Update Python bindings and docstrings. Fix examples not building.

Signed-off-by: Robert Quill <[email protected]>
KomputeProject · Aug 27, 2024 · 650665b · 650665b
1 parent b085445
commit 650665b
Show file tree

Hide file tree

Showing 16 changed files with 145 additions and 334 deletions.
diff --git a/examples/android/android-simple/app/src/main/cpp/KomputeModelML.cpp b/examples/android/android-simple/app/src/main/cpp/KomputeModelML.cpp
@@ -48,13 +48,13 @@ KomputeModelML::train(std::vector<float> yData,
         std::shared_ptr<kp::Algorithm> algorithm = mgr.algorithm(
           params, shader, kp::Workgroup({ 5 }), std::vector<float>({ 5.0 }));
 
-        mgr.sequence()->eval<kp::OpTensorSyncDevice>(params);
+        mgr.sequence()->eval<kp::OpSyncDevice>(params);
 
         std::shared_ptr<kp::Sequence> sq =
           mgr.sequence()
-            ->record<kp::OpTensorSyncDevice>({ wIn, bIn })
+            ->record<kp::OpSyncDevice>({ wIn, bIn })
             ->record<kp::OpAlgoDispatch>(algorithm)
-            ->record<kp::OpTensorSyncLocal>({ wOutI, wOutJ, bOut, lOut });
+            ->record<kp::OpSyncLocal>({ wOutI, wOutJ, bOut, lOut });
 
         // Iterate across all expected iterations
         for (size_t i = 0; i < ITERATIONS; i++) {

diff --git a/examples/array_multiplication/src/main.cpp b/examples/array_multiplication/src/main.cpp
@@ -27,9 +27,9 @@ main()
     std::shared_ptr<kp::Algorithm> algo = mgr.algorithm(params, shader);
 
     mgr.sequence()
-      ->record<kp::OpTensorSyncDevice>(params)
+      ->record<kp::OpSyncDevice>(params)
       ->record<kp::OpAlgoDispatch>(algo)
-      ->record<kp::OpTensorSyncLocal>(params)
+      ->record<kp::OpSyncLocal>(params)
       ->eval();
 
     // prints "Output {  0  4  12  }"

diff --git a/examples/logistic_regression/src/main.cpp b/examples/logistic_regression/src/main.cpp
@@ -32,7 +32,7 @@ main()
                                                         wIn, wOutI, wOutJ,
                                                         bIn, bOut,  lOut };
 
-    mgr.sequence()->eval<kp::OpTensorSyncDevice>(params);
+    mgr.sequence()->eval<kp::OpSyncDevice>(params);
 
     std::vector<uint32_t> spirv2{ 0x1, 0x2 };
 
@@ -44,9 +44,9 @@ main()
 
     std::shared_ptr<kp::Sequence> sq =
       mgr.sequence()
-        ->record<kp::OpTensorSyncDevice>({ wIn, bIn })
+        ->record<kp::OpSyncDevice>({ wIn, bIn })
         ->record<kp::OpAlgoDispatch>(algorithm)
-        ->record<kp::OpTensorSyncLocal>({ wOutI, wOutJ, bOut, lOut });
+        ->record<kp::OpSyncLocal>({ wOutI, wOutJ, bOut, lOut });
 
     // Iterate across all expected iterations
     for (size_t i = 0; i < ITERATIONS; i++) {

diff --git a/examples/neural_network_vgg7/run_vgg7.py b/examples/neural_network_vgg7/run_vgg7.py
@@ -85,7 +85,7 @@
     if i == 0:
         # For first layer, the input isn't on-device yet
         things_to_sync_to_device.append(tensor_in)
-    last_seq.eval_async(kp.OpTensorSyncDevice(things_to_sync_to_device))
+    last_seq.eval_async(kp.OpSyncDevice(things_to_sync_to_device))
     last_seq.eval_await()
 
     # Prepare
@@ -101,7 +101,7 @@
         # DEBUG:
         # We want to see the output, copy it to local
         last_seq = kpm.sequence()
-        last_seq.eval_async(kp.OpTensorSyncLocal([tensor_out]))
+        last_seq.eval_async(kp.OpSyncLocal([tensor_out]))
         last_seq.eval_await()
         tensor_out.data().astype("<f4", "C").tofile("raw" + str(i) + ".bin")
 
@@ -114,7 +114,7 @@
 
 # Download output
 fin_seq = kpm.sequence()
-fin_seq.eval_async(kp.OpTensorSyncLocal([tensor_in]))
+fin_seq.eval_async(kp.OpSyncLocal([tensor_in]))
 fin_seq.eval_await()
 
 # Output

diff --git a/examples/python_naive_matmul/benchmark.py b/examples/python_naive_matmul/benchmark.py
@@ -31,7 +31,7 @@ def main():
 
         for MatMulOp in [MatMulOp1, MatMulOp2, MatMulOp3]:
             tensor_out.data()[:] = 0
-            mgr.sequence().record(kp.OpTensorSyncDevice([tensor_out]))
+            mgr.sequence().record(kp.OpSyncDevice([tensor_out]))
             matmul_op = MatMulOp(mgr)
             matmul_op(tensor_shape, tensor_in_1, tensor_in_2, tensor_out)
 

diff --git a/examples/python_naive_matmul/first_example.py b/examples/python_naive_matmul/first_example.py
@@ -48,9 +48,9 @@ def main():
         [])  # push_consts
 
     (mgr.sequence()
-     .record(kp.OpTensorSyncDevice(params))
+     .record(kp.OpSyncDevice(params))
      .record(kp.OpAlgoDispatch(algo))
-     .record(kp.OpTensorSyncLocal(params))
+     .record(kp.OpSyncLocal(params))
      .eval())
 
     print(f'Output :\n{tensor_out.data().reshape(tensor_shape)}')

diff --git a/examples/python_naive_matmul/imp1_naive.py b/examples/python_naive_matmul/imp1_naive.py
@@ -90,9 +90,9 @@ def __call__(self, tensor_shape: tuple[int, int], tensor_in_1: kp.Tensor, tensor
                 [])  # push_consts
 
         (self.mgr.sequence()
-         .record(kp.OpTensorSyncDevice([tensor_in_1, tensor_in_2]))
+         .record(kp.OpSyncDevice([tensor_in_1, tensor_in_2]))
          .record(kp.OpAlgoDispatch(self.algo))
-         .record(kp.OpTensorSyncLocal([tensor_out]))
+         .record(kp.OpSyncLocal([tensor_out]))
          .eval())
 
 

diff --git a/examples/python_naive_matmul/imp2_tiled.py b/examples/python_naive_matmul/imp2_tiled.py
@@ -89,9 +89,9 @@ def __call__(self, tensor_shape: tuple[int, int], tensor_in_1: kp.Tensor, tensor
                 [])  # push_consts
 
         (self.mgr.sequence()
-         .record(kp.OpTensorSyncDevice([tensor_in_1, tensor_in_2]))
+         .record(kp.OpSyncDevice([tensor_in_1, tensor_in_2]))
          .record(kp.OpAlgoDispatch(self.algo))
-         .record(kp.OpTensorSyncLocal([tensor_out]))
+         .record(kp.OpSyncLocal([tensor_out]))
          .eval())
 
 

diff --git a/examples/python_naive_matmul/imp3_better_tiling.py b/examples/python_naive_matmul/imp3_better_tiling.py
@@ -110,9 +110,9 @@ def __call__(self, tensor_shape: tuple[int, int], tensor_in_1: kp.Tensor, tensor
                 [])  # push_consts
 
         (self.mgr.sequence()
-         .record(kp.OpTensorSyncDevice([tensor_in_1, tensor_in_2]))
+         .record(kp.OpSyncDevice([tensor_in_1, tensor_in_2]))
          .record(kp.OpAlgoDispatch(self.algo))
-         .record(kp.OpTensorSyncLocal([tensor_out]))
+         .record(kp.OpSyncLocal([tensor_out]))
          .eval())