Update PyTorch pin to the version which support elapsed_time; remove our patch for elapsed_time (#2952)

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

Author: anmyachev

.github/pins/pytorch-upstream.txt

@@ -1 +1 @@
-d0fd42eb3ac6939e63879bc055b6901103f713d3
+61dc5e9c0a36d590adc47b4110efd94d9eb59306

python/tutorials/08-grouped-gemm.py

@@ -171,9 +171,9 @@ def group_gemm_fn(group_A, group_B):
         g_lds += [A.stride(0), B.stride(0), C.stride(0)]
 
     # note these are device tensors
-    d_a_ptrs = torch.tensor(A_addrs, device=DEVICE)
-    d_b_ptrs = torch.tensor(B_addrs, device=DEVICE)
-    d_c_ptrs = torch.tensor(C_addrs, device=DEVICE)
+    d_a_ptrs = torch.tensor(A_addrs, device=DEVICE, dtype=torch.uint64)
+    d_b_ptrs = torch.tensor(B_addrs, device=DEVICE, dtype=torch.uint64)
+    d_c_ptrs = torch.tensor(C_addrs, device=DEVICE, dtype=torch.uint64)
     d_g_sizes = torch.tensor(g_sizes, dtype=torch.int32, device=DEVICE)
     d_g_lds = torch.tensor(g_lds, dtype=torch.int32, device=DEVICE)
     # we use a fixed number of CTA, and it's auto-tunable
@@ -277,9 +277,9 @@ def benchmark(N, provider):
         g_sizes += [N, N, N]
         g_lds += [N, N, N]
 
-    d_a_ptrs = torch.tensor(A_addrs, device=DEVICE)
-    d_b_ptrs = torch.tensor(B_addrs, device=DEVICE)
-    d_c_ptrs = torch.tensor(C_addrs, device=DEVICE)
+    d_a_ptrs = torch.tensor(A_addrs, device=DEVICE, dtype=torch.uint64)
+    d_b_ptrs = torch.tensor(B_addrs, device=DEVICE, dtype=torch.uint64)
+    d_c_ptrs = torch.tensor(C_addrs, device=DEVICE, dtype=torch.uint64)
     d_g_sizes = torch.tensor(g_sizes, dtype=torch.int32, device=DEVICE)
     d_g_lds = torch.tensor(g_lds, dtype=torch.int32, device=DEVICE)
 

scripts/patch-pytorch.sh

@@ -16,4 +16,3 @@ echo "Applying PyTorch patches in $REPO_ROOT"
 cd "$REPO_ROOT"
 
 curl -sSL https://github.com/pytorch/pytorch/pull/126516.diff | git apply -
-curl -sSL https://github.com/pytorch/pytorch/pull/126456.diff | git apply -