator.max_memory_allocated()
        prev_max_reserved = torch.accelerator.max_memory_reserved()
        self.assertGreaterEqual(prev_allocated, 0)
        self.assertGreaterEqual(prev_reserved, 0)
        self.assertGreater(prev_max_allocated, 0)
        self.assertGreater(prev_max_reserved, 0)
        tmp = torch.ones(256, device=acc)
        self.assertGreater(torch.accelerator.memory_allocated(), prev_allocated)
        self.assertGreaterEqual(torch.accelerator.memory_reserved(), prev_reserved)
        del tmp
        gc.collect()
        torch.accelerator.empty_cache()
        torch.accelerator.reset_peak_memory_stats()
        self.assertEqual(torch.accelerator.memory_allocated(), prev_allocated)
        self.assertEqual(torch.accelerator.memory_reserved(), prev_reserved)
        torch.accelerator.reset_accumulated_memory_stats()
        prev_max_allocated = torch.accelerator.max_memory_allocated()
        prev_max_reserved = torch.accelerator.max_memory_reserved()
        # Activate 1kB memory
        prev_active_current = torch.accelerator.memory_stats()[
            "active_bytes.all.current"
        ]
        tmp = torch.randn(256, device=acc)
        # Detect if the current active memory is 1kB
        self.assertEqual(
            torch.accelerator.memory_stats()["active_bytes.all.current"],
            1024 + prev_active_current,
        )
        self.assertEqual(torch.accelerator.memory_stats()["active_bytes.all.freed"], 0)
        del tmp
        gc.collect()
        torch.accelerator.empty_cache()
        self.assertEqual(
            torch.accelerator.memory_stats()["active_bytes.all.current"],
            prev_active_current,
        )
        self.assertEqual(
            torch.accelerator.memory_stats()["active_bytes.all.freed"], 1024
        )
        torch.accelerator.reset_peak_memory_stats()
        self.assertEqual(torch.accelerator.max_memory_allocated(), prev_max_allocated)
        self.assertEqual(torch.accelerator.max_memory_reserved(), prev_max_reserved)

    @unittest.skipIf(TEST_MPS, "MPS doesn't support torch.accelerator memory API!")
    def test_get_memory_info(self):
        free_bytes, total_bytes = torch.accelerator.get_memory_info()
        self.assertGreaterEqual(free_bytes, 0)
        self.assertGreaterEqual(total_bytes, 0)

    @unittest.skipIf(
        TEST_XPU,
        "bare-bones/opaque bit-field dtypes cause undefined behavior on XPU, see https://github.com/pytorch/pytorch/issues/179888",
    )
    def test_device_capability_supported_dtypes(self):
        try:
            caps = torch.accelerator.get_device_capability()
        except RuntimeError:
            self.skipTest("Backend doesn't support get_device_capability")

        supported_dtypes = caps["supported_dtypes"]
        self.assertIsInstance(supported_dtypes, set)
        self.assertGreater(len(supported_dtypes), 0)

        acc = torch.accelerator.current_accelerator()
        reference_dtype = next(iter(supported_dtypes))

        all_dtypes = [
            getattr(torch, name)
            for name in dir(torch)
            if isinstance(getattr(torch, name), torch.dtype)
        ]
        for dtype in all_dtypes:
            with self.subTest(dtype=dtype):
                ctx = (
                    nullcontext()
                    if dtype in supported_dtypes
                    else self.assertRaises((RuntimeError, TypeError))
                )
                with ctx:
                    t = torch.empty(16, dtype=dtype, device=acc)
                    t = t.to(reference_dtype)
                    t = t.to(dtype)


if __name__ == "__main__":
    run_tests()