search

NervanaSystems / deepspeech

DeepSpeech neon implementation
Apache License 2.0
222 stars 69 forks source link

Can't training on multi gpu devices. #44

Closed w4-kang closed 7 years ago

w4-kang commented 7 years ago

Hi, I have a environment which has multiple gpu devices (1080ti 2way). So, I tried run training with -m 2 options. But it uses only first gpu devices. Even I set parameter max_devices with 2 to gen_backend() statement, it still uses only 1 devices. (with increased batch_size, it throws memory error instead use second device) For your information my environment settings are below: os: ubuntu 16.04 cuda version: 8.0 gpu devices: 1080ti 2way. cpu device: intel skylake i7. virtual env: anaconda (python3).

gardenia22 commented 7 years ago

multiple gpu is not available in this open source version. see issue #5.

w4-kang commented 7 years ago

Oops.. I didn't realize that, Thanks.