public void SpeedTest() { int inwidth = 32, inchannels = 33, outchannels = 33, ksize = 3, stride = 2; int outwidth = (inwidth - ksize) / stride + 1; OverflowCheckedTensor x_tensor = new OverflowCheckedTensor(Shape.Map1D(inchannels, inwidth)); OverflowCheckedTensor w_tensor = new OverflowCheckedTensor(Shape.Kernel1D(inchannels / 3 * 4, outchannels / 3, ksize)); OverflowCheckedTensor y_tensor = new OverflowCheckedTensor(Shape.Map1D(outchannels, outwidth)); TrivectorConvolution1D ope = new TrivectorConvolution1D(inwidth, inchannels, outchannels, ksize, stride); Stopwatch sw = new Stopwatch(); sw.Start(); ope.Execute(x_tensor, w_tensor, y_tensor); ope.Execute(x_tensor, w_tensor, y_tensor); ope.Execute(x_tensor, w_tensor, y_tensor); ope.Execute(x_tensor, w_tensor, y_tensor); sw.Stop(); Console.WriteLine($"{sw.ElapsedMilliseconds / 4} msec"); }
public void ExecuteTest() { float max_err = 0; foreach (int batch in new int[] { 1, 2, 3 }) { foreach (int inchannels in new int[] { 3, 6, 9, 12 }) { foreach (int outchannels in new int[] { 3, 6, 9, 12 }) { foreach (int kwidth in new int[] { 1, 3, 5 }) { foreach (int stride in new int[] { 1, 2, 3 }) { foreach (int inwidth in new int[] { 8, 9, 13, 17 }) { int outwidth = (inwidth - kwidth) / stride + 1; float[] xval = (new float[inwidth * inchannels * batch]).Select((_, idx) => idx * 1e-3f).ToArray(); float[] wval = (new float[kwidth * inchannels * outchannels / 9 * 4]).Select((_, idx) => idx * 1e-3f).Reverse().ToArray(); Trivector[] xcval = (new Trivector[xval.Length / 3]) .Select((_, idx) => new Trivector(xval[idx * 3], xval[idx * 3 + 1], xval[idx * 3 + 2])).ToArray(); Quaternion.Quaternion[] wcval = (new Quaternion.Quaternion[wval.Length / 4]) .Select((_, idx) => new Quaternion.Quaternion(wval[idx * 4], wval[idx * 4 + 1], wval[idx * 4 + 2], wval[idx * 4 + 3])).ToArray(); TrivectorMap1D x = new TrivectorMap1D(inchannels / 3, inwidth, batch, xcval); Quaternion.QuaternionFilter1D w = new Quaternion.QuaternionFilter1D(inchannels / 3, outchannels / 3, kwidth, wcval); TrivectorMap1D y = Reference(x, w, kwidth, stride); OverflowCheckedTensor x_tensor = new OverflowCheckedTensor(Shape.Map1D(inchannels, inwidth, batch), xval); OverflowCheckedTensor w_tensor = new OverflowCheckedTensor(Shape.Kernel1D(inchannels / 3 * 4, outchannels / 3, kwidth), wval); OverflowCheckedTensor y_tensor = new OverflowCheckedTensor(Shape.Map1D(outchannels, outwidth, batch)); TrivectorConvolution1D ope = new TrivectorConvolution1D(inwidth, inchannels, outchannels, kwidth, stride, gradmode: false, batch); ope.Execute(x_tensor, w_tensor, y_tensor); float[] y_expect = y.ToArray(); float[] y_actual = y_tensor.State; CollectionAssert.AreEqual(xval, x_tensor.State); CollectionAssert.AreEqual(wval, w_tensor.State); AssertError.Tolerance(y_expect, y_actual, 1e-7f, 1e-5f, ref max_err, $"mismatch value {inchannels},{outchannels},{kwidth},{stride},{inwidth},{batch}"); Console.WriteLine($"pass: {inchannels},{outchannels},{kwidth},{stride},{inwidth},{batch}"); } } } } } } Console.WriteLine($"maxerr:{max_err}"); }
public void ExecuteTest() { int inchannels = 9, outchannels = 12, inwidth = 13, kwidth = 3, stride = 2, batch = 7; VariableField x = new Tensor(Shape.Map1D(inchannels, inwidth, batch)); Layer layer = new TrivectorConvolution1D(inchannels, outchannels, kwidth, stride, use_bias: true, pad_mode: PaddingMode.Edge, "conv"); Field y = layer.Forward(x); (Flow flow, Parameters parameters) = Flow.Optimize(y); flow.Execute(); Assert.AreEqual(2, parameters.Count); Assert.AreEqual(inchannels, layer.InChannels); Assert.AreEqual(outchannels, layer.OutChannels); Assert.AreEqual(kwidth, layer.Width); }
public void OverflowTest() { foreach (bool gradmode in new bool[] { false, true }) { foreach (int batch in new int[] { 1, 2, 3 }) { foreach (int inchannels in new int[] { 3, 6, 9, 12 }) { foreach (int outchannels in new int[] { 3, 6, 9, 12 }) { foreach (int kwidth in new int[] { 1, 3, 5 }) { foreach (int stride in new int[] { 1, 2, 3 }) { foreach (int inwidth in new int[] { 8, 9, 13, 17 }) { int outwidth = (inwidth - kwidth) / stride + 1; float[] xval = (new float[inwidth * inchannels * batch]).Select((_, idx) => idx * 1e-3f).ToArray(); float[] wval = (new float[kwidth * inchannels * outchannels / 9 * 4]).Select((_, idx) => idx * 1e-3f).Reverse().ToArray(); OverflowCheckedTensor x_tensor = new OverflowCheckedTensor(Shape.Map1D(inchannels, inwidth, batch), xval); OverflowCheckedTensor w_tensor = new OverflowCheckedTensor(Shape.Kernel1D(inchannels / 3 * 4, outchannels / 3, kwidth), wval); OverflowCheckedTensor y_tensor = new OverflowCheckedTensor(Shape.Map1D(outchannels, outwidth, batch)); TrivectorConvolution1D ope = new TrivectorConvolution1D(inwidth, inchannels, outchannels, kwidth, stride, gradmode, batch); ope.Execute(x_tensor, w_tensor, y_tensor); CollectionAssert.AreEqual(xval, x_tensor.State); CollectionAssert.AreEqual(wval, w_tensor.State); y_tensor.CheckOverflow(); Console.WriteLine($"pass: {inchannels},{outchannels},{kwidth},{stride},{inwidth},{batch},{gradmode}"); } } } } } } } }