Skip to content
This repository has been archived by the owner on Aug 17, 2021. It is now read-only.

feat/fashion-dropout: adds dropout to fashion example #15

Open
wants to merge 1 commit into
base: master
Choose a base branch
from

Conversation

subversive-owl
Copy link
Contributor

This shouldn't be merged as it is, because it changes the juice dependency to local again, but it helps test the dropout layer.

@drahnr
Copy link
Member

drahnr commented Nov 15, 2017

Does it work if you compile with native?

@subversive-owl
Copy link
Contributor Author

well, it compiles, but i don't think native can support this:

    Finished dev [unoptimized + debuginfo] target(s) in 0.0 secs
     Running `target/debug/juice-examples fashion conv --batch-size 20 --learning-rate 0.002`
thread 'main' panicked at 'called `Result::unwrap()` on an `Err` value: Plugin(Plugin("Unable to create CuDNN TensorDescriptor."))', /checkout/src/libcore/result.rs:906:4
stack backtrace:
  11: <juice::layers::common::dropout::Dropout<f32, B> as juice::layer::ComputeOutput<f32, B>>::compute_output
             at juice/src/layers/common/dropout.rs:74
  12: juice::layer::ILayer::forward
             at juice/src/layer.rs:1015
  13: <juice::layer::Layer<B>>::forward
             at juice/src/layer.rs:506
  14: <juice::layers::container::sequential::Sequential<B> as juice::layer::ILayer<B>>::forward
             at juice/src/layers/container/sequential.rs:272
  15: <juice::layer::Layer<B>>::forward
             at juice/src/layer.rs:506
  16: <juice::solver::Solver<SolverB, B>>::train_minibatch
             at juice/src/solver/mod.rs:76
  17: juice_examples::run_fashion
             at src/main.rs:524

@drahnr
Copy link
Member

drahnr commented Nov 16, 2017

It's still running with cuda.

I am gonna look deeper into this hopefully this weekend.

@subversive-owl
Copy link
Contributor Author

that's strange, i just installed CUDA 9.0 and ran into the same error:

cargo run --features cuda -- fashion conv --batch-size 20 --learning-rate 0.002
    Finished dev [unoptimized + debuginfo] target(s) in 0.0 secs
     Running `target/debug/juice-examples fashion conv --batch-size 20 --learning-rate 0.002`
thread 'main' panicked at 'called `Result::unwrap()` on an `Err` value: Plugin(Plugin("Unable to create CuDNN TensorDescriptor."))', /checkout/src/libcore/result.rs:906:4
note: Run with `RUST_BACKTRACE=1` for a backtrace.
nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2017 NVIDIA Corporation
Built on Fri_Sep__1_21:08:03_CDT_2017
Cuda compilation tools, release 9.0, V9.0.176
nvidia-smi
Sun Nov 19 10:32:49 2017       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 384.90                 Driver Version: 384.90                    |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX 106...  Off  | 00000000:01:00.0  On |                  N/A |
|  0%   37C    P8     7W / 120W |    134MiB /  6064MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

2 participants