2

I am setting up a tensorflow based machine learning workspace on my AMD (RavenRidge APU) laptop using ROCM (Ubuntu HP envy x360 Ryzen 5 2500u)

I have installed rocm and tried using a basic tensorflow script but I get HIP errors. DOes this mean that my GPU is not compatible with rocm?

2019-06-14 21:51:59.142628: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1189] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 7538 MB memory) -> physical GPU (device: 0, name: AMD Ryzen 5 2500U with Radeon Vega Mobile Gfx, pci bus id: 0000:03:00.0)
2019-06-14 21:51:59.221889: E tensorflow/stream_executor/rocm/rocm_driver.cc:601] failed to allocate 7.36G (7905050624 bytes) from device: hipError_t(1002)
2019-06-14 21:51:59.221972: E tensorflow/stream_executor/rocm/rocm_driver.cc:601] failed to allocate 6.62G (7114545152 bytes) from device: hipError_t(1002)
2019-06-14 21:51:59.221997: E tensorflow/stream_executor/rocm/rocm_driver.cc:601] failed to allocate 5.96G (6403090432 bytes) from device: hipError_t(1002)
2019-06-14 21:51:59.222021: E tensorflow/stream_executor/rocm/rocm_driver.cc:601] failed to allocate 5.37G (5762781184 bytes) from device: hipError_t(1002)
2019-06-14 21:51:59.222054: E tensorflow/stream_executor/rocm/rocm_driver.cc:601] failed to allocate 4.83G (5186503168 bytes) from device: hipError_t(1002)
2019-06-14 21:51:59.222090: E tensorflow/stream_executor/rocm/rocm_driver.cc:601] failed to allocate 4.35G (4667852800 bytes) from device: hipError_t(1002)
2019-06-14 21:52:05.830021: E tensorflow/stream_executor/rocm/rocm_driver.cc:461] failed to memset memory: HIP_ERROR_InvalidValue
user9712582
  • 1,215
  • 7
  • 16
Wilf Chun
  • 25
  • 5

2 Answers2

2

I got the same error on Ryzen 5 2400G. Here's github thread discussing the issue.

https://github.com/ROCmSoftwarePlatform/tensorflow-upstream/issues/277

From comment #2

Hi Andreas. Our APUs are currently not supported by various components in the ROCm stack. Details on hardware support can be found here: https://github.com/RadeonOpenCompute/ROCm#supported-gpus

ROCm officially supports AMD GPUs that use following chips:

    GFX8 GPUs
        "Fiji" chips, such as on the AMD Radeon R9 Fury X and Radeon Instinct MI8
        "Polaris 10" chips, such as on the AMD Radeon RX 580 and Radeon Instinct MI6
    GFX9 GPUs
        "Vega 10" chips, such as on the AMD Radeon RX Vega 64 and Radeon Instinct MI25
        "Vega 7nm" chips, such as on the Radeon Instinct MI50, Radeon Instinct MI60 or AMD Radeon VII
Scott Stensland
  • 26,870
  • 12
  • 93
  • 104
Alex Volkov
  • 2,812
  • 23
  • 27
-1

I'm afraid that ROCm-HIP is not currently fully supported because of the BIOS of the mainboard.

plz see this.

https://github.com/RadeonOpenCompute/ROCm/wiki#not-supported-or-limited-support-under-rocm

Negishi Yoshikazu
  • 14