- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
name: "suma"
input: "data"
input_shape {
dim: 1
dim:56
dim: 7
dim: 7
}
layer {
name: "relu1"
type: "PReLU"
bottom: "data"
top: "relu1"
prelu_param {
channel_shared: 1
}
}
and mvNCCheck report
Result: (56, 7, 7)
1) 2514 52.438
2) 2518 48.062
3) 2512 45.562
4) 2536 44.5
5) 2513 42.688
Expected: (56, 7, 7)
1) 2577 0.99951
2) 633 0.99854
3) 1372 0.99805
4) 1540 0.99707
5) 318 0.99365
------------------------------------------------------------
Obtained values
------------------------------------------------------------
Obtained Min Pixel Accuracy: 1036481.25% (max allowed=2%), Fail
Obtained Average Pixel Accuracy: 4342.446136474609% (max allowed=1%), Fail
Obtained Percentage of wrong values: 44.059766763848394% (max allowed=0%), Fail
Obtained Pixel-wise L2 error: 54063.709668603464% (max allowed=1%), Fail
Obtained Global Sum Difference: 119098.546875
------------------------------------------------------------
If I changed channel_shared: 1 to channel_shared: 0, to disable share a between channels, mvNCCheck passed.
Myraid2 didn't support pReLU with channel_shared enabled?
Link Copied
0 Replies
Reply
Topic Options
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page