From bd0c922a5e91a018ec289edf568fb8d752ff5ffb Mon Sep 17 00:00:00 2001 From: zhangyue Date: Tue, 10 Feb 2026 01:40:29 +0000 Subject: [PATCH 1/7] issue/1008: mv "import infinicore" ahead of "import" torch --- test/infinicore/ops/abs.py | 2 +- test/infinicore/ops/acos.py | 2 +- test/infinicore/ops/acosh.py | 2 +- test/infinicore/ops/adaptive_avg_pool1d.py | 2 +- test/infinicore/ops/adaptive_avg_pool2d.py | 2 +- test/infinicore/ops/adaptive_avg_pool3d.py | 2 +- test/infinicore/ops/adaptive_max_pool1d.py | 2 +- test/infinicore/ops/adaptive_max_pool2d.py | 2 +- test/infinicore/ops/add.py | 2 +- test/infinicore/ops/add_rms_norm.py | 2 +- test/infinicore/ops/addbmm.py | 2 +- test/infinicore/ops/addcdiv.py | 2 +- test/infinicore/ops/addcmul.py | 2 +- test/infinicore/ops/addmv.py | 2 +- test/infinicore/ops/addr.py | 2 +- test/infinicore/ops/affine_grid.py | 2 +- test/infinicore/ops/all.py | 2 +- test/infinicore/ops/alpha_dropout.py | 2 +- test/infinicore/ops/amax.py | 2 +- test/infinicore/ops/amin.py | 2 +- test/infinicore/ops/aminmax.py | 2 +- test/infinicore/ops/any.py | 2 +- test/infinicore/ops/argmax.py | 2 +- test/infinicore/ops/argmin.py | 2 +- test/infinicore/ops/argsort.py | 2 +- test/infinicore/ops/argwhere.py | 2 +- test/infinicore/ops/asin.py | 2 +- test/infinicore/ops/asinh.py | 2 +- test/infinicore/ops/atan.py | 2 +- test/infinicore/ops/atan2.py | 2 +- test/infinicore/ops/atanh.py | 2 +- test/infinicore/ops/avg_pool1d.py | 2 +- test/infinicore/ops/avg_pool2d.py | 2 +- test/infinicore/ops/avg_pool3d.py | 2 +- test/infinicore/ops/baddbmm.py | 2 +- test/infinicore/ops/batch_norm.py | 2 +- test/infinicore/ops/bilinear.py | 2 +- test/infinicore/ops/binary_cross_entropy.py | 2 +- .../infinicore/ops/binary_cross_entropy_with_logits.py | 2 +- test/infinicore/ops/bincount.py | 2 +- test/infinicore/ops/bitwise_left_shift.py | 2 +- test/infinicore/ops/bitwise_right_shift.py | 2 +- test/infinicore/ops/bitwise_xor.py | 2 +- test/infinicore/ops/block_diag.py | 2 +- test/infinicore/ops/broadcast_to.py | 2 +- test/infinicore/ops/bucketize.py | 2 +- test/infinicore/ops/cat.py | 2 +- test/infinicore/ops/causal_softmax.py | 2 +- test/infinicore/ops/cdist.py | 2 +- test/infinicore/ops/celu.py | 2 +- test/infinicore/ops/clone.py | 2 +- test/infinicore/ops/combinations.py | 2 +- test/infinicore/ops/conv1d.py | 2 +- test/infinicore/ops/conv2d.py | 2 +- test/infinicore/ops/conv3d.py | 2 +- test/infinicore/ops/conv_transpose1d.py | 2 +- test/infinicore/ops/conv_transpose2d.py | 2 +- test/infinicore/ops/conv_transpose3d.py | 2 +- test/infinicore/ops/corrcoef.py | 2 +- test/infinicore/ops/cosh.py | 2 +- test/infinicore/ops/cosine_embedding_loss.py | 2 +- test/infinicore/ops/cosine_similarity.py | 2 +- test/infinicore/ops/count_nonzero.py | 2 +- test/infinicore/ops/cov.py | 2 +- test/infinicore/ops/cross.py | 2 +- test/infinicore/ops/cross_entropy.py | 2 +- test/infinicore/ops/cummax.py | 2 +- test/infinicore/ops/cummin.py | 2 +- test/infinicore/ops/cumprod.py | 2 +- test/infinicore/ops/cumsum.py | 2 +- test/infinicore/ops/deg2rad.py | 2 +- test/infinicore/ops/det.py | 2 +- test/infinicore/ops/diag.py | 2 +- test/infinicore/ops/diag_embed.py | 2 +- test/infinicore/ops/diagflat.py | 2 +- test/infinicore/ops/diagonal.py | 2 +- test/infinicore/ops/diagonal_scatter.py | 2 +- test/infinicore/ops/diff.py | 2 +- test/infinicore/ops/digamma.py | 2 +- test/infinicore/ops/dist.py | 2 +- test/infinicore/ops/dot.py | 2 +- test/infinicore/ops/dropout1d.py | 2 +- test/infinicore/ops/dropout2d.py | 2 +- test/infinicore/ops/dropout3d.py | 2 +- test/infinicore/ops/elu.py | 2 +- test/infinicore/ops/embedding.py | 3 +-- test/infinicore/ops/empty.py | 2 +- test/infinicore/ops/empty_like.py | 2 +- test/infinicore/ops/empty_strided.py | 2 +- test/infinicore/ops/equal.py | 2 +- test/infinicore/ops/erf.py | 2 +- test/infinicore/ops/erfc.py | 2 +- test/infinicore/ops/erfinv.py | 2 +- test/infinicore/ops/exp2.py | 2 +- test/infinicore/ops/expm1.py | 2 +- test/infinicore/ops/feature_alpha_dropout.py | 2 +- test/infinicore/ops/flash_attention.py | 2 +- test/infinicore/ops/flip.py | 2 +- test/infinicore/ops/fliplr.py | 2 +- test/infinicore/ops/flipud.py | 2 +- test/infinicore/ops/float_power.py | 2 +- test/infinicore/ops/floor.py | 2 +- test/infinicore/ops/floor_divide.py | 2 +- test/infinicore/ops/fmax.py | 2 +- test/infinicore/ops/fmin.py | 2 +- test/infinicore/ops/fmod.py | 2 +- test/infinicore/ops/fold.py | 2 +- test/infinicore/ops/frac.py | 2 +- test/infinicore/ops/fractional_max_pool2d.py | 2 +- test/infinicore/ops/fractional_max_pool3d.py | 2 +- test/infinicore/ops/frexp.py | 2 +- test/infinicore/ops/full.py | 2 +- test/infinicore/ops/full_like.py | 2 +- test/infinicore/ops/gather.py | 2 +- test/infinicore/ops/gaussian_nll_loss.py | 2 +- test/infinicore/ops/gcd.py | 2 +- test/infinicore/ops/glu.py | 2 +- test/infinicore/ops/group_norm.py | 2 +- test/infinicore/ops/gt.py | 2 +- test/infinicore/ops/gumbel_softmax.py | 2 +- test/infinicore/ops/hardshrink.py | 2 +- test/infinicore/ops/hardsigmoid.py | 2 +- test/infinicore/ops/hardswish.py | 2 +- test/infinicore/ops/hardtanh.py | 2 +- test/infinicore/ops/heaviside.py | 2 +- test/infinicore/ops/hinge_embedding_loss.py | 2 +- test/infinicore/ops/histc.py | 2 +- test/infinicore/ops/histogram.py | 2 +- test/infinicore/ops/hsplit.py | 2 +- test/infinicore/ops/huber_loss.py | 2 +- test/infinicore/ops/hypot.py | 2 +- test/infinicore/ops/index_add.py | 2 +- test/infinicore/ops/index_copy.py | 2 +- test/infinicore/ops/index_reduce.py | 2 +- test/infinicore/ops/index_select.py | 2 +- test/infinicore/ops/inner.py | 2 +- test/infinicore/ops/instance_norm.py | 2 +- test/infinicore/ops/interpolate.py | 2 +- test/infinicore/ops/inverse.py | 2 +- test/infinicore/ops/isclose.py | 2 +- test/infinicore/ops/isfinite.py | 2 +- test/infinicore/ops/isin.py | 2 +- test/infinicore/ops/isneginf.py | 2 +- test/infinicore/ops/isposinf.py | 2 +- test/infinicore/ops/kl_div.py | 2 +- test/infinicore/ops/kron.py | 2 +- test/infinicore/ops/kthvalue.py | 2 +- test/infinicore/ops/kv_caching.py | 2 +- test/infinicore/ops/l1_loss.py | 2 +- test/infinicore/ops/ldexp.py | 2 +- test/infinicore/ops/leaky_relu.py | 2 +- test/infinicore/ops/lerp.py | 2 +- test/infinicore/ops/linear.py | 3 +-- test/infinicore/ops/local_response_norm.py | 2 +- test/infinicore/ops/log.py | 2 +- test/infinicore/ops/log10.py | 2 +- test/infinicore/ops/log1p.py | 2 +- test/infinicore/ops/log2.py | 2 +- test/infinicore/ops/log_softmax.py | 2 +- test/infinicore/ops/logaddexp.py | 2 +- test/infinicore/ops/logaddexp2.py | 2 +- test/infinicore/ops/logcumsumexp.py | 2 +- test/infinicore/ops/logdet.py | 2 +- test/infinicore/ops/logical_and.py | 2 +- test/infinicore/ops/logical_not.py | 2 +- test/infinicore/ops/logical_or.py | 2 +- test/infinicore/ops/logical_xor.py | 2 +- test/infinicore/ops/logsigmoid.py | 2 +- test/infinicore/ops/logsumexp.py | 2 +- test/infinicore/ops/lp_pool1d.py | 2 +- test/infinicore/ops/lp_pool2d.py | 2 +- test/infinicore/ops/lp_pool3d.py | 2 +- test/infinicore/ops/margin_ranking_loss.py | 2 +- test/infinicore/ops/masked_select.py | 2 +- test/infinicore/ops/matmul.py | 2 +- test/infinicore/ops/matrix_exp.py | 2 +- test/infinicore/ops/matrix_power.py | 2 +- test/infinicore/ops/max.py | 2 +- test/infinicore/ops/max_pool1d.py | 2 +- test/infinicore/ops/max_pool2d.py | 2 +- test/infinicore/ops/max_pool3d.py | 2 +- test/infinicore/ops/max_unpool1d.py | 2 +- test/infinicore/ops/max_unpool2d.py | 2 +- test/infinicore/ops/max_unpool3d.py | 2 +- test/infinicore/ops/maximum.py | 2 +- test/infinicore/ops/mean.py | 2 +- test/infinicore/ops/median.py | 2 +- test/infinicore/ops/min.py | 2 +- test/infinicore/ops/minimum.py | 2 +- test/infinicore/ops/mish.py | 2 +- test/infinicore/ops/mse_loss.py | 2 +- test/infinicore/ops/msort.py | 2 +- test/infinicore/ops/mul.py | 2 +- test/infinicore/ops/multi_margin_loss.py | 2 +- test/infinicore/ops/multilabel_margin_loss.py | 2 +- test/infinicore/ops/multilabel_soft_margin_loss.py | 2 +- test/infinicore/ops/mv.py | 2 +- test/infinicore/ops/narrow.py | 2 +- test/infinicore/ops/narrow_copy.py | 2 +- test/infinicore/ops/nll_loss.py | 2 +- test/infinicore/ops/nonzero.py | 2 +- test/infinicore/ops/normalize.py | 2 +- test/infinicore/ops/one_hot.py | 2 +- test/infinicore/ops/pad.py | 2 +- test/infinicore/ops/paged_attention.py | 2 +- test/infinicore/ops/paged_attention_prefill.py | 3 +-- test/infinicore/ops/paged_caching.py | 2 +- test/infinicore/ops/pairwise_distance.py | 2 +- test/infinicore/ops/pdist.py | 2 +- test/infinicore/ops/pixel_shuffle.py | 2 +- test/infinicore/ops/pixel_unshuffle.py | 2 +- test/infinicore/ops/prelu.py | 2 +- test/infinicore/ops/prod.py | 2 +- test/infinicore/ops/quantile.py | 2 +- test/infinicore/ops/random_sample.py | 2 +- test/infinicore/ops/reciprocal.py | 2 +- test/infinicore/ops/relu6.py | 2 +- test/infinicore/ops/remainder.py | 2 +- test/infinicore/ops/reshape.py | 2 +- test/infinicore/ops/rms_norm.py | 2 +- test/infinicore/ops/rope.py | 6 ++---- test/infinicore/ops/rot90.py | 2 +- test/infinicore/ops/round.py | 2 +- test/infinicore/ops/rrelu.py | 2 +- test/infinicore/ops/scaled_dot_product_attention.py | 2 +- test/infinicore/ops/scatter.py | 2 +- test/infinicore/ops/scatter_add.py | 2 +- test/infinicore/ops/scatter_reduce.py | 2 +- test/infinicore/ops/select_op.py | 2 +- test/infinicore/ops/select_scatter.py | 2 +- test/infinicore/ops/selu.py | 2 +- test/infinicore/ops/sgn.py | 2 +- test/infinicore/ops/sign.py | 2 +- test/infinicore/ops/signbit.py | 2 +- test/infinicore/ops/silu.py | 2 +- test/infinicore/ops/sinh.py | 2 +- test/infinicore/ops/slice_scatter.py | 2 +- test/infinicore/ops/slogdet.py | 2 +- test/infinicore/ops/smooth_l1_loss.py | 2 +- test/infinicore/ops/soft_margin_loss.py | 2 +- test/infinicore/ops/softmin.py | 2 +- test/infinicore/ops/softplus.py | 2 +- test/infinicore/ops/softshrink.py | 2 +- test/infinicore/ops/softsign.py | 2 +- test/infinicore/ops/sort.py | 2 +- test/infinicore/ops/split.py | 2 +- test/infinicore/ops/sqrt.py | 2 +- test/infinicore/ops/square.py | 2 +- test/infinicore/ops/stack.py | 2 +- test/infinicore/ops/std.py | 2 +- test/infinicore/ops/std_mean.py | 2 +- test/infinicore/ops/sum.py | 2 +- test/infinicore/ops/svd.py | 2 +- test/infinicore/ops/swiglu.py | 2 +- test/infinicore/ops/take.py | 2 +- test/infinicore/ops/tan.py | 2 +- test/infinicore/ops/tanhshrink.py | 2 +- test/infinicore/ops/threshold.py | 2 +- test/infinicore/ops/topk.py | 2 +- test/infinicore/ops/transpose.py | 2 +- test/infinicore/ops/triplet_margin_loss.py | 2 +- .../ops/triplet_margin_with_distance_loss.py | 2 +- test/infinicore/ops/trunc.py | 2 +- test/infinicore/ops/unfold.py | 2 +- test/infinicore/ops/unique.py | 2 +- test/infinicore/ops/upsample.py | 2 +- test/infinicore/ops/upsample_bilinear.py | 2 +- test/infinicore/ops/upsample_nearest.py | 2 +- test/infinicore/ops/vander.py | 2 +- test/infinicore/ops/var.py | 2 +- test/infinicore/ops/var_mean.py | 2 +- test/infinicore/ops/vdot.py | 2 +- test/infinicore/ops/where.py | 2 +- test/infiniop/gelu.py | 10 ++++++++++ 274 files changed, 284 insertions(+), 279 deletions(-) diff --git a/test/infinicore/ops/abs.py b/test/infinicore/ops/abs.py index ca90dd421..4e7fc3bbb 100644 --- a/test/infinicore/ops/abs.py +++ b/test/infinicore/ops/abs.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (in_shape, in_strides_or_None) diff --git a/test/infinicore/ops/acos.py b/test/infinicore/ops/acos.py index 87732413c..9babb5f9b 100644 --- a/test/infinicore/ops/acos.py +++ b/test/infinicore/ops/acos.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/acosh.py b/test/infinicore/ops/acosh.py index 359adf0a7..665fdf4d4 100644 --- a/test/infinicore/ops/acosh.py +++ b/test/infinicore/ops/acosh.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/adaptive_avg_pool1d.py b/test/infinicore/ops/adaptive_avg_pool1d.py index 05382724d..999f47062 100644 --- a/test/infinicore/ops/adaptive_avg_pool1d.py +++ b/test/infinicore/ops/adaptive_avg_pool1d.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/adaptive_avg_pool2d.py b/test/infinicore/ops/adaptive_avg_pool2d.py index 5a3d725e5..b1ecb5c58 100644 --- a/test/infinicore/ops/adaptive_avg_pool2d.py +++ b/test/infinicore/ops/adaptive_avg_pool2d.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/adaptive_avg_pool3d.py b/test/infinicore/ops/adaptive_avg_pool3d.py index 5566f3a72..37a0b7998 100644 --- a/test/infinicore/ops/adaptive_avg_pool3d.py +++ b/test/infinicore/ops/adaptive_avg_pool3d.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/adaptive_max_pool1d.py b/test/infinicore/ops/adaptive_max_pool1d.py index 0e683b4f1..00fb332fd 100644 --- a/test/infinicore/ops/adaptive_max_pool1d.py +++ b/test/infinicore/ops/adaptive_max_pool1d.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/adaptive_max_pool2d.py b/test/infinicore/ops/adaptive_max_pool2d.py index f8782d724..6ef12d184 100644 --- a/test/infinicore/ops/adaptive_max_pool2d.py +++ b/test/infinicore/ops/adaptive_max_pool2d.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, CaseResult, diff --git a/test/infinicore/ops/add.py b/test/infinicore/ops/add.py index add647be2..bbfeb0e7f 100644 --- a/test/infinicore/ops/add.py +++ b/test/infinicore/ops/add.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/add_rms_norm.py b/test/infinicore/ops/add_rms_norm.py index f6bf165a9..d629ec81e 100644 --- a/test/infinicore/ops/add_rms_norm.py +++ b/test/infinicore/ops/add_rms_norm.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/addbmm.py b/test/infinicore/ops/addbmm.py index ba37a1ac1..e14899ed7 100644 --- a/test/infinicore/ops/addbmm.py +++ b/test/infinicore/ops/addbmm.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/addcdiv.py b/test/infinicore/ops/addcdiv.py index 47d574eb1..0ade93d33 100644 --- a/test/infinicore/ops/addcdiv.py +++ b/test/infinicore/ops/addcdiv.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/addcmul.py b/test/infinicore/ops/addcmul.py index 143154f84..4032fe9c3 100644 --- a/test/infinicore/ops/addcmul.py +++ b/test/infinicore/ops/addcmul.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/addmv.py b/test/infinicore/ops/addmv.py index c06499438..78c4d419f 100644 --- a/test/infinicore/ops/addmv.py +++ b/test/infinicore/ops/addmv.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/addr.py b/test/infinicore/ops/addr.py index e1f279349..51b172538 100644 --- a/test/infinicore/ops/addr.py +++ b/test/infinicore/ops/addr.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/affine_grid.py b/test/infinicore/ops/affine_grid.py index 7e782ea6a..3cf732e22 100644 --- a/test/infinicore/ops/affine_grid.py +++ b/test/infinicore/ops/affine_grid.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/all.py b/test/infinicore/ops/all.py index bef8ba48b..cded59ee8 100644 --- a/test/infinicore/ops/all.py +++ b/test/infinicore/ops/all.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/alpha_dropout.py b/test/infinicore/ops/alpha_dropout.py index e55ce5abb..cfb8a3dbf 100644 --- a/test/infinicore/ops/alpha_dropout.py +++ b/test/infinicore/ops/alpha_dropout.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/amax.py b/test/infinicore/ops/amax.py index 442cb2b4e..64ab48533 100644 --- a/test/infinicore/ops/amax.py +++ b/test/infinicore/ops/amax.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/amin.py b/test/infinicore/ops/amin.py index a4272f098..85af5bec1 100644 --- a/test/infinicore/ops/amin.py +++ b/test/infinicore/ops/amin.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/aminmax.py b/test/infinicore/ops/aminmax.py index 9f2d43cd3..73408a6b3 100644 --- a/test/infinicore/ops/aminmax.py +++ b/test/infinicore/ops/aminmax.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/any.py b/test/infinicore/ops/any.py index c3996bc5f..6fa1a3b9e 100644 --- a/test/infinicore/ops/any.py +++ b/test/infinicore/ops/any.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/argmax.py b/test/infinicore/ops/argmax.py index 9e078775a..fb547211b 100644 --- a/test/infinicore/ops/argmax.py +++ b/test/infinicore/ops/argmax.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/argmin.py b/test/infinicore/ops/argmin.py index 4b1753a02..c431ac90a 100644 --- a/test/infinicore/ops/argmin.py +++ b/test/infinicore/ops/argmin.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/argsort.py b/test/infinicore/ops/argsort.py index 35312a235..b3bb183b8 100644 --- a/test/infinicore/ops/argsort.py +++ b/test/infinicore/ops/argsort.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/argwhere.py b/test/infinicore/ops/argwhere.py index 83fa1e6a5..7f2326a29 100644 --- a/test/infinicore/ops/argwhere.py +++ b/test/infinicore/ops/argwhere.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, input_strides_or_None) diff --git a/test/infinicore/ops/asin.py b/test/infinicore/ops/asin.py index 168c1e209..1f4d70ea3 100644 --- a/test/infinicore/ops/asin.py +++ b/test/infinicore/ops/asin.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/asinh.py b/test/infinicore/ops/asinh.py index 97bcd5edb..79452d336 100644 --- a/test/infinicore/ops/asinh.py +++ b/test/infinicore/ops/asinh.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/atan.py b/test/infinicore/ops/atan.py index 966c982ab..0f676c9af 100644 --- a/test/infinicore/ops/atan.py +++ b/test/infinicore/ops/atan.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/atan2.py b/test/infinicore/ops/atan2.py index 09f15a319..26ed2e6c1 100644 --- a/test/infinicore/ops/atan2.py +++ b/test/infinicore/ops/atan2.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/atanh.py b/test/infinicore/ops/atanh.py index 4f8cab954..dbef3920d 100644 --- a/test/infinicore/ops/atanh.py +++ b/test/infinicore/ops/atanh.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/avg_pool1d.py b/test/infinicore/ops/avg_pool1d.py index d7f396aea..5a0318571 100644 --- a/test/infinicore/ops/avg_pool1d.py +++ b/test/infinicore/ops/avg_pool1d.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/avg_pool2d.py b/test/infinicore/ops/avg_pool2d.py index 2653e153e..ee3df7496 100644 --- a/test/infinicore/ops/avg_pool2d.py +++ b/test/infinicore/ops/avg_pool2d.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/avg_pool3d.py b/test/infinicore/ops/avg_pool3d.py index adb356227..bb7b60851 100644 --- a/test/infinicore/ops/avg_pool3d.py +++ b/test/infinicore/ops/avg_pool3d.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/baddbmm.py b/test/infinicore/ops/baddbmm.py index 35b4cd625..61bb97864 100644 --- a/test/infinicore/ops/baddbmm.py +++ b/test/infinicore/ops/baddbmm.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/batch_norm.py b/test/infinicore/ops/batch_norm.py index bc5052679..9f9e6a8c9 100644 --- a/test/infinicore/ops/batch_norm.py +++ b/test/infinicore/ops/batch_norm.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, input_strides_or_None, running_mean_present_bool, running_var_present_bool, weight_bias_present_bool, training_or_None, momentum_or_None, eps_or_None) diff --git a/test/infinicore/ops/bilinear.py b/test/infinicore/ops/bilinear.py index da31e0b74..f0c606399 100644 --- a/test/infinicore/ops/bilinear.py +++ b/test/infinicore/ops/bilinear.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (in1_shape, in2_shape, weight_shape, in1_strides_or_None, in2_strides_or_None, weight_strides_or_None, bias_present_bool) diff --git a/test/infinicore/ops/binary_cross_entropy.py b/test/infinicore/ops/binary_cross_entropy.py index cca511f38..c214957c3 100644 --- a/test/infinicore/ops/binary_cross_entropy.py +++ b/test/infinicore/ops/binary_cross_entropy.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, input_strides_or_None, weight_present_bool, reduction_or_None) diff --git a/test/infinicore/ops/binary_cross_entropy_with_logits.py b/test/infinicore/ops/binary_cross_entropy_with_logits.py index 986605a38..d242deb5d 100644 --- a/test/infinicore/ops/binary_cross_entropy_with_logits.py +++ b/test/infinicore/ops/binary_cross_entropy_with_logits.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, input_strides_or_None, weight_present_bool, pos_weight_present_bool, reduction_or_None) diff --git a/test/infinicore/ops/bincount.py b/test/infinicore/ops/bincount.py index cbbbde722..446eee547 100644 --- a/test/infinicore/ops/bincount.py +++ b/test/infinicore/ops/bincount.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner from framework.tensor import TensorInitializer diff --git a/test/infinicore/ops/bitwise_left_shift.py b/test/infinicore/ops/bitwise_left_shift.py index 3b0b503e3..8d907ccf5 100644 --- a/test/infinicore/ops/bitwise_left_shift.py +++ b/test/infinicore/ops/bitwise_left_shift.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/bitwise_right_shift.py b/test/infinicore/ops/bitwise_right_shift.py index 23ed8504d..84197c7c3 100644 --- a/test/infinicore/ops/bitwise_right_shift.py +++ b/test/infinicore/ops/bitwise_right_shift.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/bitwise_xor.py b/test/infinicore/ops/bitwise_xor.py index 8eadd6fc5..077297d3b 100644 --- a/test/infinicore/ops/bitwise_xor.py +++ b/test/infinicore/ops/bitwise_xor.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/block_diag.py b/test/infinicore/ops/block_diag.py index b65e5c5ee..9c0828642 100644 --- a/test/infinicore/ops/block_diag.py +++ b/test/infinicore/ops/block_diag.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (list_of_matrix_shapes, list_of_strides_or_None, dtype) diff --git a/test/infinicore/ops/broadcast_to.py b/test/infinicore/ops/broadcast_to.py index 0813cf14f..0c5f78f80 100644 --- a/test/infinicore/ops/broadcast_to.py +++ b/test/infinicore/ops/broadcast_to.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/bucketize.py b/test/infinicore/ops/bucketize.py index fe6ccb6da..e1c6fc75a 100644 --- a/test/infinicore/ops/bucketize.py +++ b/test/infinicore/ops/bucketize.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, input_strides_or_None, boundaries_len) diff --git a/test/infinicore/ops/cat.py b/test/infinicore/ops/cat.py index 669fbfd1b..e66f29fa4 100644 --- a/test/infinicore/ops/cat.py +++ b/test/infinicore/ops/cat.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/causal_softmax.py b/test/infinicore/ops/causal_softmax.py index cc2a60574..11df922bd 100644 --- a/test/infinicore/ops/causal_softmax.py +++ b/test/infinicore/ops/causal_softmax.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/cdist.py b/test/infinicore/ops/cdist.py index 2d41668cc..3657022f0 100644 --- a/test/infinicore/ops/cdist.py +++ b/test/infinicore/ops/cdist.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (x1_shape, x2_shape, x1_strides_or_None, x2_strides_or_None, p_or_None) diff --git a/test/infinicore/ops/celu.py b/test/infinicore/ops/celu.py index 88b52bb7e..52a8b83ff 100644 --- a/test/infinicore/ops/celu.py +++ b/test/infinicore/ops/celu.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/clone.py b/test/infinicore/ops/clone.py index 89e329999..6b09b45df 100644 --- a/test/infinicore/ops/clone.py +++ b/test/infinicore/ops/clone.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (shape, input_strides_or_None) diff --git a/test/infinicore/ops/combinations.py b/test/infinicore/ops/combinations.py index ca603c6d1..e3529dfde 100644 --- a/test/infinicore/ops/combinations.py +++ b/test/infinicore/ops/combinations.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, input_strides_or_None, r) diff --git a/test/infinicore/ops/conv1d.py b/test/infinicore/ops/conv1d.py index c1904a882..0d4e22895 100644 --- a/test/infinicore/ops/conv1d.py +++ b/test/infinicore/ops/conv1d.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/conv2d.py b/test/infinicore/ops/conv2d.py index 1772ab4cf..1be7cca75 100644 --- a/test/infinicore/ops/conv2d.py +++ b/test/infinicore/ops/conv2d.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/conv3d.py b/test/infinicore/ops/conv3d.py index 07748007b..39a42dc1b 100644 --- a/test/infinicore/ops/conv3d.py +++ b/test/infinicore/ops/conv3d.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/conv_transpose1d.py b/test/infinicore/ops/conv_transpose1d.py index 5f55e802d..b7dafd4f2 100644 --- a/test/infinicore/ops/conv_transpose1d.py +++ b/test/infinicore/ops/conv_transpose1d.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/conv_transpose2d.py b/test/infinicore/ops/conv_transpose2d.py index 68f572ddc..1009f7160 100644 --- a/test/infinicore/ops/conv_transpose2d.py +++ b/test/infinicore/ops/conv_transpose2d.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/conv_transpose3d.py b/test/infinicore/ops/conv_transpose3d.py index 988af3e57..eca3addb4 100644 --- a/test/infinicore/ops/conv_transpose3d.py +++ b/test/infinicore/ops/conv_transpose3d.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/corrcoef.py b/test/infinicore/ops/corrcoef.py index f615448bc..1b14b09d2 100644 --- a/test/infinicore/ops/corrcoef.py +++ b/test/infinicore/ops/corrcoef.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, input_strides_or_None) diff --git a/test/infinicore/ops/cosh.py b/test/infinicore/ops/cosh.py index a4fb92a73..16d3c3d7b 100644 --- a/test/infinicore/ops/cosh.py +++ b/test/infinicore/ops/cosh.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/cosine_embedding_loss.py b/test/infinicore/ops/cosine_embedding_loss.py index 0ac5ff7bb..d01b33b57 100644 --- a/test/infinicore/ops/cosine_embedding_loss.py +++ b/test/infinicore/ops/cosine_embedding_loss.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input1_shape, input2_shape, target_shape, input1_strides_or_None, input2_strides_or_None, target_strides_or_None, margin_or_None) diff --git a/test/infinicore/ops/cosine_similarity.py b/test/infinicore/ops/cosine_similarity.py index 4831de9e9..754b93e12 100644 --- a/test/infinicore/ops/cosine_similarity.py +++ b/test/infinicore/ops/cosine_similarity.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/count_nonzero.py b/test/infinicore/ops/count_nonzero.py index 64c2879c7..12bdf8fad 100644 --- a/test/infinicore/ops/count_nonzero.py +++ b/test/infinicore/ops/count_nonzero.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/cov.py b/test/infinicore/ops/cov.py index 1af7ba0b9..1087e5b12 100644 --- a/test/infinicore/ops/cov.py +++ b/test/infinicore/ops/cov.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, input_strides_or_None, correction, fweights, aweights) diff --git a/test/infinicore/ops/cross.py b/test/infinicore/ops/cross.py index c6b78a55d..fa9a571c1 100644 --- a/test/infinicore/ops/cross.py +++ b/test/infinicore/ops/cross.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/cross_entropy.py b/test/infinicore/ops/cross_entropy.py index 21e5cb987..e71a30567 100644 --- a/test/infinicore/ops/cross_entropy.py +++ b/test/infinicore/ops/cross_entropy.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner from framework.tensor import TensorInitializer diff --git a/test/infinicore/ops/cummax.py b/test/infinicore/ops/cummax.py index ef10ac3ec..5dbae8fad 100644 --- a/test/infinicore/ops/cummax.py +++ b/test/infinicore/ops/cummax.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/cummin.py b/test/infinicore/ops/cummin.py index fc8075b3b..b448c7370 100644 --- a/test/infinicore/ops/cummin.py +++ b/test/infinicore/ops/cummin.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/cumprod.py b/test/infinicore/ops/cumprod.py index e4fed125a..2902d89f7 100644 --- a/test/infinicore/ops/cumprod.py +++ b/test/infinicore/ops/cumprod.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/cumsum.py b/test/infinicore/ops/cumsum.py index f381b327a..514e2ba85 100644 --- a/test/infinicore/ops/cumsum.py +++ b/test/infinicore/ops/cumsum.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/deg2rad.py b/test/infinicore/ops/deg2rad.py index a1509a41a..b454742c8 100644 --- a/test/infinicore/ops/deg2rad.py +++ b/test/infinicore/ops/deg2rad.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/det.py b/test/infinicore/ops/det.py index 9ddbbdde7..374e15e3a 100644 --- a/test/infinicore/ops/det.py +++ b/test/infinicore/ops/det.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (matrix_shape, strides_or_None) diff --git a/test/infinicore/ops/diag.py b/test/infinicore/ops/diag.py index 320987a3b..f5c52c824 100644 --- a/test/infinicore/ops/diag.py +++ b/test/infinicore/ops/diag.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, input_strides_or_None, diagonal_k_or_None) diff --git a/test/infinicore/ops/diag_embed.py b/test/infinicore/ops/diag_embed.py index f26d571a4..3fa662395 100644 --- a/test/infinicore/ops/diag_embed.py +++ b/test/infinicore/ops/diag_embed.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, input_strides_or_None, offset_or_None) diff --git a/test/infinicore/ops/diagflat.py b/test/infinicore/ops/diagflat.py index 37a17d119..b16a62630 100644 --- a/test/infinicore/ops/diagflat.py +++ b/test/infinicore/ops/diagflat.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, input_strides_or_None, offset_or_None) diff --git a/test/infinicore/ops/diagonal.py b/test/infinicore/ops/diagonal.py index 1c6add36c..598d5a1f7 100644 --- a/test/infinicore/ops/diagonal.py +++ b/test/infinicore/ops/diagonal.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, input_strides_or_None, offset_or_None, dim1_or_None, dim2_or_None) diff --git a/test/infinicore/ops/diagonal_scatter.py b/test/infinicore/ops/diagonal_scatter.py index 598ef6671..33547f697 100644 --- a/test/infinicore/ops/diagonal_scatter.py +++ b/test/infinicore/ops/diagonal_scatter.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/diff.py b/test/infinicore/ops/diff.py index 7a9c8b97f..7f33e402c 100644 --- a/test/infinicore/ops/diff.py +++ b/test/infinicore/ops/diff.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/digamma.py b/test/infinicore/ops/digamma.py index caa197360..52d3070d7 100644 --- a/test/infinicore/ops/digamma.py +++ b/test/infinicore/ops/digamma.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/dist.py b/test/infinicore/ops/dist.py index 6eae582c9..3008801f1 100644 --- a/test/infinicore/ops/dist.py +++ b/test/infinicore/ops/dist.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (shape, a_strides_or_None, b_strides_or_None, p_or_None) diff --git a/test/infinicore/ops/dot.py b/test/infinicore/ops/dot.py index 5d2300d24..cf0bee10d 100644 --- a/test/infinicore/ops/dot.py +++ b/test/infinicore/ops/dot.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (vec1_shape, vec2_shape, vec1_strides_or_None, vec2_strides_or_None) diff --git a/test/infinicore/ops/dropout1d.py b/test/infinicore/ops/dropout1d.py index d292d9edc..c4a21f945 100644 --- a/test/infinicore/ops/dropout1d.py +++ b/test/infinicore/ops/dropout1d.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/dropout2d.py b/test/infinicore/ops/dropout2d.py index 226f5c4e3..41f2243d6 100644 --- a/test/infinicore/ops/dropout2d.py +++ b/test/infinicore/ops/dropout2d.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/dropout3d.py b/test/infinicore/ops/dropout3d.py index f749b2ebc..dad06c789 100644 --- a/test/infinicore/ops/dropout3d.py +++ b/test/infinicore/ops/dropout3d.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/elu.py b/test/infinicore/ops/elu.py index 3c5416cc5..b40fb4dc5 100644 --- a/test/infinicore/ops/elu.py +++ b/test/infinicore/ops/elu.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/embedding.py b/test/infinicore/ops/embedding.py index 6cb7755af..c7a1007ff 100644 --- a/test/infinicore/ops/embedding.py +++ b/test/infinicore/ops/embedding.py @@ -3,6 +3,7 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) +import infinicore import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner from framework.tensor import TensorInitializer @@ -12,8 +13,6 @@ to_torch_dtype, ) -import infinicore - # ============================================================================== # Operator-specific configuration # ============================================================================== diff --git a/test/infinicore/ops/empty.py b/test/infinicore/ops/empty.py index b333d76e7..ed7f9f326 100644 --- a/test/infinicore/ops/empty.py +++ b/test/infinicore/ops/empty.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner from framework.datatypes import to_torch_dtype diff --git a/test/infinicore/ops/empty_like.py b/test/infinicore/ops/empty_like.py index bf6989490..c1d16fd4e 100644 --- a/test/infinicore/ops/empty_like.py +++ b/test/infinicore/ops/empty_like.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner from framework.datatypes import to_torch_dtype diff --git a/test/infinicore/ops/empty_strided.py b/test/infinicore/ops/empty_strided.py index 0c3911255..aa3f172cb 100644 --- a/test/infinicore/ops/empty_strided.py +++ b/test/infinicore/ops/empty_strided.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner from framework.datatypes import to_torch_dtype diff --git a/test/infinicore/ops/equal.py b/test/infinicore/ops/equal.py index 126439cb6..10aae3fcb 100644 --- a/test/infinicore/ops/equal.py +++ b/test/infinicore/ops/equal.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/erf.py b/test/infinicore/ops/erf.py index 5334e27bf..e40f3a4f7 100644 --- a/test/infinicore/ops/erf.py +++ b/test/infinicore/ops/erf.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/erfc.py b/test/infinicore/ops/erfc.py index ed5fcef9e..0e513d0f1 100644 --- a/test/infinicore/ops/erfc.py +++ b/test/infinicore/ops/erfc.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/erfinv.py b/test/infinicore/ops/erfinv.py index 3a29a2dd8..c6071dd1b 100644 --- a/test/infinicore/ops/erfinv.py +++ b/test/infinicore/ops/erfinv.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/exp2.py b/test/infinicore/ops/exp2.py index 96d6bd11f..95b175dc6 100644 --- a/test/infinicore/ops/exp2.py +++ b/test/infinicore/ops/exp2.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/expm1.py b/test/infinicore/ops/expm1.py index 72f6b1ba1..4501d3e7e 100644 --- a/test/infinicore/ops/expm1.py +++ b/test/infinicore/ops/expm1.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/feature_alpha_dropout.py b/test/infinicore/ops/feature_alpha_dropout.py index 81a9798f5..0fb019b83 100644 --- a/test/infinicore/ops/feature_alpha_dropout.py +++ b/test/infinicore/ops/feature_alpha_dropout.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/flash_attention.py b/test/infinicore/ops/flash_attention.py index 2d4b09599..a4181c6e8 100644 --- a/test/infinicore/ops/flash_attention.py +++ b/test/infinicore/ops/flash_attention.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/flip.py b/test/infinicore/ops/flip.py index 2bea23274..04b6e9e63 100644 --- a/test/infinicore/ops/flip.py +++ b/test/infinicore/ops/flip.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/fliplr.py b/test/infinicore/ops/fliplr.py index d832c881a..43a40e3e2 100644 --- a/test/infinicore/ops/fliplr.py +++ b/test/infinicore/ops/fliplr.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/flipud.py b/test/infinicore/ops/flipud.py index ee7f22ad7..b92762f49 100644 --- a/test/infinicore/ops/flipud.py +++ b/test/infinicore/ops/flipud.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/float_power.py b/test/infinicore/ops/float_power.py index 6df07b5af..2548c57ac 100644 --- a/test/infinicore/ops/float_power.py +++ b/test/infinicore/ops/float_power.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/floor.py b/test/infinicore/ops/floor.py index c7663d39b..708636b78 100644 --- a/test/infinicore/ops/floor.py +++ b/test/infinicore/ops/floor.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/floor_divide.py b/test/infinicore/ops/floor_divide.py index 28e3dc77c..9aaebfe62 100644 --- a/test/infinicore/ops/floor_divide.py +++ b/test/infinicore/ops/floor_divide.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/fmax.py b/test/infinicore/ops/fmax.py index 046710f35..43a603417 100644 --- a/test/infinicore/ops/fmax.py +++ b/test/infinicore/ops/fmax.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/fmin.py b/test/infinicore/ops/fmin.py index 280d6bf4e..49ea01447 100644 --- a/test/infinicore/ops/fmin.py +++ b/test/infinicore/ops/fmin.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/fmod.py b/test/infinicore/ops/fmod.py index 6ef862154..66bdee38d 100644 --- a/test/infinicore/ops/fmod.py +++ b/test/infinicore/ops/fmod.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/fold.py b/test/infinicore/ops/fold.py index f9fb6f99c..ee145938a 100644 --- a/test/infinicore/ops/fold.py +++ b/test/infinicore/ops/fold.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (in_shape, in_strides_or_None, output_size, kernel_size, dilation, padding, stride) diff --git a/test/infinicore/ops/frac.py b/test/infinicore/ops/frac.py index 8d3cf7bed..40a203c65 100644 --- a/test/infinicore/ops/frac.py +++ b/test/infinicore/ops/frac.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/fractional_max_pool2d.py b/test/infinicore/ops/fractional_max_pool2d.py index 89849730b..09a04e8c2 100644 --- a/test/infinicore/ops/fractional_max_pool2d.py +++ b/test/infinicore/ops/fractional_max_pool2d.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (in_shape, in_strides_or_None, kernel_size, output_size_or_None, return_indices) diff --git a/test/infinicore/ops/fractional_max_pool3d.py b/test/infinicore/ops/fractional_max_pool3d.py index 45d061a24..918874623 100644 --- a/test/infinicore/ops/fractional_max_pool3d.py +++ b/test/infinicore/ops/fractional_max_pool3d.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (in_shape, in_strides_or_None, kernel_size, output_size_or_None, return_indices) diff --git a/test/infinicore/ops/frexp.py b/test/infinicore/ops/frexp.py index ac8a9c23a..9a2eb7d1b 100644 --- a/test/infinicore/ops/frexp.py +++ b/test/infinicore/ops/frexp.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (in_shape, in_strides_or_None) diff --git a/test/infinicore/ops/full.py b/test/infinicore/ops/full.py index 44f686bd9..6a8be9d6f 100644 --- a/test/infinicore/ops/full.py +++ b/test/infinicore/ops/full.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner from framework.datatypes import to_torch_dtype diff --git a/test/infinicore/ops/full_like.py b/test/infinicore/ops/full_like.py index db2aedf89..579dfd360 100644 --- a/test/infinicore/ops/full_like.py +++ b/test/infinicore/ops/full_like.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner from framework.datatypes import to_torch_dtype diff --git a/test/infinicore/ops/gather.py b/test/infinicore/ops/gather.py index 06de91c1e..08f2e512b 100644 --- a/test/infinicore/ops/gather.py +++ b/test/infinicore/ops/gather.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/gaussian_nll_loss.py b/test/infinicore/ops/gaussian_nll_loss.py index 8e31065c3..f81cf2c99 100644 --- a/test/infinicore/ops/gaussian_nll_loss.py +++ b/test/infinicore/ops/gaussian_nll_loss.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, var_present_bool, full_or_None, eps_or_None, input_strides_or_None) diff --git a/test/infinicore/ops/gcd.py b/test/infinicore/ops/gcd.py index 9484511f8..150bf22a8 100644 --- a/test/infinicore/ops/gcd.py +++ b/test/infinicore/ops/gcd.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/glu.py b/test/infinicore/ops/glu.py index f53f3097d..9e2faa0e8 100644 --- a/test/infinicore/ops/glu.py +++ b/test/infinicore/ops/glu.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/group_norm.py b/test/infinicore/ops/group_norm.py index 969c43c02..2da3bc150 100644 --- a/test/infinicore/ops/group_norm.py +++ b/test/infinicore/ops/group_norm.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, input_strides_or_None, num_groups, weight_bias_present_bool, eps_or_None) diff --git a/test/infinicore/ops/gt.py b/test/infinicore/ops/gt.py index b4809493e..39689174a 100644 --- a/test/infinicore/ops/gt.py +++ b/test/infinicore/ops/gt.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/gumbel_softmax.py b/test/infinicore/ops/gumbel_softmax.py index 2c9bb27a9..cb7f74e9d 100644 --- a/test/infinicore/ops/gumbel_softmax.py +++ b/test/infinicore/ops/gumbel_softmax.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/hardshrink.py b/test/infinicore/ops/hardshrink.py index ccd3fbc35..8bf2389e1 100644 --- a/test/infinicore/ops/hardshrink.py +++ b/test/infinicore/ops/hardshrink.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/hardsigmoid.py b/test/infinicore/ops/hardsigmoid.py index bbe0411ab..51016dade 100644 --- a/test/infinicore/ops/hardsigmoid.py +++ b/test/infinicore/ops/hardsigmoid.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/hardswish.py b/test/infinicore/ops/hardswish.py index bbb47bdcd..9f31cdc62 100644 --- a/test/infinicore/ops/hardswish.py +++ b/test/infinicore/ops/hardswish.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/hardtanh.py b/test/infinicore/ops/hardtanh.py index fa2f0fd12..6861e464e 100644 --- a/test/infinicore/ops/hardtanh.py +++ b/test/infinicore/ops/hardtanh.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/heaviside.py b/test/infinicore/ops/heaviside.py index cbe17e242..14e075de2 100644 --- a/test/infinicore/ops/heaviside.py +++ b/test/infinicore/ops/heaviside.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/hinge_embedding_loss.py b/test/infinicore/ops/hinge_embedding_loss.py index bf8135a69..2d2817e9d 100644 --- a/test/infinicore/ops/hinge_embedding_loss.py +++ b/test/infinicore/ops/hinge_embedding_loss.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, target_shape, input_strides_or_None, target_strides_or_None, margin_or_None) diff --git a/test/infinicore/ops/histc.py b/test/infinicore/ops/histc.py index 26ba35986..5ae03ef7b 100644 --- a/test/infinicore/ops/histc.py +++ b/test/infinicore/ops/histc.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, input_strides_or_None, bins, min_val, max_val) diff --git a/test/infinicore/ops/histogram.py b/test/infinicore/ops/histogram.py index 9e379b183..bc3e7166f 100644 --- a/test/infinicore/ops/histogram.py +++ b/test/infinicore/ops/histogram.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, input_strides_or_None, bins_or_sequence, range_or_None) diff --git a/test/infinicore/ops/hsplit.py b/test/infinicore/ops/hsplit.py index b5a23fa25..8fde1c1ae 100644 --- a/test/infinicore/ops/hsplit.py +++ b/test/infinicore/ops/hsplit.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, input_strides_or_None, sections_or_None) diff --git a/test/infinicore/ops/huber_loss.py b/test/infinicore/ops/huber_loss.py index 592695993..f4c588f4e 100644 --- a/test/infinicore/ops/huber_loss.py +++ b/test/infinicore/ops/huber_loss.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, target_shape, input_strides_or_None, target_strides_or_None, delta_or_None) diff --git a/test/infinicore/ops/hypot.py b/test/infinicore/ops/hypot.py index 283214c0e..df8dbd2f4 100644 --- a/test/infinicore/ops/hypot.py +++ b/test/infinicore/ops/hypot.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/index_add.py b/test/infinicore/ops/index_add.py index ee136cf31..024cd89e1 100644 --- a/test/infinicore/ops/index_add.py +++ b/test/infinicore/ops/index_add.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/index_copy.py b/test/infinicore/ops/index_copy.py index f38403320..8050d0f0b 100644 --- a/test/infinicore/ops/index_copy.py +++ b/test/infinicore/ops/index_copy.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/index_reduce.py b/test/infinicore/ops/index_reduce.py index c6192350a..90b2a90d1 100644 --- a/test/infinicore/ops/index_reduce.py +++ b/test/infinicore/ops/index_reduce.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/index_select.py b/test/infinicore/ops/index_select.py index a28bcc223..145a42f3e 100644 --- a/test/infinicore/ops/index_select.py +++ b/test/infinicore/ops/index_select.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, input_strides_or_None, dim, index_shape) diff --git a/test/infinicore/ops/inner.py b/test/infinicore/ops/inner.py index 3f75a6b64..3766ca860 100644 --- a/test/infinicore/ops/inner.py +++ b/test/infinicore/ops/inner.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (a_shape, b_shape, a_strides_or_None, b_strides_or_None) diff --git a/test/infinicore/ops/instance_norm.py b/test/infinicore/ops/instance_norm.py index 02fdea744..e5871c98e 100644 --- a/test/infinicore/ops/instance_norm.py +++ b/test/infinicore/ops/instance_norm.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, input_strides_or_None, running_mean_present_bool, running_var_present_bool, weight_bias_present_bool, use_input_stats_or_None, momentum_or_None, eps_or_None) diff --git a/test/infinicore/ops/interpolate.py b/test/infinicore/ops/interpolate.py index 1d4236614..ebeb47ca3 100644 --- a/test/infinicore/ops/interpolate.py +++ b/test/infinicore/ops/interpolate.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/inverse.py b/test/infinicore/ops/inverse.py index 34bd48740..a7a80a9fc 100644 --- a/test/infinicore/ops/inverse.py +++ b/test/infinicore/ops/inverse.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (matrix_shape, strides_or_None) diff --git a/test/infinicore/ops/isclose.py b/test/infinicore/ops/isclose.py index 440bfc88b..c1c311ffc 100644 --- a/test/infinicore/ops/isclose.py +++ b/test/infinicore/ops/isclose.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/isfinite.py b/test/infinicore/ops/isfinite.py index 81b298d93..bb3bc8850 100644 --- a/test/infinicore/ops/isfinite.py +++ b/test/infinicore/ops/isfinite.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/isin.py b/test/infinicore/ops/isin.py index 4ad8abb8a..c2b4a6972 100644 --- a/test/infinicore/ops/isin.py +++ b/test/infinicore/ops/isin.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/isneginf.py b/test/infinicore/ops/isneginf.py index 446c918b4..34f93460b 100644 --- a/test/infinicore/ops/isneginf.py +++ b/test/infinicore/ops/isneginf.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/isposinf.py b/test/infinicore/ops/isposinf.py index c4d0ee11e..1e7a0f1a7 100644 --- a/test/infinicore/ops/isposinf.py +++ b/test/infinicore/ops/isposinf.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/kl_div.py b/test/infinicore/ops/kl_div.py index 8da6e5ff4..ec6f776fd 100644 --- a/test/infinicore/ops/kl_div.py +++ b/test/infinicore/ops/kl_div.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, input_strides_or_None, reduction_or_None, log_target_bool_or_None) diff --git a/test/infinicore/ops/kron.py b/test/infinicore/ops/kron.py index 4b7dee217..08ecafa0e 100644 --- a/test/infinicore/ops/kron.py +++ b/test/infinicore/ops/kron.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (a_shape, b_shape, a_strides_or_None, b_strides_or_None) diff --git a/test/infinicore/ops/kthvalue.py b/test/infinicore/ops/kthvalue.py index 893fd6f29..856eabd2f 100644 --- a/test/infinicore/ops/kthvalue.py +++ b/test/infinicore/ops/kthvalue.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/kv_caching.py b/test/infinicore/ops/kv_caching.py index 4ca857586..2598d02e0 100644 --- a/test/infinicore/ops/kv_caching.py +++ b/test/infinicore/ops/kv_caching.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/l1_loss.py b/test/infinicore/ops/l1_loss.py index 91ddebf7b..d7a4be467 100644 --- a/test/infinicore/ops/l1_loss.py +++ b/test/infinicore/ops/l1_loss.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, target_shape, input_strides_or_None, reduction_or_None) diff --git a/test/infinicore/ops/ldexp.py b/test/infinicore/ops/ldexp.py index 89559e512..3cda26e44 100644 --- a/test/infinicore/ops/ldexp.py +++ b/test/infinicore/ops/ldexp.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # ldexp(input, other) computes input * (2**other) diff --git a/test/infinicore/ops/leaky_relu.py b/test/infinicore/ops/leaky_relu.py index 4faaa3d86..d667c012f 100644 --- a/test/infinicore/ops/leaky_relu.py +++ b/test/infinicore/ops/leaky_relu.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/lerp.py b/test/infinicore/ops/lerp.py index 810858739..107d370f9 100644 --- a/test/infinicore/ops/lerp.py +++ b/test/infinicore/ops/lerp.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/linear.py b/test/infinicore/ops/linear.py index 90c402af0..5e9070bdb 100644 --- a/test/infinicore/ops/linear.py +++ b/test/infinicore/ops/linear.py @@ -3,6 +3,7 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) +import infinicore import torch from framework import ( BaseOperatorTest, @@ -12,8 +13,6 @@ is_broadcast, ) -import infinicore - # ============================================================================== # Operator-specific configuration # ============================================================================== diff --git a/test/infinicore/ops/local_response_norm.py b/test/infinicore/ops/local_response_norm.py index f6b62c49e..6981e110a 100644 --- a/test/infinicore/ops/local_response_norm.py +++ b/test/infinicore/ops/local_response_norm.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, input_strides_or_None, size, alpha_or_None, beta_or_None, k_or_None) diff --git a/test/infinicore/ops/log.py b/test/infinicore/ops/log.py index 98ceb8a1c..82ed5ba5c 100644 --- a/test/infinicore/ops/log.py +++ b/test/infinicore/ops/log.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/log10.py b/test/infinicore/ops/log10.py index fbb0863e3..75f5bde3b 100644 --- a/test/infinicore/ops/log10.py +++ b/test/infinicore/ops/log10.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/log1p.py b/test/infinicore/ops/log1p.py index 32996927f..7b9e6609b 100644 --- a/test/infinicore/ops/log1p.py +++ b/test/infinicore/ops/log1p.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/log2.py b/test/infinicore/ops/log2.py index 6c4ebd740..68037c681 100644 --- a/test/infinicore/ops/log2.py +++ b/test/infinicore/ops/log2.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/log_softmax.py b/test/infinicore/ops/log_softmax.py index 68af20f1e..795adfe7b 100644 --- a/test/infinicore/ops/log_softmax.py +++ b/test/infinicore/ops/log_softmax.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/logaddexp.py b/test/infinicore/ops/logaddexp.py index 0266e8276..25c8a52c1 100644 --- a/test/infinicore/ops/logaddexp.py +++ b/test/infinicore/ops/logaddexp.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/logaddexp2.py b/test/infinicore/ops/logaddexp2.py index 6ee1b984c..d775a392f 100644 --- a/test/infinicore/ops/logaddexp2.py +++ b/test/infinicore/ops/logaddexp2.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/logcumsumexp.py b/test/infinicore/ops/logcumsumexp.py index 82514064b..f0c1a9fd9 100644 --- a/test/infinicore/ops/logcumsumexp.py +++ b/test/infinicore/ops/logcumsumexp.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/logdet.py b/test/infinicore/ops/logdet.py index 6b9ef1395..dfb28ffb4 100644 --- a/test/infinicore/ops/logdet.py +++ b/test/infinicore/ops/logdet.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (matrix_shape, strides_or_None) diff --git a/test/infinicore/ops/logical_and.py b/test/infinicore/ops/logical_and.py index 85f5438b9..f61836a57 100644 --- a/test/infinicore/ops/logical_and.py +++ b/test/infinicore/ops/logical_and.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/logical_not.py b/test/infinicore/ops/logical_not.py index c2e0d363d..c797b8675 100644 --- a/test/infinicore/ops/logical_not.py +++ b/test/infinicore/ops/logical_not.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/logical_or.py b/test/infinicore/ops/logical_or.py index a51b7384f..fd22eaad9 100644 --- a/test/infinicore/ops/logical_or.py +++ b/test/infinicore/ops/logical_or.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/logical_xor.py b/test/infinicore/ops/logical_xor.py index 4f82cc613..b72722f00 100644 --- a/test/infinicore/ops/logical_xor.py +++ b/test/infinicore/ops/logical_xor.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/logsigmoid.py b/test/infinicore/ops/logsigmoid.py index 0c9131323..e5e0eeb86 100644 --- a/test/infinicore/ops/logsigmoid.py +++ b/test/infinicore/ops/logsigmoid.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/logsumexp.py b/test/infinicore/ops/logsumexp.py index ff5df7a61..71b2d0278 100644 --- a/test/infinicore/ops/logsumexp.py +++ b/test/infinicore/ops/logsumexp.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/lp_pool1d.py b/test/infinicore/ops/lp_pool1d.py index 6efd30de2..620b54af5 100644 --- a/test/infinicore/ops/lp_pool1d.py +++ b/test/infinicore/ops/lp_pool1d.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (in_shape, in_strides_or_None, norm_type, kernel_size, stride_or_None, ceil_mode) diff --git a/test/infinicore/ops/lp_pool2d.py b/test/infinicore/ops/lp_pool2d.py index 4d8ec03cd..94212ea5c 100644 --- a/test/infinicore/ops/lp_pool2d.py +++ b/test/infinicore/ops/lp_pool2d.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (in_shape, in_strides_or_None, norm_type, kernel_size, stride_or_None, ceil_mode) diff --git a/test/infinicore/ops/lp_pool3d.py b/test/infinicore/ops/lp_pool3d.py index c3acc22a9..1ed221204 100644 --- a/test/infinicore/ops/lp_pool3d.py +++ b/test/infinicore/ops/lp_pool3d.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (in_shape, in_strides_or_None, norm_type, kernel_size, stride_or_None, ceil_mode) diff --git a/test/infinicore/ops/margin_ranking_loss.py b/test/infinicore/ops/margin_ranking_loss.py index 5ada6799b..eb8bb24c6 100644 --- a/test/infinicore/ops/margin_ranking_loss.py +++ b/test/infinicore/ops/margin_ranking_loss.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input1_shape, input2_shape, target_shape, input1_strides_or_None, input2_strides_or_None, target_strides_or_None, margin_or_None, p_or_None) diff --git a/test/infinicore/ops/masked_select.py b/test/infinicore/ops/masked_select.py index 711a0cef0..e7230b661 100644 --- a/test/infinicore/ops/masked_select.py +++ b/test/infinicore/ops/masked_select.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, input_strides_or_None, mask_shape) diff --git a/test/infinicore/ops/matmul.py b/test/infinicore/ops/matmul.py index 06a62efb9..79acc5cf7 100644 --- a/test/infinicore/ops/matmul.py +++ b/test/infinicore/ops/matmul.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/matrix_exp.py b/test/infinicore/ops/matrix_exp.py index f6d6a7f23..44b806d15 100644 --- a/test/infinicore/ops/matrix_exp.py +++ b/test/infinicore/ops/matrix_exp.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (matrix_shape, strides_or_None) diff --git a/test/infinicore/ops/matrix_power.py b/test/infinicore/ops/matrix_power.py index 95a33a0bb..ff69d910a 100644 --- a/test/infinicore/ops/matrix_power.py +++ b/test/infinicore/ops/matrix_power.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (matrix_shape, strides_or_None, n_or_None) diff --git a/test/infinicore/ops/max.py b/test/infinicore/ops/max.py index 89308c0f5..95c0fdeff 100644 --- a/test/infinicore/ops/max.py +++ b/test/infinicore/ops/max.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/max_pool1d.py b/test/infinicore/ops/max_pool1d.py index 2fa8fd995..21183f599 100644 --- a/test/infinicore/ops/max_pool1d.py +++ b/test/infinicore/ops/max_pool1d.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (in_shape, in_strides_or_None, kernel_size, stride_or_None, padding, dilation, ceil_mode) diff --git a/test/infinicore/ops/max_pool2d.py b/test/infinicore/ops/max_pool2d.py index f37e7beb7..70163e1fc 100644 --- a/test/infinicore/ops/max_pool2d.py +++ b/test/infinicore/ops/max_pool2d.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (in_shape, in_strides_or_None, kernel_size, stride_or_None, padding, dilation, ceil_mode) diff --git a/test/infinicore/ops/max_pool3d.py b/test/infinicore/ops/max_pool3d.py index e44009ebf..9e2db894d 100644 --- a/test/infinicore/ops/max_pool3d.py +++ b/test/infinicore/ops/max_pool3d.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (in_shape, in_strides_or_None, kernel_size, stride_or_None, padding, dilation, ceil_mode) diff --git a/test/infinicore/ops/max_unpool1d.py b/test/infinicore/ops/max_unpool1d.py index 3cb4acf9d..fee10c5df 100644 --- a/test/infinicore/ops/max_unpool1d.py +++ b/test/infinicore/ops/max_unpool1d.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner from framework.tensor import TensorInitializer diff --git a/test/infinicore/ops/max_unpool2d.py b/test/infinicore/ops/max_unpool2d.py index 1bbd8e181..eb295f0ff 100644 --- a/test/infinicore/ops/max_unpool2d.py +++ b/test/infinicore/ops/max_unpool2d.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner from framework.tensor import TensorInitializer diff --git a/test/infinicore/ops/max_unpool3d.py b/test/infinicore/ops/max_unpool3d.py index ea1d45569..7f04e56a3 100644 --- a/test/infinicore/ops/max_unpool3d.py +++ b/test/infinicore/ops/max_unpool3d.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner from framework.tensor import TensorInitializer diff --git a/test/infinicore/ops/maximum.py b/test/infinicore/ops/maximum.py index 0e3fc19c6..812ebbccb 100644 --- a/test/infinicore/ops/maximum.py +++ b/test/infinicore/ops/maximum.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/mean.py b/test/infinicore/ops/mean.py index 1b461d9af..780d562e2 100644 --- a/test/infinicore/ops/mean.py +++ b/test/infinicore/ops/mean.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/median.py b/test/infinicore/ops/median.py index 7a4c1531a..2cec283a2 100644 --- a/test/infinicore/ops/median.py +++ b/test/infinicore/ops/median.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/min.py b/test/infinicore/ops/min.py index 0901f4ee9..1ce1360a7 100644 --- a/test/infinicore/ops/min.py +++ b/test/infinicore/ops/min.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/minimum.py b/test/infinicore/ops/minimum.py index 4538665b6..e6a452d14 100644 --- a/test/infinicore/ops/minimum.py +++ b/test/infinicore/ops/minimum.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/mish.py b/test/infinicore/ops/mish.py index 087cabeac..5c5bfd5c0 100644 --- a/test/infinicore/ops/mish.py +++ b/test/infinicore/ops/mish.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/mse_loss.py b/test/infinicore/ops/mse_loss.py index 5231feec8..f871b7dce 100644 --- a/test/infinicore/ops/mse_loss.py +++ b/test/infinicore/ops/mse_loss.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, target_shape, input_strides_or_None, reduction_or_None) diff --git a/test/infinicore/ops/msort.py b/test/infinicore/ops/msort.py index 762f1125a..58068118f 100644 --- a/test/infinicore/ops/msort.py +++ b/test/infinicore/ops/msort.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/mul.py b/test/infinicore/ops/mul.py index 8a50cf019..f32f9449d 100644 --- a/test/infinicore/ops/mul.py +++ b/test/infinicore/ops/mul.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/multi_margin_loss.py b/test/infinicore/ops/multi_margin_loss.py index 1a0abf528..0e12bd608 100644 --- a/test/infinicore/ops/multi_margin_loss.py +++ b/test/infinicore/ops/multi_margin_loss.py @@ -3,9 +3,9 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) +import infinicore import torch import torch.nn.functional as F -import infinicore from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner from framework.tensor import TensorInitializer diff --git a/test/infinicore/ops/multilabel_margin_loss.py b/test/infinicore/ops/multilabel_margin_loss.py index 4137dd3ab..7968433ee 100644 --- a/test/infinicore/ops/multilabel_margin_loss.py +++ b/test/infinicore/ops/multilabel_margin_loss.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner from framework.tensor import TensorInitializer diff --git a/test/infinicore/ops/multilabel_soft_margin_loss.py b/test/infinicore/ops/multilabel_soft_margin_loss.py index d1b3b4ddb..b63be8db7 100644 --- a/test/infinicore/ops/multilabel_soft_margin_loss.py +++ b/test/infinicore/ops/multilabel_soft_margin_loss.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, target_shape, input_strides_or_None, reduction_or_None) diff --git a/test/infinicore/ops/mv.py b/test/infinicore/ops/mv.py index f9224c59f..a446a008b 100644 --- a/test/infinicore/ops/mv.py +++ b/test/infinicore/ops/mv.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/narrow.py b/test/infinicore/ops/narrow.py index 85f2b81b3..6630ab457 100644 --- a/test/infinicore/ops/narrow.py +++ b/test/infinicore/ops/narrow.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, input_strides_or_None, dim, start, length) diff --git a/test/infinicore/ops/narrow_copy.py b/test/infinicore/ops/narrow_copy.py index 44c1b0fb0..d9700b049 100644 --- a/test/infinicore/ops/narrow_copy.py +++ b/test/infinicore/ops/narrow_copy.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/nll_loss.py b/test/infinicore/ops/nll_loss.py index 03291ae4b..c44e08af1 100644 --- a/test/infinicore/ops/nll_loss.py +++ b/test/infinicore/ops/nll_loss.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner from framework.tensor import TensorInitializer diff --git a/test/infinicore/ops/nonzero.py b/test/infinicore/ops/nonzero.py index ef2bee325..51c186747 100644 --- a/test/infinicore/ops/nonzero.py +++ b/test/infinicore/ops/nonzero.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, input_strides_or_None) diff --git a/test/infinicore/ops/normalize.py b/test/infinicore/ops/normalize.py index 545ac59ce..b72f5ce77 100644 --- a/test/infinicore/ops/normalize.py +++ b/test/infinicore/ops/normalize.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, input_strides_or_None, p_or_None, dim_or_None, eps_or_None) diff --git a/test/infinicore/ops/one_hot.py b/test/infinicore/ops/one_hot.py index 22bb5a8bd..79989efa7 100644 --- a/test/infinicore/ops/one_hot.py +++ b/test/infinicore/ops/one_hot.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner from framework.tensor import TensorInitializer diff --git a/test/infinicore/ops/pad.py b/test/infinicore/ops/pad.py index 984a7e7cd..5dc933548 100644 --- a/test/infinicore/ops/pad.py +++ b/test/infinicore/ops/pad.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/paged_attention.py b/test/infinicore/ops/paged_attention.py index 0f7675101..9f66120d5 100644 --- a/test/infinicore/ops/paged_attention.py +++ b/test/infinicore/ops/paged_attention.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/paged_attention_prefill.py b/test/infinicore/ops/paged_attention_prefill.py index d4d97f073..a5a21cbf9 100644 --- a/test/infinicore/ops/paged_attention_prefill.py +++ b/test/infinicore/ops/paged_attention_prefill.py @@ -1,9 +1,8 @@ import os import sys -import torch - import infinicore +import torch sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) diff --git a/test/infinicore/ops/paged_caching.py b/test/infinicore/ops/paged_caching.py index 65e59eaae..7971beb52 100644 --- a/test/infinicore/ops/paged_caching.py +++ b/test/infinicore/ops/paged_caching.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/pairwise_distance.py b/test/infinicore/ops/pairwise_distance.py index 4a446613c..c63a6e2f3 100644 --- a/test/infinicore/ops/pairwise_distance.py +++ b/test/infinicore/ops/pairwise_distance.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/pdist.py b/test/infinicore/ops/pdist.py index 593a8bd4d..c2b3a1ed3 100644 --- a/test/infinicore/ops/pdist.py +++ b/test/infinicore/ops/pdist.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/pixel_shuffle.py b/test/infinicore/ops/pixel_shuffle.py index d9404189f..8fb813210 100644 --- a/test/infinicore/ops/pixel_shuffle.py +++ b/test/infinicore/ops/pixel_shuffle.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/pixel_unshuffle.py b/test/infinicore/ops/pixel_unshuffle.py index 4d86ff4e2..d459df874 100644 --- a/test/infinicore/ops/pixel_unshuffle.py +++ b/test/infinicore/ops/pixel_unshuffle.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/prelu.py b/test/infinicore/ops/prelu.py index 22ede01e9..c5327ec40 100644 --- a/test/infinicore/ops/prelu.py +++ b/test/infinicore/ops/prelu.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/prod.py b/test/infinicore/ops/prod.py index 550a49e1b..aada5819f 100644 --- a/test/infinicore/ops/prod.py +++ b/test/infinicore/ops/prod.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/quantile.py b/test/infinicore/ops/quantile.py index 547eba7ac..3cd223a6d 100644 --- a/test/infinicore/ops/quantile.py +++ b/test/infinicore/ops/quantile.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/random_sample.py b/test/infinicore/ops/random_sample.py index 60669cd81..7d1bb7c48 100644 --- a/test/infinicore/ops/random_sample.py +++ b/test/infinicore/ops/random_sample.py @@ -3,9 +3,9 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore import infinicore.nn.functional as F +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner from framework.tensor import TensorInitializer diff --git a/test/infinicore/ops/reciprocal.py b/test/infinicore/ops/reciprocal.py index 9d4ba0849..7548bc4a5 100644 --- a/test/infinicore/ops/reciprocal.py +++ b/test/infinicore/ops/reciprocal.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/relu6.py b/test/infinicore/ops/relu6.py index 23952d6d5..fe1ad7c63 100644 --- a/test/infinicore/ops/relu6.py +++ b/test/infinicore/ops/relu6.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/remainder.py b/test/infinicore/ops/remainder.py index f86e96514..653ef699e 100644 --- a/test/infinicore/ops/remainder.py +++ b/test/infinicore/ops/remainder.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/reshape.py b/test/infinicore/ops/reshape.py index 79710f567..e2b9741bb 100644 --- a/test/infinicore/ops/reshape.py +++ b/test/infinicore/ops/reshape.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/rms_norm.py b/test/infinicore/ops/rms_norm.py index 6ea3c2d9c..3f1886283 100644 --- a/test/infinicore/ops/rms_norm.py +++ b/test/infinicore/ops/rms_norm.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/rope.py b/test/infinicore/ops/rope.py index b36c568db..ea5f71a80 100644 --- a/test/infinicore/ops/rope.py +++ b/test/infinicore/ops/rope.py @@ -3,7 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) - +import infinicore +from infinicore.nn.functional import RopeAlgo import torch from framework import ( BaseOperatorTest, @@ -12,9 +13,6 @@ GenericTestRunner, is_broadcast, ) -from infinicore.nn.functional import RopeAlgo - -import infinicore # ============================================================================== # Operator-specific configuration diff --git a/test/infinicore/ops/rot90.py b/test/infinicore/ops/rot90.py index d842f4876..7e569c54a 100644 --- a/test/infinicore/ops/rot90.py +++ b/test/infinicore/ops/rot90.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/round.py b/test/infinicore/ops/round.py index e9e244b0d..7cd48f660 100644 --- a/test/infinicore/ops/round.py +++ b/test/infinicore/ops/round.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/rrelu.py b/test/infinicore/ops/rrelu.py index ff5a8b212..3d51de62a 100644 --- a/test/infinicore/ops/rrelu.py +++ b/test/infinicore/ops/rrelu.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/scaled_dot_product_attention.py b/test/infinicore/ops/scaled_dot_product_attention.py index 218420d72..2097db9d1 100644 --- a/test/infinicore/ops/scaled_dot_product_attention.py +++ b/test/infinicore/ops/scaled_dot_product_attention.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (q_shape, k_shape, v_shape, attn_mask_or_None, dropout_p, is_causal) diff --git a/test/infinicore/ops/scatter.py b/test/infinicore/ops/scatter.py index d99cfb57b..31e6a4c8d 100644 --- a/test/infinicore/ops/scatter.py +++ b/test/infinicore/ops/scatter.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/scatter_add.py b/test/infinicore/ops/scatter_add.py index abfe1caa3..323e4ae8e 100644 --- a/test/infinicore/ops/scatter_add.py +++ b/test/infinicore/ops/scatter_add.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/scatter_reduce.py b/test/infinicore/ops/scatter_reduce.py index 88833c7ac..ab0ce089c 100644 --- a/test/infinicore/ops/scatter_reduce.py +++ b/test/infinicore/ops/scatter_reduce.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/select_op.py b/test/infinicore/ops/select_op.py index 422a130d7..357b3eac5 100644 --- a/test/infinicore/ops/select_op.py +++ b/test/infinicore/ops/select_op.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, input_strides_or_None, dim, index) diff --git a/test/infinicore/ops/select_scatter.py b/test/infinicore/ops/select_scatter.py index 12375ce15..f3e3b7b01 100644 --- a/test/infinicore/ops/select_scatter.py +++ b/test/infinicore/ops/select_scatter.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/selu.py b/test/infinicore/ops/selu.py index 484bd78c7..f6f44682a 100644 --- a/test/infinicore/ops/selu.py +++ b/test/infinicore/ops/selu.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/sgn.py b/test/infinicore/ops/sgn.py index 9631f2ab3..b8f10c707 100644 --- a/test/infinicore/ops/sgn.py +++ b/test/infinicore/ops/sgn.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (in_shape, in_strides_or_None) diff --git a/test/infinicore/ops/sign.py b/test/infinicore/ops/sign.py index 729b6c4ae..9aafd9387 100644 --- a/test/infinicore/ops/sign.py +++ b/test/infinicore/ops/sign.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (in_shape, in_strides_or_None) diff --git a/test/infinicore/ops/signbit.py b/test/infinicore/ops/signbit.py index 93b4b5118..fec36df86 100644 --- a/test/infinicore/ops/signbit.py +++ b/test/infinicore/ops/signbit.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (in_shape, in_strides_or_None) diff --git a/test/infinicore/ops/silu.py b/test/infinicore/ops/silu.py index 0b6a44beb..79314c511 100644 --- a/test/infinicore/ops/silu.py +++ b/test/infinicore/ops/silu.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/sinh.py b/test/infinicore/ops/sinh.py index 06c6ca608..5f55c5b99 100644 --- a/test/infinicore/ops/sinh.py +++ b/test/infinicore/ops/sinh.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/slice_scatter.py b/test/infinicore/ops/slice_scatter.py index aa619093e..6a6357f7f 100644 --- a/test/infinicore/ops/slice_scatter.py +++ b/test/infinicore/ops/slice_scatter.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/slogdet.py b/test/infinicore/ops/slogdet.py index 015915daa..402081f53 100644 --- a/test/infinicore/ops/slogdet.py +++ b/test/infinicore/ops/slogdet.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (matrix_shape, strides_or_None) diff --git a/test/infinicore/ops/smooth_l1_loss.py b/test/infinicore/ops/smooth_l1_loss.py index 7f0ca670c..c50b316b2 100644 --- a/test/infinicore/ops/smooth_l1_loss.py +++ b/test/infinicore/ops/smooth_l1_loss.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, target_shape, input_strides_or_None, beta_or_None, reduction_or_None) diff --git a/test/infinicore/ops/soft_margin_loss.py b/test/infinicore/ops/soft_margin_loss.py index a86df9a31..01a601565 100644 --- a/test/infinicore/ops/soft_margin_loss.py +++ b/test/infinicore/ops/soft_margin_loss.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, target_shape, input_strides_or_None, reduction_or_None) diff --git a/test/infinicore/ops/softmin.py b/test/infinicore/ops/softmin.py index 9f5a5e0a7..e2f224449 100644 --- a/test/infinicore/ops/softmin.py +++ b/test/infinicore/ops/softmin.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/softplus.py b/test/infinicore/ops/softplus.py index c01c6d632..9a88f69db 100644 --- a/test/infinicore/ops/softplus.py +++ b/test/infinicore/ops/softplus.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/softshrink.py b/test/infinicore/ops/softshrink.py index 0dd58b311..14e9e3840 100644 --- a/test/infinicore/ops/softshrink.py +++ b/test/infinicore/ops/softshrink.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/softsign.py b/test/infinicore/ops/softsign.py index c482e72e5..f41a42729 100644 --- a/test/infinicore/ops/softsign.py +++ b/test/infinicore/ops/softsign.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/sort.py b/test/infinicore/ops/sort.py index d01d43a93..7bcff0402 100644 --- a/test/infinicore/ops/sort.py +++ b/test/infinicore/ops/sort.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, CaseResult, diff --git a/test/infinicore/ops/split.py b/test/infinicore/ops/split.py index 3a639ed1b..ca327413f 100644 --- a/test/infinicore/ops/split.py +++ b/test/infinicore/ops/split.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, input_strides_or_None, split_size_or_sections, dim_or_None) diff --git a/test/infinicore/ops/sqrt.py b/test/infinicore/ops/sqrt.py index df1de761f..72d8db3fa 100644 --- a/test/infinicore/ops/sqrt.py +++ b/test/infinicore/ops/sqrt.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/square.py b/test/infinicore/ops/square.py index 35e50a377..e66957751 100644 --- a/test/infinicore/ops/square.py +++ b/test/infinicore/ops/square.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/stack.py b/test/infinicore/ops/stack.py index 4c576f1a8..77f194e70 100644 --- a/test/infinicore/ops/stack.py +++ b/test/infinicore/ops/stack.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # ============================================================================== diff --git a/test/infinicore/ops/std.py b/test/infinicore/ops/std.py index bab678144..296fd2286 100644 --- a/test/infinicore/ops/std.py +++ b/test/infinicore/ops/std.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, CaseResult, diff --git a/test/infinicore/ops/std_mean.py b/test/infinicore/ops/std_mean.py index ffcc888cc..542858bf2 100644 --- a/test/infinicore/ops/std_mean.py +++ b/test/infinicore/ops/std_mean.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/sum.py b/test/infinicore/ops/sum.py index 7cf4be80d..74d9e29fc 100644 --- a/test/infinicore/ops/sum.py +++ b/test/infinicore/ops/sum.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/svd.py b/test/infinicore/ops/svd.py index 498f25fa4..f62052674 100644 --- a/test/infinicore/ops/svd.py +++ b/test/infinicore/ops/svd.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (matrix_shape, strides_or_None, compute_uv_or_None) diff --git a/test/infinicore/ops/swiglu.py b/test/infinicore/ops/swiglu.py index f239c8472..4f8e391b5 100644 --- a/test/infinicore/ops/swiglu.py +++ b/test/infinicore/ops/swiglu.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/take.py b/test/infinicore/ops/take.py index 5ca28b86e..3cfbffc75 100644 --- a/test/infinicore/ops/take.py +++ b/test/infinicore/ops/take.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner from framework.tensor import TensorInitializer diff --git a/test/infinicore/ops/tan.py b/test/infinicore/ops/tan.py index c9948c3c6..acb1cac5a 100644 --- a/test/infinicore/ops/tan.py +++ b/test/infinicore/ops/tan.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/tanhshrink.py b/test/infinicore/ops/tanhshrink.py index ca3559f22..4e60eb264 100644 --- a/test/infinicore/ops/tanhshrink.py +++ b/test/infinicore/ops/tanhshrink.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/threshold.py b/test/infinicore/ops/threshold.py index 6a6eaad7a..b95798a37 100644 --- a/test/infinicore/ops/threshold.py +++ b/test/infinicore/ops/threshold.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/topk.py b/test/infinicore/ops/topk.py index b07f9ed7a..a0b9bdcd8 100644 --- a/test/infinicore/ops/topk.py +++ b/test/infinicore/ops/topk.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/transpose.py b/test/infinicore/ops/transpose.py index 8bf00dde4..5c4520349 100644 --- a/test/infinicore/ops/transpose.py +++ b/test/infinicore/ops/transpose.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/triplet_margin_loss.py b/test/infinicore/ops/triplet_margin_loss.py index a0cbc8ff7..f25a04245 100644 --- a/test/infinicore/ops/triplet_margin_loss.py +++ b/test/infinicore/ops/triplet_margin_loss.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (anchor_shape, positive_shape, negative_shape, strides_or_None, margin_or_None, p_or_None, eps_or_None, swap_or_None) diff --git a/test/infinicore/ops/triplet_margin_with_distance_loss.py b/test/infinicore/ops/triplet_margin_with_distance_loss.py index 35ca4a1ea..6b6ce174e 100644 --- a/test/infinicore/ops/triplet_margin_with_distance_loss.py +++ b/test/infinicore/ops/triplet_margin_with_distance_loss.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (anchor_shape, positive_shape, negative_shape, strides_or_None, margin_or_None, swap_or_None) diff --git a/test/infinicore/ops/trunc.py b/test/infinicore/ops/trunc.py index 83fdfbb2e..5988e7ff5 100644 --- a/test/infinicore/ops/trunc.py +++ b/test/infinicore/ops/trunc.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/unfold.py b/test/infinicore/ops/unfold.py index 2d2b30603..32f983b8a 100644 --- a/test/infinicore/ops/unfold.py +++ b/test/infinicore/ops/unfold.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (in_shape, in_strides_or_None, kernel_size, dilation, padding, stride) diff --git a/test/infinicore/ops/unique.py b/test/infinicore/ops/unique.py index 876d5691e..f2218d610 100644 --- a/test/infinicore/ops/unique.py +++ b/test/infinicore/ops/unique.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/upsample.py b/test/infinicore/ops/upsample.py index 13357de4e..4f599933d 100644 --- a/test/infinicore/ops/upsample.py +++ b/test/infinicore/ops/upsample.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/upsample_bilinear.py b/test/infinicore/ops/upsample_bilinear.py index 65c62f2a8..087e1ca26 100644 --- a/test/infinicore/ops/upsample_bilinear.py +++ b/test/infinicore/ops/upsample_bilinear.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/upsample_nearest.py b/test/infinicore/ops/upsample_nearest.py index 58c6d4e96..09d2bae27 100644 --- a/test/infinicore/ops/upsample_nearest.py +++ b/test/infinicore/ops/upsample_nearest.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/vander.py b/test/infinicore/ops/vander.py index 25040d474..bb89c004f 100644 --- a/test/infinicore/ops/vander.py +++ b/test/infinicore/ops/vander.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (input_shape, input_strides_or_None, N) diff --git a/test/infinicore/ops/var.py b/test/infinicore/ops/var.py index 1869085ec..e0ce9f463 100644 --- a/test/infinicore/ops/var.py +++ b/test/infinicore/ops/var.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/var_mean.py b/test/infinicore/ops/var_mean.py index 5a696fdf9..18015d2cd 100644 --- a/test/infinicore/ops/var_mean.py +++ b/test/infinicore/ops/var_mean.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import ( BaseOperatorTest, TensorSpec, diff --git a/test/infinicore/ops/vdot.py b/test/infinicore/ops/vdot.py index 2baf715f0..50e845f6b 100644 --- a/test/infinicore/ops/vdot.py +++ b/test/infinicore/ops/vdot.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (vec1_shape, vec2_shape, vec1_strides_or_None, vec2_strides_or_None) diff --git a/test/infinicore/ops/where.py b/test/infinicore/ops/where.py index bc0013bd6..a510c70c6 100644 --- a/test/infinicore/ops/where.py +++ b/test/infinicore/ops/where.py @@ -3,8 +3,8 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..")) -import torch import infinicore +import torch from framework import BaseOperatorTest, TensorSpec, TestCase, GenericTestRunner # Test cases format: (condition_shape, cond_strides_or_None, x_shape_or_None, y_shape_or_None) diff --git a/test/infiniop/gelu.py b/test/infiniop/gelu.py index fd1e4eebc..b9adbc115 100644 --- a/test/infiniop/gelu.py +++ b/test/infiniop/gelu.py @@ -15,6 +15,7 @@ InfiniDtype, InfiniDtypeNames, InfiniDeviceNames, + InfiniDeviceEnum, infiniopOperatorDescriptor_t, ) from enum import Enum, auto @@ -83,6 +84,12 @@ def test( dtype=torch.float16, sync=None, ): + # Skip strided cases on Iluvatar: GELU with non-contiguous tensors can hang the GPU (requires ixsmi -r to recover) + if device == InfiniDeviceEnum.ILUVATAR and ( + input_stride is not None or output_stride is not None + ): + return + input = TestTensor(shape, input_stride, dtype, device) if inplace == Inplace.INPLACE: if input_stride != output_stride: @@ -141,6 +148,9 @@ def lib_gelu(): lib_gelu() + if sync is not None: + sync() + atol, rtol = get_tolerance(_TOLERANCE_MAP, dtype) if DEBUG: debug(output.actual_tensor(), output.torch_tensor(), atol=atol, rtol=rtol) From f46e9f65b1a04302a9c7f7adaae9884de2d17860 Mon Sep 17 00:00:00 2001 From: zhangyue Date: Thu, 12 Feb 2026 02:40:32 +0000 Subject: [PATCH 2/7] issue/1008: adapt lpnorm layernorm softmax rearrange paged_attention for iluvatar --- scripts/python_test.py | 9 +++++--- .../layer_norm/nvidia/layer_norm_nvidia.cu | 2 ++ src/infiniop/ops/layer_norm/operator.cc | 3 +++ .../logsoftmax/nvidia/logsoftmax_nvidia.cu | 5 ++++ src/infiniop/ops/logsoftmax/operator.cc | 8 +++---- .../ops/lp_norm/nvidia/lp_norm_nvidia.cu | 2 ++ .../ops/rearrange/nvidia/rearrange_kernel.cuh | 23 ++++++++++++++++--- src/infiniop/ops/sigmoid/operator.cc | 17 ++++++++++---- .../ops/softmax/nvidia/softmax_nvidia.cu | 3 +++ src/infiniop/ops/topksoftmax/operator.cc | 14 ++++++++++- test/infiniop/ones.py | 7 ++++++ test/infiniop/zeros.py | 7 ++++++ xmake.lua | 8 ++++--- xmake/iluvatar.lua | 10 ++++---- 14 files changed, 96 insertions(+), 22 deletions(-) diff --git a/scripts/python_test.py b/scripts/python_test.py index 06af369ef..eece1218c 100644 --- a/scripts/python_test.py +++ b/scripts/python_test.py @@ -20,9 +20,9 @@ def run_tests(args): #"dequantize_awq.py", "gelu.py", "gemm.py", - #"layer_norm.py", + "layer_norm.py", "logsoftmax.py", - #"lp_norm.py", + "lp_norm.py", "mul.py", "ones.py", "random_sample.py", @@ -31,7 +31,7 @@ def run_tests(args): "rms_norm.py", "rope.py", "sigmoid.py", - #"softmax.py", + "softmax.py", "softplus.py", "sub.py", "swiglu.py", @@ -39,6 +39,9 @@ def run_tests(args): "topkrouter.py", "topksoftmax.py", "zeros.py", + "paged_attention.py", + "paged_caching.py", + "paged_attention_prefill.py" ]: result = subprocess.run( f"python {test} {args} --debug", text=True, encoding="utf-8", shell=True diff --git a/src/infiniop/ops/layer_norm/nvidia/layer_norm_nvidia.cu b/src/infiniop/ops/layer_norm/nvidia/layer_norm_nvidia.cu index 937567958..5546f5e2f 100644 --- a/src/infiniop/ops/layer_norm/nvidia/layer_norm_nvidia.cu +++ b/src/infiniop/ops/layer_norm/nvidia/layer_norm_nvidia.cu @@ -255,6 +255,8 @@ infiniStatus_t Descriptor::calculate( CALCULATE_LAYER_NORM_WITH_BLOCK_SIZE(CUDA_BLOCK_SIZE_512) } else if (_opaque->internal->maxThreadsPerBlock() == CUDA_BLOCK_SIZE_4096) { CALCULATE_LAYER_NORM_WITH_BLOCK_SIZE(CUDA_BLOCK_SIZE_4096) + } else if (_opaque->internal->maxThreadsPerBlock() == CUDA_BLOCK_SIZE_2048) { + CALCULATE_LAYER_NORM_WITH_BLOCK_SIZE(CUDA_BLOCK_SIZE_2048) } else { return INFINI_STATUS_DEVICE_ARCHITECTURE_NOT_SUPPORTED; } diff --git a/src/infiniop/ops/layer_norm/operator.cc b/src/infiniop/ops/layer_norm/operator.cc index 1554e8b3b..c03b8ac92 100644 --- a/src/infiniop/ops/layer_norm/operator.cc +++ b/src/infiniop/ops/layer_norm/operator.cc @@ -174,6 +174,9 @@ infiniopDestroyLayerNormDescriptor(infiniopLayerNormDescriptor_t desc) { #ifdef ENABLE_METAX_API DELETE(INFINI_DEVICE_METAX, metax); #endif +#ifdef ENABLE_ILUVATAR_API + DELETE(INFINI_DEVICE_ILUVATAR, nvidia); +#endif default: return INFINI_STATUS_DEVICE_TYPE_NOT_SUPPORTED; diff --git a/src/infiniop/ops/logsoftmax/nvidia/logsoftmax_nvidia.cu b/src/infiniop/ops/logsoftmax/nvidia/logsoftmax_nvidia.cu index 1235b2aaf..7cfe4f3a0 100644 --- a/src/infiniop/ops/logsoftmax/nvidia/logsoftmax_nvidia.cu +++ b/src/infiniop/ops/logsoftmax/nvidia/logsoftmax_nvidia.cu @@ -117,6 +117,11 @@ infiniStatus_t Descriptor::calculate(void *workspace, size_t workspace_size, y, x, _info.x_dtype, _info.y_dtype, _info.batch_size, _info.probs_size, _info.ndim, _info.seq_len, _info.y_stride_b, _info.y_stride_p, _info.x_stride_b, _info.x_stride_p, _info.y_stride_0, _info.y_stride_1, _info.x_stride_0, _info.x_stride_1, stream)); + } else if (_opaque->internal->maxThreadsPerBlock() == CUDA_BLOCK_SIZE_2048) { + CHECK_STATUS(launchKernel( + y, x, _info.x_dtype, _info.y_dtype, _info.batch_size, _info.probs_size, _info.ndim, _info.seq_len, + _info.y_stride_b, _info.y_stride_p, _info.x_stride_b, _info.x_stride_p, + _info.y_stride_0, _info.y_stride_1, _info.x_stride_0, _info.x_stride_1, stream)); } else if (_opaque->internal->maxThreadsPerBlock() == CUDA_BLOCK_SIZE_4096) { CHECK_STATUS(launchKernel( y, x, _info.x_dtype, _info.y_dtype, _info.batch_size, _info.probs_size, _info.ndim, _info.seq_len, diff --git a/src/infiniop/ops/logsoftmax/operator.cc b/src/infiniop/ops/logsoftmax/operator.cc index aa11cd91a..1a44a87c4 100644 --- a/src/infiniop/ops/logsoftmax/operator.cc +++ b/src/infiniop/ops/logsoftmax/operator.cc @@ -40,7 +40,7 @@ __C infiniStatus_t infiniopCreateLogSoftmaxDescriptor( CREATE(INFINI_DEVICE_ALI, nvidia); #endif #ifdef ENABLE_ILUVATAR_API - // CREATE(INFINI_DEVICE_ILUVATAR, nvidia); + CREATE(INFINI_DEVICE_ILUVATAR, nvidia); #endif #ifdef ENABLE_QY_API CREATE(INFINI_DEVICE_QY, nvidia); @@ -73,7 +73,7 @@ __C infiniStatus_t infiniopGetLogSoftmaxWorkspaceSize(infiniopLogSoftmaxDescript GET(INFINI_DEVICE_ALI, nvidia); #endif #ifdef ENABLE_ILUVATAR_API - // GET(INFINI_DEVICE_ILUVATAR, nvidia); + GET(INFINI_DEVICE_ILUVATAR, nvidia); #endif #ifdef ENABLE_QY_API GET(INFINI_DEVICE_QY, nvidia); @@ -111,7 +111,7 @@ __C infiniStatus_t infiniopLogSoftmax( CALCULATE(INFINI_DEVICE_ALI, nvidia); #endif #ifdef ENABLE_ILUVATAR_API - // CALCULATE(INFINI_DEVICE_ILUVATAR, nvidia); + CALCULATE(INFINI_DEVICE_ILUVATAR, nvidia); #endif #ifdef ENABLE_QY_API CALCULATE(INFINI_DEVICE_QY, nvidia); @@ -144,7 +144,7 @@ __C infiniStatus_t infiniopDestroyLogSoftmaxDescriptor(infiniopLogSoftmaxDescrip DESTROY(INFINI_DEVICE_ALI, nvidia); #endif #ifdef ENABLE_ILUVATAR_API - // DESTROY(INFINI_DEVICE_ILUVATAR, nvidia); + DESTROY(INFINI_DEVICE_ILUVATAR, nvidia); #endif #ifdef ENABLE_QY_API DESTROY(INFINI_DEVICE_QY, nvidia); diff --git a/src/infiniop/ops/lp_norm/nvidia/lp_norm_nvidia.cu b/src/infiniop/ops/lp_norm/nvidia/lp_norm_nvidia.cu index 48a992df1..ab463a55d 100644 --- a/src/infiniop/ops/lp_norm/nvidia/lp_norm_nvidia.cu +++ b/src/infiniop/ops/lp_norm/nvidia/lp_norm_nvidia.cu @@ -155,6 +155,8 @@ infiniStatus_t Descriptor::calculate(void *workspace, size_t workspace_size, CALCULATE_LP_NORM_WITH_BLOCK_SIZE(CUDA_BLOCK_SIZE_1024) } else if (_opaque->internal->maxThreadsPerBlock() == CUDA_BLOCK_SIZE_512) { CALCULATE_LP_NORM_WITH_BLOCK_SIZE(CUDA_BLOCK_SIZE_512) + } else if (_opaque->internal->maxThreadsPerBlock() == CUDA_BLOCK_SIZE_2048) { + CALCULATE_LP_NORM_WITH_BLOCK_SIZE(CUDA_BLOCK_SIZE_2048) } else if (_opaque->internal->maxThreadsPerBlock() == CUDA_BLOCK_SIZE_4096) { CALCULATE_LP_NORM_WITH_BLOCK_SIZE(CUDA_BLOCK_SIZE_4096) } else { diff --git a/src/infiniop/ops/rearrange/nvidia/rearrange_kernel.cuh b/src/infiniop/ops/rearrange/nvidia/rearrange_kernel.cuh index f273bfeba..0bd8d8b68 100644 --- a/src/infiniop/ops/rearrange/nvidia/rearrange_kernel.cuh +++ b/src/infiniop/ops/rearrange/nvidia/rearrange_kernel.cuh @@ -8,8 +8,8 @@ #define ARRAY_TYPE_SIZE size_t // 与 DEFINE_KERNELS_BY_CONSTRAINT 耦合，需要同时修改 -#define MAX_BLOCK_ARRAY_SIZE 5 -#define MAX_GRID_ARRAY_SIZE 5 +#define MAX_BLOCK_ARRAY_SIZE 6 +#define MAX_GRID_ARRAY_SIZE 6 template struct ArrayStruct { @@ -185,32 +185,43 @@ struct Constraint { DEFINE_REARRANGE_KERNEL(double4, constraint_num, block_array_size, grid_array_size) // 与 MAX_BLOCK_ARRAY_SIZE 和 MAX_GRID_ARRAY_SIZE 耦合，需要同时修改 -// 为1-5和1-5的所有组合生成内核 +// 为1-6和1-6的所有组合生成内核 DEFINE_KERNELS_BY_CONSTRAINT(1, 1) DEFINE_KERNELS_BY_CONSTRAINT(1, 2) DEFINE_KERNELS_BY_CONSTRAINT(1, 3) DEFINE_KERNELS_BY_CONSTRAINT(1, 4) DEFINE_KERNELS_BY_CONSTRAINT(1, 5) +DEFINE_KERNELS_BY_CONSTRAINT(1, 6) DEFINE_KERNELS_BY_CONSTRAINT(2, 1) DEFINE_KERNELS_BY_CONSTRAINT(2, 2) DEFINE_KERNELS_BY_CONSTRAINT(2, 3) DEFINE_KERNELS_BY_CONSTRAINT(2, 4) DEFINE_KERNELS_BY_CONSTRAINT(2, 5) +DEFINE_KERNELS_BY_CONSTRAINT(2, 6) DEFINE_KERNELS_BY_CONSTRAINT(3, 1) DEFINE_KERNELS_BY_CONSTRAINT(3, 2) DEFINE_KERNELS_BY_CONSTRAINT(3, 3) DEFINE_KERNELS_BY_CONSTRAINT(3, 4) DEFINE_KERNELS_BY_CONSTRAINT(3, 5) +DEFINE_KERNELS_BY_CONSTRAINT(3, 6) DEFINE_KERNELS_BY_CONSTRAINT(4, 1) DEFINE_KERNELS_BY_CONSTRAINT(4, 2) DEFINE_KERNELS_BY_CONSTRAINT(4, 3) DEFINE_KERNELS_BY_CONSTRAINT(4, 4) DEFINE_KERNELS_BY_CONSTRAINT(4, 5) +DEFINE_KERNELS_BY_CONSTRAINT(4, 6) DEFINE_KERNELS_BY_CONSTRAINT(5, 1) DEFINE_KERNELS_BY_CONSTRAINT(5, 2) DEFINE_KERNELS_BY_CONSTRAINT(5, 3) DEFINE_KERNELS_BY_CONSTRAINT(5, 4) DEFINE_KERNELS_BY_CONSTRAINT(5, 5) +DEFINE_KERNELS_BY_CONSTRAINT(5, 6) +DEFINE_KERNELS_BY_CONSTRAINT(6, 1) +DEFINE_KERNELS_BY_CONSTRAINT(6, 2) +DEFINE_KERNELS_BY_CONSTRAINT(6, 3) +DEFINE_KERNELS_BY_CONSTRAINT(6, 4) +DEFINE_KERNELS_BY_CONSTRAINT(6, 5) +DEFINE_KERNELS_BY_CONSTRAINT(6, 6) // 准备参数结构体 struct RearrangeParams { @@ -294,6 +305,9 @@ utils::Result getRearrangeKernel(const RearrangeParams ¶ms) { case 5: \ GET_REARRANGE_KERNEL_BY_CONSTRAINT(block_array_size, 5); \ break; \ + case 6: \ + GET_REARRANGE_KERNEL_BY_CONSTRAINT(block_array_size, 6); \ + break; \ } #define GET_REARRANGE_KERNEL_BY_BLOCK_NUM \ @@ -313,6 +327,9 @@ utils::Result getRearrangeKernel(const RearrangeParams ¶ms) { case 5: \ GET_REARRANGE_KERNEL_BY_GRID_NUM(5); \ break; \ + case 6: \ + GET_REARRANGE_KERNEL_BY_GRID_NUM(6); \ + break; \ } GET_REARRANGE_KERNEL_BY_BLOCK_NUM diff --git a/src/infiniop/ops/sigmoid/operator.cc b/src/infiniop/ops/sigmoid/operator.cc index 80d0487df..f9df2c8e2 100644 --- a/src/infiniop/ops/sigmoid/operator.cc +++ b/src/infiniop/ops/sigmoid/operator.cc @@ -5,7 +5,7 @@ #ifdef ENABLE_CPU_API #include "cpu/sigmoid_cpu.h" #endif -#if defined(ENABLE_NVIDIA_API) || defined(ENABLE_QY_API) || defined(ENABLE_ALI_API) +#if defined(ENABLE_NVIDIA_API) || defined(ENABLE_QY_API) || defined(ENABLE_ALI_API) || defined(ENABLE_ILUVATAR_API) #include "nvidia/sigmoid_nvidia.cuh" #endif @@ -37,6 +37,9 @@ __C infiniStatus_t infiniopCreateSigmoidDescriptor( #ifdef ENABLE_ALI_API CREATE(INFINI_DEVICE_ALI, nvidia); #endif +#ifdef ENABLE_ILUVATAR_API + CREATE(INFINI_DEVICE_ILUVATAR, nvidia); +#endif default: return INFINI_STATUS_DEVICE_TYPE_NOT_SUPPORTED; @@ -65,7 +68,9 @@ __C infiniStatus_t infiniopGetSigmoidWorkspaceSize(infiniopSigmoidDescriptor_t d #ifdef ENABLE_ALI_API GET(INFINI_DEVICE_ALI, nvidia); #endif - +#ifdef ENABLE_ILUVATAR_API + GET(INFINI_DEVICE_ILUVATAR, nvidia); +#endif default: return INFINI_STATUS_DEVICE_TYPE_NOT_SUPPORTED; } @@ -101,7 +106,9 @@ __C infiniStatus_t infiniopSigmoid( #ifdef ENABLE_ALI_API CALCULATE(INFINI_DEVICE_ALI, nvidia); #endif - +#ifdef ENABLE_ILUVATAR_API + CALCULATE(INFINI_DEVICE_ILUVATAR, nvidia); +#endif default: return INFINI_STATUS_DEVICE_TYPE_NOT_SUPPORTED; } @@ -131,7 +138,9 @@ infiniopDestroySigmoidDescriptor(infiniopSigmoidDescriptor_t desc) { #ifdef ENABLE_ALI_API DELETE(INFINI_DEVICE_ALI, nvidia); #endif - +#ifdef ENABLE_ILUVATAR_API + DELETE(INFINI_DEVICE_ILUVATAR, nvidia); +#endif default: return INFINI_STATUS_DEVICE_TYPE_NOT_SUPPORTED; } diff --git a/src/infiniop/ops/softmax/nvidia/softmax_nvidia.cu b/src/infiniop/ops/softmax/nvidia/softmax_nvidia.cu index d87fe8167..a523bc41d 100644 --- a/src/infiniop/ops/softmax/nvidia/softmax_nvidia.cu +++ b/src/infiniop/ops/softmax/nvidia/softmax_nvidia.cu @@ -128,6 +128,9 @@ infiniStatus_t Descriptor::calculate(void *workspace, size_t workspace_size, } else if (_opaque->internal->maxThreadsPerBlock() == CUDA_BLOCK_SIZE_4096) { CHECK_STATUS(launchKernel( y, x, _info.dtype, _info.othersize, _info.dimsize, _info.stride, stream)); + } else if (_opaque->internal->maxThreadsPerBlock() == CUDA_BLOCK_SIZE_2048) { + CHECK_STATUS(launchKernel( + y, x, _info.dtype, _info.othersize, _info.dimsize, _info.stride, stream)); } else { return INFINI_STATUS_DEVICE_ARCHITECTURE_NOT_SUPPORTED; } diff --git a/src/infiniop/ops/topksoftmax/operator.cc b/src/infiniop/ops/topksoftmax/operator.cc index 622c05387..ad0ff36a0 100644 --- a/src/infiniop/ops/topksoftmax/operator.cc +++ b/src/infiniop/ops/topksoftmax/operator.cc @@ -5,7 +5,7 @@ #ifdef ENABLE_CPU_API #include "cpu/topksoftmax_cpu.h" #endif -#if defined(ENABLE_NVIDIA_API) || defined(ENABLE_QY_API) || defined(ENABLE_ALI_API) +#if defined(ENABLE_NVIDIA_API) || defined(ENABLE_QY_API) || defined(ENABLE_ALI_API) || defined(ENABLE_ILUVATAR_API) #include "nvidia/topksoftmax_nvidia.cuh" #endif #ifdef ENABLE_METAX_API @@ -36,6 +36,9 @@ __C infiniStatus_t infiniopCreateTopksoftmaxDescriptor(infiniopHandle_t handle, #endif #ifdef ENABLE_ALI_API CREATE(INFINI_DEVICE_ALI, nvidia); +#endif +#ifdef ENABLE_ILUVATAR_API + CREATE(INFINI_DEVICE_ILUVATAR, nvidia); #endif } @@ -66,6 +69,9 @@ __C infiniStatus_t infiniopGetTopksoftmaxWorkspaceSize(infiniopTopksoftmaxDescri #endif #ifdef ENABLE_ALI_API GET(INFINI_DEVICE_ALI, nvidia); +#endif +#ifdef ENABLE_ILUVATAR_API + GET(INFINI_DEVICE_ILUVATAR, nvidia); #endif } @@ -101,6 +107,9 @@ __C infiniStatus_t infiniopTopksoftmax(infiniopTopksoftmaxDescriptor_t desc, voi #endif #ifdef ENABLE_ALI_API CALCULATE(INFINI_DEVICE_ALI, nvidia); +#endif +#ifdef ENABLE_ILUVATAR_API + CALCULATE(INFINI_DEVICE_ILUVATAR, nvidia); #endif } @@ -131,6 +140,9 @@ __C infiniStatus_t infiniopDestroyTopksoftmaxDescriptor(infiniopTopksoftmaxDescr #endif #ifdef ENABLE_ALI_API DESTROY(INFINI_DEVICE_ALI, nvidia); +#endif +#ifdef ENABLE_ILUVATAR_API + DESTROY(INFINI_DEVICE_ILUVATAR, nvidia); #endif } diff --git a/test/infiniop/ones.py b/test/infiniop/ones.py index dab29d74f..d29d13ce1 100644 --- a/test/infiniop/ones.py +++ b/test/infiniop/ones.py @@ -15,6 +15,7 @@ InfiniDtype, InfiniDtypeNames, InfiniDeviceNames, + InfiniDeviceEnum, infiniopOperatorDescriptor_t, ) from enum import Enum, auto @@ -112,6 +113,12 @@ def test( dtype=None, sync=None, ): + # Skip strided cases on Iluvatar: Ones with non-contiguous tensors can hang the GPU (requires ixsmi -r to recover) + if device == InfiniDeviceEnum.ILUVATAR and ( + x_stride is not None or y_stride is not None + ): + return + if dtype in [InfiniDtype.F16, InfiniDtype.BF16, InfiniDtype.F32, InfiniDtype.F64]: x = TestTensor(shape, x_stride, dtype, device) elif dtype in [InfiniDtype.BYTE, InfiniDtype.U8, InfiniDtype.U16, InfiniDtype.U32, InfiniDtype.U64, diff --git a/test/infiniop/zeros.py b/test/infiniop/zeros.py index 7f6067ade..8c14ca86b 100644 --- a/test/infiniop/zeros.py +++ b/test/infiniop/zeros.py @@ -15,6 +15,7 @@ InfiniDtype, InfiniDtypeNames, InfiniDeviceNames, + InfiniDeviceEnum, infiniopOperatorDescriptor_t, ) from enum import Enum, auto @@ -114,6 +115,12 @@ def test( dtype=None, sync=None, ): + # Skip strided cases on Iluvatar: Zeros with non-contiguous tensors can hang the GPU (requires ixsmi -r to recover) + if device == InfiniDeviceEnum.ILUVATAR and ( + x_stride is not None or y_stride is not None + ): + return + if dtype in [InfiniDtype.F16, InfiniDtype.BF16, InfiniDtype.F32, InfiniDtype.F64]: x = TestTensor(shape, x_stride, dtype, device) elif dtype in [InfiniDtype.BYTE, InfiniDtype.U8, InfiniDtype.U16, InfiniDtype.U32, InfiniDtype.U64, diff --git a/xmake.lua b/xmake.lua index 05ecee7a9..58a0c9fac 100644 --- a/xmake.lua +++ b/xmake.lua @@ -115,10 +115,12 @@ option("iluvatar-gpu") set_description("Whether to compile implementations for Iluvatar GPU") option_end() -option("ivcore-20") - set_default(false) +option("iluvatar_arch") + set_default("ivcore20") set_showmenu(true) - set_description("Use ivcore20") + set_description("Set Iluvatar GPU architecture (e.g. ivcore20)") + set_values("ivcore20") + set_category("option") option_end() if has_config("iluvatar-gpu") then diff --git a/xmake/iluvatar.lua b/xmake/iluvatar.lua index 4c641d459..8cec28e3e 100644 --- a/xmake/iluvatar.lua +++ b/xmake/iluvatar.lua @@ -1,4 +1,6 @@ -toolchain("iluvatar.toolchain") +local iluvatar_arch = get_config("iluvatar_arch") or "ivcore20" + +toolchain("iluvatar.toolchain") set_toolset("cc" , "clang" ) set_toolset("cxx" , "clang++") set_toolset("cu" , "clang++") @@ -44,9 +46,7 @@ target("infiniop-iluvatar") set_warnings("all", "error") add_cuflags("-Wno-error=unused-private-field", "-Wno-error=unused-variable", "-Wno-unused-variable") add_cuflags("-fPIC", "-x", "ivcore", "-std=c++17", {force = true}) - if has_config("ivcore-20") then - add_cuflags("--cuda-gpu-arch=ivcore20", {force = true}) - end + add_cuflags("--cuda-gpu-arch=" .. iluvatar_arch, {force = true}) add_culdflags("-fPIC") add_cxflags("-fPIC", "-Wno-error=unused-variable", "-Wno-unused-variable") add_cxxflags("-fPIC", "-Wno-error=unused-variable", "-Wno-unused-variable") @@ -75,6 +75,7 @@ target("infinirt-iluvatar") set_warnings("all", "error") add_cuflags("-fPIC", "-x", "ivcore", "-std=c++17", {force = true}) + add_cuflags("--cuda-gpu-arch=" .. iluvatar_arch, {force = true}) add_culdflags("-fPIC") add_cxflags("-fPIC") add_cxxflags("-fPIC") @@ -97,6 +98,7 @@ target("infiniccl-iluvatar") set_warnings("all", "error") add_cuflags("-fPIC", "-x", "ivcore", "-std=c++17", {force = true}) + add_cuflags("--cuda-gpu-arch=" .. iluvatar_arch, {force = true}) add_culdflags("-fPIC") add_cxflags("-fPIC") add_cxxflags("-fPIC") From 7377e711bba190eea80d84bc0f143bf2eebf627d Mon Sep 17 00:00:00 2001 From: zhangyue Date: Thu, 12 Feb 2026 05:39:19 +0000 Subject: [PATCH 3/7] issue/1008: adapt paged_attention_prefill --- .../ops/paged_attention/cuda/kernel_v2.cuh | 49 +++++++++++++++++++ .../cuda/kernel_v2.cuh | 30 ++++++------ .../nvidia/paged_attention_prefill_nvidia.cu | 5 ++ 3 files changed, 69 insertions(+), 15 deletions(-) diff --git a/src/infiniop/ops/paged_attention/cuda/kernel_v2.cuh b/src/infiniop/ops/paged_attention/cuda/kernel_v2.cuh index 2b603217b..305820862 100644 --- a/src/infiniop/ops/paged_attention/cuda/kernel_v2.cuh +++ b/src/infiniop/ops/paged_attention/cuda/kernel_v2.cuh @@ -16,17 +16,66 @@ struct OnlineSoftmaxState { } }; __device__ __forceinline__ float warpReduceSum(float x) { +#if defined(ENABLE_ILUVATAR_API) + // Iluvatar may use warp size 64; __shfl_sync(0xffffffff) only covers 32 threads. + // Use shared-memory tree reduce for portability across warp sizes. + constexpr int kMaxWarps = 16; + __shared__ float _reduce_buf[kMaxWarps * 32]; + const int lane = threadIdx.x & 31; + const int warp_id = threadIdx.x / 32; + _reduce_buf[threadIdx.x] = x; + __syncthreads(); + for (int offset = 16; offset > 0; offset >>= 1) { + if (lane < offset) { + _reduce_buf[warp_id * 32 + lane] += _reduce_buf[warp_id * 32 + lane + offset]; + } + __syncthreads(); + } + return _reduce_buf[warp_id * 32]; +#else for (int offset = 16; offset > 0; offset >>= 1) { x += __shfl_down_sync(0xffffffff, x, offset); } return x; +#endif +} + +__device__ __forceinline__ float warpBroadcast(float x, int src_lane) { +#if defined(ENABLE_ILUVATAR_API) + __shared__ float _bcast_buf[16]; + const int warp_id = threadIdx.x / 32; + if ((threadIdx.x & 31) == src_lane) { + _bcast_buf[warp_id] = x; + } + __syncthreads(); + return _bcast_buf[warp_id]; +#else + return __shfl_sync(0xffffffff, x, src_lane); +#endif } __device__ __forceinline__ float warpReduceMax(float x) { +#if defined(ENABLE_ILUVATAR_API) + __shared__ float _reduce_buf[16 * 32]; + const int lane = threadIdx.x & 31; + const int warp_id = threadIdx.x / 32; + _reduce_buf[threadIdx.x] = x; + __syncthreads(); + for (int offset = 16; offset > 0; offset >>= 1) { + if (lane < offset) { + float other = _reduce_buf[warp_id * 32 + lane + offset]; + float cur = _reduce_buf[warp_id * 32 + lane]; + _reduce_buf[warp_id * 32 + lane] = fmaxf(cur, other); + } + __syncthreads(); + } + return _reduce_buf[warp_id * 32]; +#else for (int offset = 16; offset > 0; offset >>= 1) { x = fmaxf(x, __shfl_down_sync(0xffffffff, x, offset)); } return x; +#endif } __device__ __forceinline__ unsigned int cvtaToShared(const void *ptr) { diff --git a/src/infiniop/ops/paged_attention_prefill/cuda/kernel_v2.cuh b/src/infiniop/ops/paged_attention_prefill/cuda/kernel_v2.cuh index 281f918ea..dd7dfdda4 100644 --- a/src/infiniop/ops/paged_attention_prefill/cuda/kernel_v2.cuh +++ b/src/infiniop/ops/paged_attention_prefill/cuda/kernel_v2.cuh @@ -1,7 +1,7 @@ #ifndef __PAGED_ATTENTION_PREFILL_KERNEL_V2_CUH__ #define __PAGED_ATTENTION_PREFILL_KERNEL_V2_CUH__ -#if defined(ENABLE_NVIDIA_API) || defined(ENABLE_ALI_API) +#if defined(ENABLE_NVIDIA_API) || defined(ENABLE_ALI_API) || defined(ENABLE_ILUVATAR_API) #include #include #include @@ -194,8 +194,8 @@ __device__ void PagedAttentionPrefillWarpKernel( l = l * alpha + beta; m = m_new; } - alpha = __shfl_sync(0xffffffff, alpha, 0); - beta = __shfl_sync(0xffffffff, beta, 0); + alpha = op::paged_attention::cuda::warpBroadcast(alpha, 0); + beta = op::paged_attention::cuda::warpBroadcast(beta, 0); #if defined(__CUDA_ARCH__) if constexpr (std::is_same_v) { @@ -233,7 +233,7 @@ __device__ void PagedAttentionPrefillWarpKernel( if (lane == 0) { inv_l = 1.0f / (l + 1e-6f); } - inv_l = __shfl_sync(0xffffffff, inv_l, 0); + inv_l = op::paged_attention::cuda::warpBroadcast(inv_l, 0); #pragma unroll for (int i = 0; i < DIMS_PER_THREAD; ++i) { @@ -411,8 +411,8 @@ __global__ void PagedAttentionPrefillWarpGlobalKernel( l = l * alpha + beta; m = m_new; } - alpha = __shfl_sync(0xffffffff, alpha, 0); - beta = __shfl_sync(0xffffffff, beta, 0); + alpha = op::paged_attention::cuda::warpBroadcast(alpha, 0); + beta = op::paged_attention::cuda::warpBroadcast(beta, 0); #if defined(__CUDA_ARCH__) if constexpr (std::is_same_v) { @@ -450,7 +450,7 @@ __global__ void PagedAttentionPrefillWarpGlobalKernel( if (lane == 0) { inv_l = 1.0f / (l + 1e-6f); } - inv_l = __shfl_sync(0xffffffff, inv_l, 0); + inv_l = op::paged_attention::cuda::warpBroadcast(inv_l, 0); #pragma unroll for (int i = 0; i < DIMS_PER_THREAD; ++i) { @@ -785,8 +785,8 @@ __device__ void PagedAttentionPrefillWarpCtaKernel( l = l * alpha + beta; m = m_new; } - alpha = __shfl_sync(0xffffffff, alpha, 0); - beta = __shfl_sync(0xffffffff, beta, 0); + alpha = op::paged_attention::cuda::warpBroadcast(alpha, 0); + beta = op::paged_attention::cuda::warpBroadcast(beta, 0); #if defined(__CUDA_ARCH__) if constexpr (std::is_same_v) { @@ -826,7 +826,7 @@ __device__ void PagedAttentionPrefillWarpCtaKernel( if (lane == 0) { inv_l = 1.0f / (l + 1e-6f); } - inv_l = __shfl_sync(0xffffffff, inv_l, 0); + inv_l = op::paged_attention::cuda::warpBroadcast(inv_l, 0); #pragma unroll for (int i = 0; i < DIMS_PER_THREAD; ++i) { @@ -1270,7 +1270,7 @@ __device__ void PagedAttentionPrefillWarpCtaKernelPipelined( if (lane == 0) { inv_l = 1.0f / (l + 1e-6f); } - inv_l = __shfl_sync(0xffffffff, inv_l, 0); + inv_l = op::paged_attention::cuda::warpBroadcast(inv_l, 0); #pragma unroll for (int i = 0; i < DIMS_PER_THREAD; ++i) { @@ -1961,8 +1961,8 @@ __device__ void PagedAttentionPrefillWarpCtaKernelKOnly( l = l * alpha + beta; m = m_new; } - alpha = __shfl_sync(0xffffffff, alpha, 0); - beta = __shfl_sync(0xffffffff, beta, 0); + alpha = op::paged_attention::cuda::warpBroadcast(alpha, 0); + beta = op::paged_attention::cuda::warpBroadcast(beta, 0); #if defined(__CUDA_ARCH__) if constexpr (std::is_same_v) { @@ -2002,7 +2002,7 @@ __device__ void PagedAttentionPrefillWarpCtaKernelKOnly( if (lane == 0) { inv_l = 1.0f / (l + 1e-6f); } - inv_l = __shfl_sync(0xffffffff, inv_l, 0); + inv_l = op::paged_attention::cuda::warpBroadcast(inv_l, 0); #pragma unroll for (int i = 0; i < DIMS_PER_THREAD; ++i) { @@ -2131,7 +2131,7 @@ __device__ __forceinline__ void PagedAttentionPrefillMmaScoreWriteRow( if (lane == 0) { inv_l = 1.0f / (l + 1e-6f); } - inv_l = __shfl_sync(0xffffffff, inv_l, 0); + inv_l = op::paged_attention::cuda::warpBroadcast(inv_l, 0); const int64_t q_token = q_start + static_cast(q_token_local); half *out_ptr = out_ + q_token * o_stride + static_cast(head_idx) * o_head_stride; diff --git a/src/infiniop/ops/paged_attention_prefill/nvidia/paged_attention_prefill_nvidia.cu b/src/infiniop/ops/paged_attention_prefill/nvidia/paged_attention_prefill_nvidia.cu index b8e98338a..04107fb4e 100644 --- a/src/infiniop/ops/paged_attention_prefill/nvidia/paged_attention_prefill_nvidia.cu +++ b/src/infiniop/ops/paged_attention_prefill/nvidia/paged_attention_prefill_nvidia.cu @@ -21,6 +21,11 @@ constexpr size_t ceilDiv(size_t a, size_t b) { } inline const char *default_prefill_kernel(const PagedAttentionPrefillInfo &info) { + // Iluvatar: use warp (stable). Users can override via INFINIOP_FLASH_PREFILL_KERNEL. +#ifdef ENABLE_ILUVATAR_API + (void)info; + return "warp"; +#endif // Heuristic auto-dispatch (v0.4): // - Prefer the pipelined + tile-wise softmax kernel on FA2-compatible block_size=256. // - Keep a conservative fallback for other shapes / older GPUs (cp.async is a no-op below SM80). From 034b18953128947ac5bf5d2efc75b002ede9fdca Mon Sep 17 00:00:00 2001 From: zhangyue Date: Thu, 12 Feb 2026 06:23:04 +0000 Subject: [PATCH 4/7] issue/1008 skip scale_mm compile in iluvatar --- xmake/iluvatar.lua | 2 ++ 1 file changed, 2 insertions(+) diff --git a/xmake/iluvatar.lua b/xmake/iluvatar.lua index 8cec28e3e..8fb69e8eb 100644 --- a/xmake/iluvatar.lua +++ b/xmake/iluvatar.lua @@ -53,6 +53,8 @@ target("infiniop-iluvatar") -- set_languages("cxx17") 天数似乎不能用这个配置 add_files("../src/infiniop/devices/nvidia/*.cu", "../src/infiniop/ops/*/nvidia/*.cu") + -- skip scaled_mm, adapt it later + remove_files("../src/infiniop/ops/scaled_mm/nvidia/*.cu") -- 天数平台不支持部分 NVIDIA PTX 指令，AWQ 反量化改用 CUDA C++ 实现 add_files("../src/infiniop/ops/dequantize_awq/iluvatar/*.cu") From 1c32d14d703e0da9f1554ded3751bae65a3cd987 Mon Sep 17 00:00:00 2001 From: zhangyue Date: Thu, 12 Feb 2026 07:37:28 +0000 Subject: [PATCH 5/7] issue/1008: wrap iluvatar change in #ifdef ENABLE_ILUVATAR_API --- scripts/python_test.py | 10 ++--- .../cuda/kernel_v2.cuh | 44 +++++++++++++++++++ .../ops/scaled_mm/nvidia/int8_gemm_nvidia.cu | 2 + xmake/iluvatar.lua | 2 +- 4 files changed, 52 insertions(+), 6 deletions(-) diff --git a/scripts/python_test.py b/scripts/python_test.py index eece1218c..d87763422 100644 --- a/scripts/python_test.py +++ b/scripts/python_test.py @@ -20,7 +20,7 @@ def run_tests(args): #"dequantize_awq.py", "gelu.py", "gemm.py", - "layer_norm.py", + # "layer_norm.py", "logsoftmax.py", "lp_norm.py", "mul.py", @@ -31,7 +31,7 @@ def run_tests(args): "rms_norm.py", "rope.py", "sigmoid.py", - "softmax.py", + # "softmax.py", "softplus.py", "sub.py", "swiglu.py", @@ -39,9 +39,9 @@ def run_tests(args): "topkrouter.py", "topksoftmax.py", "zeros.py", - "paged_attention.py", - "paged_caching.py", - "paged_attention_prefill.py" + # "paged_attention.py", + # "paged_caching.py", + # "paged_attention_prefill.py" ]: result = subprocess.run( f"python {test} {args} --debug", text=True, encoding="utf-8", shell=True diff --git a/src/infiniop/ops/paged_attention_prefill/cuda/kernel_v2.cuh b/src/infiniop/ops/paged_attention_prefill/cuda/kernel_v2.cuh index dd7dfdda4..15b5c6c40 100644 --- a/src/infiniop/ops/paged_attention_prefill/cuda/kernel_v2.cuh +++ b/src/infiniop/ops/paged_attention_prefill/cuda/kernel_v2.cuh @@ -194,8 +194,13 @@ __device__ void PagedAttentionPrefillWarpKernel( l = l * alpha + beta; m = m_new; } +#ifdef ENABLE_ILUVATAR_API alpha = op::paged_attention::cuda::warpBroadcast(alpha, 0); beta = op::paged_attention::cuda::warpBroadcast(beta, 0); +#else + alpha = __shfl_sync(0xffffffff, alpha, 0); + beta = __shfl_sync(0xffffffff, beta, 0); +#endif #if defined(__CUDA_ARCH__) if constexpr (std::is_same_v) { @@ -233,7 +238,11 @@ __device__ void PagedAttentionPrefillWarpKernel( if (lane == 0) { inv_l = 1.0f / (l + 1e-6f); } +#ifdef ENABLE_ILUVATAR_API inv_l = op::paged_attention::cuda::warpBroadcast(inv_l, 0); +#else + inv_l = __shfl_sync(0xffffffff, inv_l, 0); +#endif #pragma unroll for (int i = 0; i < DIMS_PER_THREAD; ++i) { @@ -411,8 +420,13 @@ __global__ void PagedAttentionPrefillWarpGlobalKernel( l = l * alpha + beta; m = m_new; } +#ifdef ENABLE_ILUVATAR_API alpha = op::paged_attention::cuda::warpBroadcast(alpha, 0); beta = op::paged_attention::cuda::warpBroadcast(beta, 0); +#else + alpha = __shfl_sync(0xffffffff, alpha, 0); + beta = __shfl_sync(0xffffffff, beta, 0); +#endif #if defined(__CUDA_ARCH__) if constexpr (std::is_same_v) { @@ -450,7 +464,11 @@ __global__ void PagedAttentionPrefillWarpGlobalKernel( if (lane == 0) { inv_l = 1.0f / (l + 1e-6f); } +#ifdef ENABLE_ILUVATAR_API inv_l = op::paged_attention::cuda::warpBroadcast(inv_l, 0); +#else + inv_l = __shfl_sync(0xffffffff, inv_l, 0); +#endif #pragma unroll for (int i = 0; i < DIMS_PER_THREAD; ++i) { @@ -785,8 +803,13 @@ __device__ void PagedAttentionPrefillWarpCtaKernel( l = l * alpha + beta; m = m_new; } +#ifdef ENABLE_ILUVATAR_API alpha = op::paged_attention::cuda::warpBroadcast(alpha, 0); beta = op::paged_attention::cuda::warpBroadcast(beta, 0); +#else + alpha = __shfl_sync(0xffffffff, alpha, 0); + beta = __shfl_sync(0xffffffff, beta, 0); +#endif #if defined(__CUDA_ARCH__) if constexpr (std::is_same_v) { @@ -826,7 +849,11 @@ __device__ void PagedAttentionPrefillWarpCtaKernel( if (lane == 0) { inv_l = 1.0f / (l + 1e-6f); } +#ifdef ENABLE_ILUVATAR_API inv_l = op::paged_attention::cuda::warpBroadcast(inv_l, 0); +#else + inv_l = __shfl_sync(0xffffffff, inv_l, 0); +#endif #pragma unroll for (int i = 0; i < DIMS_PER_THREAD; ++i) { @@ -1270,7 +1297,11 @@ __device__ void PagedAttentionPrefillWarpCtaKernelPipelined( if (lane == 0) { inv_l = 1.0f / (l + 1e-6f); } +#ifdef ENABLE_ILUVATAR_API inv_l = op::paged_attention::cuda::warpBroadcast(inv_l, 0); +#else + inv_l = __shfl_sync(0xffffffff, inv_l, 0); +#endif #pragma unroll for (int i = 0; i < DIMS_PER_THREAD; ++i) { @@ -1961,8 +1992,13 @@ __device__ void PagedAttentionPrefillWarpCtaKernelKOnly( l = l * alpha + beta; m = m_new; } +#ifdef ENABLE_ILUVATAR_API alpha = op::paged_attention::cuda::warpBroadcast(alpha, 0); beta = op::paged_attention::cuda::warpBroadcast(beta, 0); +#else + alpha = __shfl_sync(0xffffffff, alpha, 0); + beta = __shfl_sync(0xffffffff, beta, 0); +#endif #if defined(__CUDA_ARCH__) if constexpr (std::is_same_v) { @@ -2002,7 +2038,11 @@ __device__ void PagedAttentionPrefillWarpCtaKernelKOnly( if (lane == 0) { inv_l = 1.0f / (l + 1e-6f); } +#ifdef ENABLE_ILUVATAR_API inv_l = op::paged_attention::cuda::warpBroadcast(inv_l, 0); +#else + inv_l = __shfl_sync(0xffffffff, inv_l, 0); +#endif #pragma unroll for (int i = 0; i < DIMS_PER_THREAD; ++i) { @@ -2131,7 +2171,11 @@ __device__ __forceinline__ void PagedAttentionPrefillMmaScoreWriteRow( if (lane == 0) { inv_l = 1.0f / (l + 1e-6f); } +#ifdef ENABLE_ILUVATAR_API inv_l = op::paged_attention::cuda::warpBroadcast(inv_l, 0); +#else + inv_l = __shfl_sync(0xffffffff, inv_l, 0); +#endif const int64_t q_token = q_start + static_cast(q_token_local); half *out_ptr = out_ + q_token * o_stride + static_cast(head_idx) * o_head_stride; diff --git a/src/infiniop/ops/scaled_mm/nvidia/int8_gemm_nvidia.cu b/src/infiniop/ops/scaled_mm/nvidia/int8_gemm_nvidia.cu index c429408f1..36351cf42 100644 --- a/src/infiniop/ops/scaled_mm/nvidia/int8_gemm_nvidia.cu +++ b/src/infiniop/ops/scaled_mm/nvidia/int8_gemm_nvidia.cu @@ -64,6 +64,7 @@ infiniStatus_t Descriptor::create( return INFINI_STATUS_SUCCESS; } +#ifdef ENABLE_QY_API template infiniStatus_t Descriptor::launchKernel(const I8GemmInfo &info, Tdata *y, const Tdata *bias, const int8_t *x_packed, const float *x_scale, const int8_t *w_packed, const float *w_scale, void *stream_, void *workspace) const { cudaStream_t stream = (cudaStream_t)stream_; @@ -112,6 +113,7 @@ infiniStatus_t Descriptor::launchKernel(const I8GemmInfo &info, Tdata *y, const return INFINI_STATUS_SUCCESS; } +#endif infiniStatus_t Descriptor::calculate( void *workspace, diff --git a/xmake/iluvatar.lua b/xmake/iluvatar.lua index 8fb69e8eb..9c3160092 100644 --- a/xmake/iluvatar.lua +++ b/xmake/iluvatar.lua @@ -54,7 +54,7 @@ target("infiniop-iluvatar") -- set_languages("cxx17") 天数似乎不能用这个配置 add_files("../src/infiniop/devices/nvidia/*.cu", "../src/infiniop/ops/*/nvidia/*.cu") -- skip scaled_mm, adapt it later - remove_files("../src/infiniop/ops/scaled_mm/nvidia/*.cu") + -- remove_files("../src/infiniop/ops/scaled_mm/nvidia/*.cu") -- 天数平台不支持部分 NVIDIA PTX 指令，AWQ 反量化改用 CUDA C++ 实现 add_files("../src/infiniop/ops/dequantize_awq/iluvatar/*.cu") From 3d54ce8c73586eae29a732c2ccb7a13a65aa6eb0 Mon Sep 17 00:00:00 2001 From: zhangyue Date: Thu, 12 Feb 2026 07:56:39 +0000 Subject: [PATCH 6/7] issue/1008: revert python_test.py --- scripts/python_test.py | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/scripts/python_test.py b/scripts/python_test.py index d87763422..0bd8bc26d 100644 --- a/scripts/python_test.py +++ b/scripts/python_test.py @@ -20,9 +20,9 @@ def run_tests(args): #"dequantize_awq.py", "gelu.py", "gemm.py", - # "layer_norm.py", + #"layer_norm.py", "logsoftmax.py", - "lp_norm.py", + #"lp_norm.py", "mul.py", "ones.py", "random_sample.py", @@ -31,7 +31,7 @@ def run_tests(args): "rms_norm.py", "rope.py", "sigmoid.py", - # "softmax.py", + #"softmax.py", "softplus.py", "sub.py", "swiglu.py", From 68026bd17b628da26d76ea70c4ac6b8f6ed33957 Mon Sep 17 00:00:00 2001 From: zhangyue Date: Thu, 12 Feb 2026 08:37:25 +0000 Subject: [PATCH 7/7] issue/1008: use warpBroadcast api --- .../cuda/kernel_v2.cuh | 40 ------------------- 1 file changed, 40 deletions(-) diff --git a/src/infiniop/ops/paged_attention_prefill/cuda/kernel_v2.cuh b/src/infiniop/ops/paged_attention_prefill/cuda/kernel_v2.cuh index 15b5c6c40..3fbaf40f8 100644 --- a/src/infiniop/ops/paged_attention_prefill/cuda/kernel_v2.cuh +++ b/src/infiniop/ops/paged_attention_prefill/cuda/kernel_v2.cuh @@ -194,13 +194,8 @@ __device__ void PagedAttentionPrefillWarpKernel( l = l * alpha + beta; m = m_new; } -#ifdef ENABLE_ILUVATAR_API alpha = op::paged_attention::cuda::warpBroadcast(alpha, 0); beta = op::paged_attention::cuda::warpBroadcast(beta, 0); -#else - alpha = __shfl_sync(0xffffffff, alpha, 0); - beta = __shfl_sync(0xffffffff, beta, 0); -#endif #if defined(__CUDA_ARCH__) if constexpr (std::is_same_v) { @@ -238,11 +233,7 @@ __device__ void PagedAttentionPrefillWarpKernel( if (lane == 0) { inv_l = 1.0f / (l + 1e-6f); } -#ifdef ENABLE_ILUVATAR_API inv_l = op::paged_attention::cuda::warpBroadcast(inv_l, 0); -#else - inv_l = __shfl_sync(0xffffffff, inv_l, 0); -#endif #pragma unroll for (int i = 0; i < DIMS_PER_THREAD; ++i) { @@ -420,13 +411,8 @@ __global__ void PagedAttentionPrefillWarpGlobalKernel( l = l * alpha + beta; m = m_new; } -#ifdef ENABLE_ILUVATAR_API alpha = op::paged_attention::cuda::warpBroadcast(alpha, 0); beta = op::paged_attention::cuda::warpBroadcast(beta, 0); -#else - alpha = __shfl_sync(0xffffffff, alpha, 0); - beta = __shfl_sync(0xffffffff, beta, 0); -#endif #if defined(__CUDA_ARCH__) if constexpr (std::is_same_v) { @@ -803,13 +789,8 @@ __device__ void PagedAttentionPrefillWarpCtaKernel( l = l * alpha + beta; m = m_new; } -#ifdef ENABLE_ILUVATAR_API alpha = op::paged_attention::cuda::warpBroadcast(alpha, 0); beta = op::paged_attention::cuda::warpBroadcast(beta, 0); -#else - alpha = __shfl_sync(0xffffffff, alpha, 0); - beta = __shfl_sync(0xffffffff, beta, 0); -#endif #if defined(__CUDA_ARCH__) if constexpr (std::is_same_v) { @@ -849,11 +830,7 @@ __device__ void PagedAttentionPrefillWarpCtaKernel( if (lane == 0) { inv_l = 1.0f / (l + 1e-6f); } -#ifdef ENABLE_ILUVATAR_API inv_l = op::paged_attention::cuda::warpBroadcast(inv_l, 0); -#else - inv_l = __shfl_sync(0xffffffff, inv_l, 0); -#endif #pragma unroll for (int i = 0; i < DIMS_PER_THREAD; ++i) { @@ -1297,11 +1274,7 @@ __device__ void PagedAttentionPrefillWarpCtaKernelPipelined( if (lane == 0) { inv_l = 1.0f / (l + 1e-6f); } -#ifdef ENABLE_ILUVATAR_API inv_l = op::paged_attention::cuda::warpBroadcast(inv_l, 0); -#else - inv_l = __shfl_sync(0xffffffff, inv_l, 0); -#endif #pragma unroll for (int i = 0; i < DIMS_PER_THREAD; ++i) { @@ -1992,13 +1965,8 @@ __device__ void PagedAttentionPrefillWarpCtaKernelKOnly( l = l * alpha + beta; m = m_new; } -#ifdef ENABLE_ILUVATAR_API alpha = op::paged_attention::cuda::warpBroadcast(alpha, 0); beta = op::paged_attention::cuda::warpBroadcast(beta, 0); -#else - alpha = __shfl_sync(0xffffffff, alpha, 0); - beta = __shfl_sync(0xffffffff, beta, 0); -#endif #if defined(__CUDA_ARCH__) if constexpr (std::is_same_v) { @@ -2038,11 +2006,7 @@ __device__ void PagedAttentionPrefillWarpCtaKernelKOnly( if (lane == 0) { inv_l = 1.0f / (l + 1e-6f); } -#ifdef ENABLE_ILUVATAR_API inv_l = op::paged_attention::cuda::warpBroadcast(inv_l, 0); -#else - inv_l = __shfl_sync(0xffffffff, inv_l, 0); -#endif #pragma unroll for (int i = 0; i < DIMS_PER_THREAD; ++i) { @@ -2171,11 +2135,7 @@ __device__ __forceinline__ void PagedAttentionPrefillMmaScoreWriteRow( if (lane == 0) { inv_l = 1.0f / (l + 1e-6f); } -#ifdef ENABLE_ILUVATAR_API inv_l = op::paged_attention::cuda::warpBroadcast(inv_l, 0); -#else - inv_l = __shfl_sync(0xffffffff, inv_l, 0); -#endif const int64_t q_token = q_start + static_cast(q_token_local); half *out_ptr = out_ + q_token * o_stride + static_cast(head_idx) * o_head_stride;