Enable packaging for ROCm 6.2 (#1367)

* Enable 6.2 build

* Update documentation for 6.2.0 pip install

.github/workflows/python-package.yml

@@ -107,7 +107,7 @@ jobs:
         os: [ubuntu-latest]
         arch: [x86_64]
         rocm_version:
-          ["6.1.2"]
+          ["6.1.2", "6.2"]
     runs-on: ${{ matrix.os }} # One day, we could run them on native agents. Azure supports this now but it's planned only for Q3 2023 for hosted agents
     steps:
       - uses: actions/checkout@v4

docs/source/installation.mdx

@@ -167,7 +167,7 @@ apt-get update && apt-get install -y git && cd home
 pip install torch --index-url https://download.pytorch.org/whl/rocm6.1/
 
 # Install bitsandbytes from PyPI
-# (This is supported on Ubuntu 22.04, Python 3.10, ROCm 6.1.0/6.1.1/6.1.2 and gpu arch - gfx90a, gfx942, gfx1100
+# (This is supported on Ubuntu 22.04, Python 3.10, ROCm 6.1.0/6.1.1/6.1.2/6.2.0 and gpu arch - gfx90a, gfx942, gfx1100
 # Please install from source if your configuration doesn't match with these)
 pip install bitsandbytes
 

tests/test_functional.py

@@ -2303,6 +2303,7 @@ def test_gemv_4bit(dtype, storage_type, quant_storage, double_quant, kind):
             assert maxratio < 1.02 and maxratio > 0.98
 
 
+@pytest.mark.skipif(HIP_ENVIRONMENT, reason="this test is not supported on ROCm yet")
 @pytest.mark.parametrize("kind", ["fc1", "fc2", "attn", "attn_packed"])
 @pytest.mark.parametrize("quant_type", ["nf4", "fp4"])
 @pytest.mark.parametrize("dtype", [torch.float16, torch.bfloat16, torch.float32], ids=describe_dtype)