Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

add change log for 6.3 #3386

Open
wants to merge 3 commits into
base: develop
Choose a base branch
from
+22 −7
Open

add change log for 6.3 #3386

Changes from 2 commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
c65ff10
add change log for 6.3
bghimireamd Nov 15, 2024
668af36
fix typo
bghimireamd Nov 15, 2024
cbca225
fix review comments
bghimireamd Nov 15, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
27 changes: 20 additions & 7 deletions CHANGELOG.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,7 +3,21 @@

Full documentation for MIOpen is available [here](https://rocm.docs.amd.com/projects/MIOpen/en/latest/)

## MIOpen-3.2.0 for ROCm 6.2.0
## MIOpen 3.3.0 for ROCm 6.3.0
### Added
- [RNN] LSTM fwd
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Should this be forward or forwarding? Or is fwd okay?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think fwd is ok

- [Mha] Mask is added for Forward pass
- [GLU] Gated Linear Unit (experimental)
- [PReLU] Implement PReLU backward (experimental)
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I believe PReLU Backward is an official name, so Backward s/b capitalized. Can you confirm?


### Fixed
- Fixed stream not being set when calling hipMemsetAsync
- Fixed memory leak issue caused by incorrect transpose in find 2.0 (#3285)
- Fixed memcopy data race by replacing hipMemcpy with hipMemcpyWithStream
## Perfomance
- MI300 TunaNet Update: CK FWD and WRW Solvers Updated

## MIOpen 3.2.0 for ROCm 6.2.0
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
## MIOpen 3.3.0 for ROCm 6.3.0
### Added
- [RNN] LSTM fwd
- [Mha] Mask is added for Forward pass
- [GLU] Gated Linear Unit (experimental)
- [PReLU] Implement PReLU backward (experimental)
### Fixed
- Fixed stream not being set when calling hipMemsetAsync
- Fixed memory leak issue caused by incorrect transpose in find 2.0 (#3285)
- Fixed memcopy data race by replacing hipMemcpy with hipMemcpyWithStream
## Perfomance
- MI300 TunaNet Update: CK FWD and WRW Solvers Updated
## MIOpen 3.2.0 for ROCm 6.2.0
## MIOpen 3.3.0 for ROCm 6.3.0
### Added
* [RNN] LSTM fwd
* [Mha] Mask is added for Forward pass
* [GLU] Gated Linear Unit (this is an experimental feature)
* [PReLU] Implemented PReLU backward (this is an experimental feature)
### Optimized
- MI300 TunaNet Update: CK FWD and WRW Solvers Updated
### Resolved issues
- Fixed unset stream when calling `hipMemsetAsync`
- Fixed a memory leak issue caused by an incorrect transpose in find 2.0 (see issue #3285 on GitHub)
- Fixed a `memcopy` data race by replacing `hipMemcpy` with `hipMemcpyWithStream`
## MIOpen 3.2.0 for ROCm 6.2.0

### Added
- [Conv] bilinear (alpha beta) solvers
- [Conv] enable bf16 for ck-based solvers
Expand All @@ -12,7 +26,6 @@ Full documentation for MIOpen is available [here](https://rocm.docs.amd.com/proj
- [RNN] multi-stream as default solution.
- TunaNetv2.0 for MI300
- Added adam and amp adam optimizer

### Fixed
- Memory access fault caused by GemmBwdRest
- Context configuration in GetWorkSpaceSize
Expand All @@ -21,7 +34,7 @@ Full documentation for MIOpen is available [here](https://rocm.docs.amd.com/proj
### Performance
- Find: Improve precision of benchmarking

## MIOpen-3.1.0 for ROCm 6.1.0
## MIOpen 3.1.0 for ROCm 6.1.0
### Added
- CK-based 2d/3d convolution solvers to support nchw/ncdhw layout
- Fused solver for Fwd Convolution with Residual, Bias and activation
Expand All @@ -40,7 +53,7 @@ Full documentation for MIOpen is available [here](https://rocm.docs.amd.com/proj
### Removed
- clamping to MAX from CastTensor used in Bwd and WrW convolution

## MIOpen-3.0.0 for ROCm 6.0.0
## MIOpen 3.0.0 for ROCm 6.0.0
- This release adds 3D convolution, enablement of fp8 convolution, NHWC batch norm, RNN padding support. It also removes
INI8x4 support and fix minor issues and bugs.
### Notes
Expand All @@ -63,7 +76,7 @@ INI8x4 support and fix minor issues and bugs.
- Remove INT8x4 support
- Remove target ids from kdb args

## MIOpen-2.21.0 for ROCm 5.7.0
## MIOpen 2.21.0 for ROCm 5.7.0
### Added
- AI Heuristic for Immediate Mode Fallback
- CK group forward convolution integration
Expand All @@ -73,7 +86,7 @@ INI8x4 support and fix minor issues and bugs.
### Changed
- Update fdb data to use solver keys [MI100][MI200]

## MIOpen-2.20.0 for ROCm 5.6.0
## MIOpen 2.20.0 for ROCm 5.6.0
### Added
- AI Based Heuristic for Kernel Parameter Prediction
- LSTM multi-stream solver
Expand All @@ -83,7 +96,7 @@ INI8x4 support and fix minor issues and bugs.
- Winograd kernel failure
- Perf DB updates for gfx908 and gfx90a

## MIOpen-2.19.0 for ROCm 5.5.0
## MIOpen 2.19.0 for ROCm 5.5.0
### Added
- ROCm 5.5 support for gfx1101 (Navi32)
### Changed
Expand Down