Merge pull request #91 from ryanvergel/2.5

Updated pytorchsubgraphs
Xilinx · Mar 21, 2023 · 9b0e9c3 · 9b0e9c3
2 parents 3990372 + 1739865
commit 9b0e9c3
Show file tree

Hide file tree

Showing 2 changed files with 4 additions and 36 deletions.
diff --git a/Tutorials/pytorch-subgraphs/README.md b/Tutorials/pytorch-subgraphs/README.md
@@ -27,15 +27,11 @@ Author: Daniele Bagni, Xilinx Inc
 </table>
 </div>
 
-### Current status
 
-1. Tested with PyTorch 1.7.1 within [Vitis AI 2.0](https://github.com/Xilinx/Vitis-AI/tree/v2.0) on an Ubuntu 18.04.5 Desktop PC and tested in hardware on VCK190 Production board (``xilinx-vck190-dpu-v2021.2-v2.2.0.img.gz``) and ZCU102 board (``xilinx-zcu102-dpu-v2021.2-v2.0.0.img.gz``) both from the Vitis AI 2.0.
+- Version:      Vitis AI 2.5
+- Support:      ZCU102
+- Last update:  21 Mar. 2023
 
-1. Tested with PyTorch 1.7.1 within [Vitis AI 2.5](https://github.com/Xilinx/Vitis-AI) on an Ubuntu 18.04.5 Desktop PC and tested in hardware on VCK190 Production board (``xilinx-vck190-dpu-v2022.1-v2.5.0.img.gz``) and ZCU102 board (``xilinx-zcu102-dpu-v2022.1-v2.5.0.img.gz``) both from the Vitis AI 2.5.
-
-
-
-#### Date: 20 June 2022
 
 
 

diff --git a/Tutorials/pytorch-subgraphs/files/application/main_subgraphs.py b/Tutorials/pytorch-subgraphs/files/application/main_subgraphs.py
@@ -18,8 +18,7 @@
 """
 
 """
-Author:    Daniele Bagni & Jon Cory
-date:      20 June 2022
+date:      20 Mar 2023
 """
 
 from ctypes import *
@@ -116,33 +115,6 @@ def Linear(xx):
         np.save('cnn_out.bin', y)
     return y
 
-"""
-# yet to be checked
-def LeakyReLU(K, x):
-    data = np.asarray( x, dtype="float32" )
-    print("LKRE inp shape ", data.shape)
-    size  = data[0].size
-    shape = data[0].shape
-    d = data[0].reshape([1, size])
-    print("LKRE inp: ", d)
-    pos_index = (d >= 0);
-    neg_index = (d <  0); Linear(out6)
-    coef = np.float32(K)
-    y_n = d[np.ix_(neg_index)]*coef
-    y_p = d[np.ix_(pos_index)]
-    y = [y_n, y_p]
-    print("LKRE out: ", y)
-    x[0] = y.reshape([shape])
-    return x
-"""
-
-def fix2float(fix_point, value):
-    return value.astype(np.float32) * np.exp2(fix_point, dtype=np.float32)
-
-
-def float2fix(fix_point, value):
-    return value.astype(np.float32) / np.exp2(fix_point, dtype=np.float32)
-
 
 def execute_async(dpu, tensor_buffers_dict):
     input_tensor_buffers = [