Skip to content

GitLab

Switch branch/tag
  1. 05 Dec, 2023 2 commits
  3. 27 Nov, 2023 1 commit
  5. 30 Oct, 2023 1 commit
  7. 28 Sep, 2023 1 commit
    • Umang Yadav's avatar
      Add options to set tolerances inside MIGraphX driver (#2213) · 69d8d789
      Umang Yadav authored 
      MIGraphX verification by default uses normalized RMS error as the basis for the verification.  This change adds some logic to allow migraphx to do "np.allclose" type of elementwise verification using atol and rtol.

Commit also includes changes to consistently pass "gold" or "expected" results as the second argument for "verify_range()" calls.  Default RMS tolerance inside driver is set to 0.001 which IMO is high for FP32 compared to what we had earlier. Need better defaults
      69d8d789
  9. 08 Aug, 2023 1 commit
  11. 08 May, 2023 1 commit
  13. 28 Apr, 2023 1 commit
  15. 25 Apr, 2023 1 commit
  17. 06 Apr, 2023 1 commit
    • Charlie Lin's avatar
      Driver dynamic batch update (#1652) · adccec52
      Charlie Lin authored 
      Examples..

bin/driver verify /codes/onnx_models/resnet50-v1-7/resnet50-v1-7.onnx --split-single-dyn-dim --batch 3 --dyn-input-dim @data "[{min:1, max:4}, 3, 224, 224]"

bin/driver compile /codes/onnx_models/resnet50-v1-7/resnet50-v1-7.onnx --split-single-dyn-dim --default-dyn-dim "{min:1, max:10}" --output resnet50_batch1-10.mxr

bin/driver perf resnet50_batch1-10.mxr --batch 4
      adccec52
  19. 28 Mar, 2023 1 commit
  21. 25 Mar, 2023 1 commit
  23. 16 Feb, 2023 1 commit
  25. 30 Nov, 2022 1 commit
  27. 23 Nov, 2022 1 commit
  29. 03 Oct, 2022 1 commit
    • Umang Yadav's avatar
      Add output_alias and runs_on_offload_target flags for the custom ops (#1309) · c9ffb38d
      Umang Yadav authored 
      Adds two methods for the custom_ops virtual class.

bool runs_on_offload_target(), if the custom op runs directly on the gpu then it should be set to true. in this case, custom op expects its parameters to reside in GPU memory and writes output to the GPU memory. If it is set to false then, custom op expects it's parameter to reside on the host and puts back the result into the host memory.

output_alias, if output of the custom op is aliasing the input buffer. i.e. interpreting the same input buffer with differnet shape and strides.

Update as_vector() in C++ API to handle non-standard shapes. It required exposing element_index to space_index conversion method for the shape class.
      c9ffb38d
  31. 06 Sep, 2022 1 commit
  33. 09 Aug, 2022 1 commit
  35. 29 Jun, 2022 2 commits
  37. 28 Jun, 2022 2 commits
  39. 23 Jun, 2022 2 commits
  41. 22 Jun, 2022 1 commit
  43. 31 May, 2022 1 commit
  45. 27 May, 2022 1 commit
  47. 25 May, 2022 2 commits
  49. 11 May, 2022 1 commit
  51. 14 Apr, 2022 1 commit
  53. 17 Feb, 2022 1 commit
  55. 11 Feb, 2022 1 commit
  57. 08 Feb, 2022 1 commit
  59. 28 Jan, 2022 1 commit
  61. 10 Dec, 2021 1 commit
  63. 25 Nov, 2021 1 commit
  65. 06 Oct, 2021 1 commit
  67. 08 Sep, 2021 1 commit
  69. 07 Sep, 2021 1 commit