Update PyTorch pin (#4629)

Due to new merge
(pytorch/pytorch@47757b2)
in pytorch/pytorch#143553

Signed-off-by: Anatoly Myachev <[email protected]>

Author: anmyachev

.github/pins/pytorch.txt

@@ -1 +1 @@
-64f2ec77f869a7d495694519fb482e9ecaaa6da1
+815545f2dd6ade563cb1263f8bb7813f355edb2e

scripts/patch-pytorch.sh

@@ -35,7 +35,6 @@ apply_patch() {
 echo "Applying PyTorch patches in $REPO_ROOT"
 
 # put your patch applies here
-# trigger build
 apply_patch ./patch/flex_attn_143553.patch
 apply_patch pytorch_fp64.patch
 apply_patch ./patch/pytorch_global_scratch.patch

scripts/patch/flex_attn_143553.patch

@@ -1,5 +1,5 @@
 diff --git a/.ci/docker/common/install_xpu.sh b/.ci/docker/common/install_xpu.sh
-index 51e9df623d5d1..647e77f6d17bd 100644
+index 51e9df623d5d1a..647e77f6d17bdc 100644
 --- a/.ci/docker/common/install_xpu.sh
 +++ b/.ci/docker/common/install_xpu.sh
 @@ -35,12 +35,12 @@ function install_ubuntu() {
@@ -21,7 +21,7 @@ index 51e9df623d5d1..647e77f6d17bd 100644
          apt-get install -y intel-ocloc
      fi
 diff --git a/.ci/docker/ubuntu-xpu/Dockerfile b/.ci/docker/ubuntu-xpu/Dockerfile
-index a0e7dce3df4d5..9cd30e0178bf9 100644
+index a0e7dce3df4d55..9cd30e0178bf92 100644
 --- a/.ci/docker/ubuntu-xpu/Dockerfile
 +++ b/.ci/docker/ubuntu-xpu/Dockerfile
 @@ -63,6 +63,7 @@ RUN rm install_inductor_benchmark_deps.sh common_utils.sh timm.txt huggingface.t
@@ -33,7 +33,7 @@ index a0e7dce3df4d5..9cd30e0178bf9 100644
  RUN bash ./install_xpu.sh && rm install_xpu.sh
 
 diff --git a/test/inductor/test_flex_attention.py b/test/inductor/test_flex_attention.py
-index e6b46ae8edc9d..f8c8d9f31cde3 100644
+index 4d14555800c8c4..ab40d19d2ff5ee 100644
 --- a/test/inductor/test_flex_attention.py
 +++ b/test/inductor/test_flex_attention.py
 @@ -41,20 +41,26 @@
@@ -525,7 +525,7 @@ index e6b46ae8edc9d..f8c8d9f31cde3 100644
  if __name__ == "__main__":
      from torch._inductor.test_case import run_tests
 diff --git a/test/inductor/test_flex_decoding.py b/test/inductor/test_flex_decoding.py
-index 2125e78eea8db..f59433c7e77f2 100644
+index 3b4905fc356168..3a165e5fff2eda 100644
 --- a/test/inductor/test_flex_decoding.py
 +++ b/test/inductor/test_flex_decoding.py
 @@ -27,6 +27,7 @@
@@ -691,7 +691,7 @@ index 2125e78eea8db..f59433c7e77f2 100644
  if __name__ == "__main__":
      from torch._inductor.test_case import run_tests
 diff --git a/torch/_inductor/kernel/flex_attention.py b/torch/_inductor/kernel/flex_attention.py
-index 99e869dc8fdb7..1426f000d191d 100644
+index 99e869dc8fdb71..1426f000d191d2 100644
 --- a/torch/_inductor/kernel/flex_attention.py
 +++ b/torch/_inductor/kernel/flex_attention.py
 @@ -1441,7 +1441,9 @@ def flex_attention(
@@ -717,7 +717,7 @@ index 99e869dc8fdb7..1426f000d191d 100644
      # Default config for warp specialization
      num_consumer_groups, num_buffers_warp_spec = 0, 0
 diff --git a/torch/_inductor/kernel/flex_decoding.py b/torch/_inductor/kernel/flex_decoding.py
-index 7e0aef9818560..343086e5b2d16 100644
+index 7e0aef98185603..343086e5b2d16a 100644
 --- a/torch/_inductor/kernel/flex_decoding.py
 +++ b/torch/_inductor/kernel/flex_decoding.py
 @@ -310,7 +310,10 @@ def flex_decoding_grid(batch_size, kv_heads, gqa_group_size, n_keys, d_model, me
@@ -744,7 +744,7 @@ index 7e0aef9818560..343086e5b2d16 100644
      # TODO: fix autotuning.
 
 diff --git a/torch/_inductor/template_heuristics.py b/torch/_inductor/template_heuristics.py
-index dfd37523a3702..b830ec6369a9d 100644
+index dfd37523a37027..b830ec6369a9d7 100644
 --- a/torch/_inductor/template_heuristics.py
 +++ b/torch/_inductor/template_heuristics.py
 @@ -1178,3 +1178,87 @@ class XPUConfigHeuristic(BaseConfigHeuristic):
@@ -836,7 +836,7 @@ index dfd37523a3702..b830ec6369a9d 100644
 +
 +        return flex_decode_configs
 diff --git a/torch/_ops.py b/torch/_ops.py
-index 337b9a11e6a18..5e3423285e02b 100644
+index 337b9a11e6a180..5e3423285e02b5 100644
 --- a/torch/_ops.py
 +++ b/torch/_ops.py
 @@ -267,6 +267,7 @@ def resolve_key(op: OperatorBase, k: DispatchKey):  # type: ignore[valid-type]
@@ -848,7 +848,7 @@ index 337b9a11e6a18..5e3423285e02b 100644
 
 
 diff --git a/torch/nn/attention/flex_attention.py b/torch/nn/attention/flex_attention.py
-index 15a00e1a9d342..3c9b3a2017399 100644
+index 15a00e1a9d342b..3c9b3a20173997 100644
 --- a/torch/nn/attention/flex_attention.py
 +++ b/torch/nn/attention/flex_attention.py
 @@ -1142,11 +1142,8 @@ def _validate_device(query: Tensor, key: Tensor, value: Tensor):
@@ -866,7 +866,7 @@ index 15a00e1a9d342..3c9b3a2017399 100644
              "FlexAttention is only supported on CUDA, CPU or HPU devices. "
              f"Found input tensors on {query.device.type} device."
 diff --git a/torch/testing/_internal/common_device_type.py b/torch/testing/_internal/common_device_type.py
-index 01499280da8f5..6a5951fde65dc 100644
+index 01499280da8f5d..6a5951fde65dca 100644
 --- a/torch/testing/_internal/common_device_type.py
 +++ b/torch/testing/_internal/common_device_type.py
 @@ -1342,8 +1342,8 @@ def dep_fn(self, *args, **kwargs):