Ifu rocm enabled 09 23 2025 - #90

Merged

lcskrishna merged 63 commits into

rocm_enabledfrom

IFU-rocm_enabled-09-23-2025

Sep 24, 2025

Commits on Jun 20, 2025

Enable ROCm backend with custom ops integration (#1683)

authored
888788d

Commits on Jun 23, 2025

Fix AdamW documentation (#1686)
Aman Gupta
authored
fd2949a

Commits on Jun 24, 2025

Make minor improvements to optimizer.py (#1687)
Aman Gupta
authored
aca9778

Commits on Jun 27, 2025

Add CUDA 12.9 build (#1689)
matthewdouglas
authored
1abd5e7

Commits on Jun 30, 2025

Temporarily disable HPU tests
matthewdouglas
authored
6d0a5cd

Commits on Jul 1, 2025

Commits on Jul 2, 2025

Added inference benchmark
Egor-Krivov
committed
3278614

Commits on Jul 3, 2025

fix log
jiqing-feng
committed
ea4b59f

Commits on Jul 8, 2025

Commits on Jul 11, 2025

Add interface for 8bit optimizer
Egor-Krivov
committed
b43edf5
Fixed bugs
Egor-Krivov
committed
35ce337

Commits on Jul 14, 2025

enabled tests
Egor-Krivov
committed
abf4a1e
Add 32bit optimizer interface
Egor-Krivov
committed
3b89a05
Add no_cpu for optimizers
Egor-Krivov
committed
223fea5
Update to kernel registration
Egor-Krivov
committed
4075a64
Reverse lion
Egor-Krivov
committed
236124e
Changed number of errors
Egor-Krivov
committed
36f5c4f
Removed cpu
Egor-Krivov
committed
24d9139
Added mutated args to the schema
Egor-Krivov
committed
e33ba1c
Fixed default args
Egor-Krivov
committed
0f6fe6b
Merge pull request #1706 from Egor-Krivov/egor/8bit_int
matthewdouglas
authored
941681d
Test fix
matthewdouglas
committed
14147f6

Commits on Jul 21, 2025

Commits on Jul 31, 2025

Fix Params4bit tensor subclass handling
ved1beta
committed
1dbe602

Commits on Aug 2, 2025

Fixing quantization uint8 packing bug for NF4 and FP4
Mhmd-Hisham
committed
639f8c0
test_params4bit_torch_chunk_split
ved1beta
committed
2938c73

Commits on Aug 4, 2025

lint
ved1beta
committed
0ecb8fb

Commits on Aug 6, 2025

Commits on Aug 11, 2025

Temporary updates for release
matthewdouglas
committed
5959389
Release 0.47.0
matthewdouglas
committed
c0dcdf2
Bump dev version
matthewdouglas
committed
9088107
Restore temporary changes from release
matthewdouglas
committed
7bfe923

Commits on Aug 25, 2025

add py.typed (#1726)
cyyever
authored
ff389db

Commits on Sep 2, 2025

Enable F841 (#1727)
cyyever
authored
c76e208

Commits on Sep 3, 2025

Commits on Sep 8, 2025

Commits on Sep 9, 2025

Test improvements (#1750)
matthewdouglas
authored
6a07ffe

Commits on Sep 15, 2025

Lint fix
matthewdouglas
committed
d848d4d
[XPU] Implemented 32bit optimizers in triton (#1710)

YangKai0616
and
faaany
authored
275671b
Add SYCL Kernels for XPU backend (#1679)

authored
1813b05
Lint fix
matthewdouglas
committed
4b02574

Commits on Sep 16, 2025

[XPU] Implemented 8bit optimizers in triton (#1692)
Egor-Krivov
authored
404e277
Drop Maxwell (sm50) build from distribution (#1755)
matthewdouglas
authored
dd1929b
Bump minimum PyTorch to 2.3 (#1754)
matthewdouglas
authored
c9bce2b

Commits on Sep 18, 2025

[CUDA] Branchless NF4/FP4 kDequantizeBlockwise kernel for faster dequantization (#1746)
Mhmd-Hisham
authored
b1f80b8

Commits on Sep 19, 2025

Update log (#1758)
YangKai0616
authored
b2a8a15
Add function to reverse 4bit weights for HPU (#1757)
vivekgoe
authored
2adcb7a

Commits on Sep 22, 2025

Update README.md
matthewdouglas
authored
e817036

Commits on Sep 23, 2025

Merge upstream/main into ROCm/rocm_enabled
root
committed
0507a45
Skip unsupported tests on ROCm
pnunna93
committed
359d545
update kernels.hip with latest upstream
pnunna93
committed
9f74744
Import missing modules
pnunna93
committed
7ba4fb4
Fix lint errors.
pnunna93
committed
36da3e1