[pt][quant] Support either min or max in qclamp (pytorch#45937)

Summary: Pull Request resolved: pytorch#45937 torch.clamp can now be used with quantized tensors with either min argument or max argument only Fixes pytorch#45928 ghstack-source-id: 114085914 Test Plan: buck test mode/dev caffe2/test:quantization -- 'test_qclamp' --print-passing-details ``` Started reporting to test run: https://our.intern.facebook.com/intern/testinfra/testrun/4222124686876909 ✓ ListingSuccess: caffe2/test:quantization - main (7.602) ✓ Pass: caffe2/test:quantization - test_qclamp (quantization.test_quantized_op.TestQuantizedOps) (7.233) Summary Pass: 1 ListingSuccess: 1 Finished test run: https://our.intern.facebook.com/intern/testinfra/testrun/4222124686876909 ``` Reviewed By: jerryzh168 Differential Revision: D24153431 fbshipit-source-id: 9735635a48bcdd88d1dd6dc2f18b59311d45ad90
lcy-seso · Oct 12, 2020 · 87a4baf · 87a4baf
1 parent bed3b40
commit 87a4baf
Show file tree

Hide file tree

Showing 5 changed files with 99 additions and 12 deletions.
diff --git a/aten/src/ATen/native/quantized/cpu/kernels/QuantizedOpKernels.cpp b/aten/src/ATen/native/quantized/cpu/kernels/QuantizedOpKernels.cpp
@@ -629,6 +629,56 @@ void qclamp_kernel(
   });
 }
 
+void qclamp_min_kernel(const Tensor& qx, Scalar min_scalar, Tensor& qy) {
+  AT_DISPATCH_QINT_TYPES(qx.scalar_type(), "qclamp", [&]() {
+    qy = at::_empty_affine_quantized(
+        qx.sizes(),
+        at::device(kCPU)
+            .dtype(SCALAR_TYPE)
+            .memory_format(qx.suggest_memory_format()),
+        qx.q_scale(),
+        qx.q_zero_point(),
+        c10::nullopt);
+    using Vec = Vec256<scalar_t>;
+    auto iter = TensorIterator::unary_op(qy, qx);
+    auto min = min_scalar.to<float>();
+    scalar_t min_q = at::native::quantize_val<scalar_t>(
+        qx.q_scale(), qx.q_zero_point(), min);
+    auto min_vec = Vec(min_q);
+    cpu_kernel_vec(
+        iter,
+        [&](scalar_t value) -> scalar_t {
+          return scalar_t(std::max<underlying_t>(value.val_, min_q.val_));
+        },
+        [&](Vec val) -> Vec { return val.maximum(min_vec); });
+  });
+}
+
+void qclamp_max_kernel(const Tensor& qx, Scalar max_scalar, Tensor& qy) {
+  AT_DISPATCH_QINT_TYPES(qx.scalar_type(), "qclamp", [&]() {
+    qy = at::_empty_affine_quantized(
+        qx.sizes(),
+        at::device(kCPU)
+            .dtype(SCALAR_TYPE)
+            .memory_format(qx.suggest_memory_format()),
+        qx.q_scale(),
+        qx.q_zero_point(),
+        c10::nullopt);
+    using Vec = Vec256<scalar_t>;
+    auto iter = TensorIterator::unary_op(qy, qx);
+    auto max = max_scalar.to<float>();
+    scalar_t max_q = at::native::quantize_val<scalar_t>(
+        qx.q_scale(), qx.q_zero_point(), max);
+    auto max_vec = Vec(max_q);
+    cpu_kernel_vec(
+        iter,
+        [&](scalar_t value) -> scalar_t {
+          return scalar_t(std::min<underlying_t>(value.val_, max_q.val_));
+        },
+        [&](Vec val) -> Vec { return val.minimum(max_vec); });
+  });
+}
+
 void qthreshold_kernel(
   // TODO: For future tasks, since output quantization parameters are set equal to
   // the input ones, it might make sense to implement this completely in the
@@ -2811,6 +2861,8 @@ REGISTER_DISPATCH(qbatch_norm_stub, &q_batch_norm_kernel<false>);
 REGISTER_DISPATCH(qcat_nhwc_stub, &qcat_nhwc_kernel<false>);
 REGISTER_DISPATCH(qcat_relu_nhwc_stub, &qcat_nhwc_kernel<true>);
 REGISTER_DISPATCH(qclamp_stub, &qclamp_kernel);
+REGISTER_DISPATCH(qclamp_min_stub, &qclamp_min_kernel);
+REGISTER_DISPATCH(qclamp_max_stub, &qclamp_max_kernel);
 REGISTER_DISPATCH(qelu_stub, &qelu_kernel);
 REGISTER_DISPATCH(qhardsigmoid_stub, &qhardsigmoid_kernel);
 REGISTER_DISPATCH(qhardswish_stub, &qhardswish_kernel);

diff --git a/aten/src/ATen/native/quantized/cpu/qclamp.cpp b/aten/src/ATen/native/quantized/cpu/qclamp.cpp
@@ -15,6 +15,8 @@ namespace at {
 namespace native {
 
 DEFINE_DISPATCH(qclamp_stub);
+DEFINE_DISPATCH(qclamp_min_stub);
+DEFINE_DISPATCH(qclamp_max_stub);
 
 namespace {
 
@@ -84,14 +86,26 @@ Tensor quantized_clamp_impl(
   Tensor qy;
   if (min && max) {
 #ifdef USE_PYTORCH_QNNPACK
-    if (at::globalContext().qEngine() == at::QEngine::QNNPACK && qx.scalar_type() == kQUInt8) {
+    if (at::globalContext().qEngine() == at::QEngine::QNNPACK &&
+        qx.scalar_type() == kQUInt8) {
       return qnnpack_clamp(qx, *min, *max);
     }
 #endif
     qclamp_stub(qx.device().type(), qx, *min, *max, qy);
   } else {
-    TORCH_CHECK(
-        false, "Both min and max should be specified for quantized clamp!");
+#ifdef USE_PYTORCH_QNNPACK
+    if (at::globalContext().qEngine() == at::QEngine::QNNPACK) {
+      TORCH_CHECK(
+          false, "Both min and max should be specified for quantized clamp!");
+    }
+#endif
+    if (max) {
+      qclamp_max_stub(qx.device().type(), qx, *max, qy);
+    } else if (min) {
+      qclamp_min_stub(qx.device().type(), qx, *min, qy);
+    } else {
+      TORCH_CHECK(false, "At least one of 'min' or 'max' must not be None");
+    }
   }
   return qy;
 }

diff --git a/aten/src/ATen/native/quantized/cpu/quantized_ops.h b/aten/src/ATen/native/quantized/cpu/quantized_ops.h
@@ -15,6 +15,10 @@ using qclamp_fn = void (*)(
     Scalar min,
     Scalar max,
     at::Tensor& /*qy*/);
+using qclamp_minmax_fn = void (*)(
+    const at::Tensor& /*qx*/,
+    Scalar /*min or max*/,
+    at::Tensor& /*qy*/);
 using qthreshold_fn = void (*)(
     const at::Tensor& /*qx*/,
     Scalar threshold,
@@ -167,6 +171,8 @@ DECLARE_DISPATCH(qbinary_fn, qmul_stub);
 DECLARE_DISPATCH(qcat_nhwc_fn, qcat_nhwc_stub);
 DECLARE_DISPATCH(qcat_nhwc_fn, qcat_relu_nhwc_stub);
 DECLARE_DISPATCH(qclamp_fn, qclamp_stub);
+DECLARE_DISPATCH(qclamp_minmax_fn, qclamp_min_stub);
+DECLARE_DISPATCH(qclamp_minmax_fn, qclamp_max_stub);
 DECLARE_DISPATCH(qelu_fn, qelu_stub);
 DECLARE_DISPATCH(qhardsigmoid_fn, qhardsigmoid_stub);
 DECLARE_DISPATCH(qhardswish_fn, qhardswish_stub);

diff --git a/aten/src/ATen/native/quantized/library.cpp b/aten/src/ATen/native/quantized/library.cpp
@@ -54,7 +54,7 @@ TORCH_LIBRARY(quantized, m) {
   m.def(TORCH_SELECTIVE_SCHEMA("quantized::batch_norm2d_relu(Tensor qx, Tensor? weight, Tensor? bias, Tensor mean, Tensor var, float eps, float output_scale, int output_zero_point) -> Tensor"));
   m.def(TORCH_SELECTIVE_SCHEMA("quantized::batch_norm3d(Tensor qx, Tensor? weight, Tensor? bias, Tensor mean, Tensor var, float eps, float output_scale, int output_zero_point) -> Tensor"));
   m.def(TORCH_SELECTIVE_SCHEMA("quantized::batch_norm3d_relu(Tensor qx, Tensor? weight, Tensor? bias, Tensor mean, Tensor var, float eps, float output_scale, int output_zero_point) -> Tensor"));
-  m.def(TORCH_SELECTIVE_SCHEMA("quantized::clamp(Tensor qx, Scalar? min, Scalar? max) -> Tensor qy"));
+  m.def(TORCH_SELECTIVE_SCHEMA("quantized::clamp(Tensor qx, Scalar? min=None, Scalar? max=None) -> Tensor qy"));
   m.def(TORCH_SELECTIVE_SCHEMA("quantized::threshold(Tensor qx, Scalar threshold, Scalar value) -> Tensor qy"));
   m.def(TORCH_SELECTIVE_SCHEMA("quantized::cat(Tensor[] qx, int dim, float? scale, int? zero_point) -> Tensor"));
   m.def(TORCH_SELECTIVE_SCHEMA("quantized::cat_relu(Tensor[] qx, int dim, float? scale, int? zero_point) -> Tensor"));

diff --git a/test/quantization/test_quantized_op.py b/test/quantization/test_quantized_op.py
@@ -575,22 +575,37 @@ def test_qclamp(self, X, min_val, max_val):
         X, (scale, zero_point, torch_type) = X
 
         assume(min_val <= max_val)
-        Y = X.copy()
-        Y[Y < min_val] = min_val
-        Y[Y > max_val] = max_val
-        qY = torch.quantize_per_tensor(torch.from_numpy(Y), scale=scale,
-                                       zero_point=zero_point, dtype=torch_type)
+        Y_clamp = torch.clamp(torch.from_numpy(X), min=min_val, max=max_val)
+        qY_clamp = torch.quantize_per_tensor(Y_clamp, scale=scale,
+                                             zero_point=zero_point, dtype=torch_type)
+
         X = torch.from_numpy(X)
         qX = torch.quantize_per_tensor(X, scale=scale, zero_point=zero_point,
                                        dtype=torch_type)
-
         ops_under_test = {
             'ops.quantized': torch.ops.quantized.clamp,
         }
 
         for name, op in ops_under_test.items():
-            qY_hat = op(qX, min_val, max_val)
-            self.assertEqual(qY, qY_hat, msg="{} qclamp failed".format(name))
+            qY_clamp_hat = op(qX, min=min_val, max=max_val)
+            self.assertEqual(qY_clamp, qY_clamp_hat, msg="{} qclamp failed".format(name))
+
+        if torch.backends.quantized.engine == 'fbgemm':
+            with override_quantized_engine('fbgemm'):
+                Y_min_clamp = torch.clamp(X, min=min_val)
+                Y_max_clamp = torch.clamp(X, max=max_val)
+
+                qY_min_clamp = torch.quantize_per_tensor(Y_min_clamp, scale=scale,
+                                                         zero_point=zero_point, dtype=torch_type)
+                qY_max_clamp = torch.quantize_per_tensor(Y_max_clamp, scale=scale,
+                                                         zero_point=zero_point, dtype=torch_type)
+
+
+                for name, op in ops_under_test.items():
+                    qY_min_clamp_hat = op(qX, min=min_val)
+                    self.assertEqual(qY_min_clamp, qY_min_clamp_hat, msg="{} qclamp failed".format(name))
+                    qY_max_clamp_hat = op(qX, max=max_val)
+                    self.assertEqual(qY_max_clamp, qY_max_clamp_hat, msg="{} qclamp failed".format(name))
 
     """Tests the correctness of the quantized::hardtanh op."""
     @skipIfNoFBGEMM