Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

ensure scalars are float32 in test cases for dropout #174

Merged
merged 1 commit into from
Aug 23, 2024
Merged

ensure scalars are float32 in test cases for dropout #174

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 10 additions & 2 deletions tests/test_special_ops.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,5 +1,6 @@
from typing import Optional

import numpy as np
import pytest
import torch

Expand All @@ -24,6 +25,11 @@ def test_accuracy_dropout(shape, p, dtype):
inp = torch.randn(shape, dtype=dtype, device="cuda", requires_grad=True)
ref_inp = to_reference(inp)

# NOTE: ensure that scalars are float32(instead of float64)
# in some cases, casting up then casting down have different result
p = np.float32(p)
one_minus_p = np.float32(1.0) - p

ref_out = torch.nn.functional.dropout(ref_inp, p, True)
with flag_gems.use_gems():
res_out = torch.nn.functional.dropout(inp, p, True)
Expand All @@ -37,13 +43,15 @@ def test_accuracy_dropout(shape, p, dtype):
res_out = to_reference(res_out)
res_in_grad = to_reference(res_in_grad)

exp_equal = (p * p + (1 - p) * (1 - p)) * inp.numel()
exp_equal = (p * p + one_minus_p * one_minus_p) * inp.numel()
num_equal = torch.sum(torch.isclose(ref_out, res_out)).item()
if TO_CPU:
zero_equal = torch.eq(res_out, torch.zeros_like(res_out))
num_zero = torch.sum(zero_equal).item()
assert abs(num_zero / inp.numel() - p) <= 0.05
scale_equal = torch.isclose(res_out, ref_inp / (1 - p), rtol=RESOLUTION[dtype])
scale_equal = torch.isclose(
res_out, ref_inp / one_minus_p, rtol=RESOLUTION[dtype]
)
assert torch.all(torch.logical_or(zero_equal, scale_equal))
else:
assert (
Expand Down
Loading