Merge branch 'deepspeedai:master' into avoid-graph-break-caused-by-in…

…ner-classes

.github/workflows/no-torch.yml

@@ -32,11 +32,12 @@ jobs:
         run: |
           pip uninstall torch --yes
           pip install setuptools
+          pip install build
           pip list
 
       - name: Build deepspeed
         run: |
-          DS_BUILD_STRING=" " python setup.py sdist
+          DS_BUILD_STRING=" " python -m build --sdist
 
       - name: Open GitHub issue if nightly CI fails
         if: ${{ failure() && (github.event_name == 'schedule') }}

.github/workflows/release.yml

@@ -26,7 +26,8 @@ jobs:
     - name: Build DeepSpeed
       run: |
         pip install setuptools
-        DS_BUILD_STRING=" " python setup.py sdist
+        pip install build
+        DS_BUILD_STRING=" " python -m build --sdist
     - name: Publish to PyPI
       uses: pypa/gh-action-pypi-publish@release/v1
       with:

README.md

@@ -172,6 +172,7 @@ dynamically link them at runtime.
 | Intel       | Intel(R) Gaudi(R) 2 AI accelerator  | hpu              | Yes | Yes                |
 | Intel       | Intel(R) Xeon(R) Processors         | cpu              | Yes | Yes                |
 | Intel       | Intel(R) Data Center GPU Max series | xpu              | Yes | Yes                |
+| Tecorigin   | Scalable Data Analytics Accelerator | sdaa             | Yes | No                 |
 
 ## PyPI
 We regularly push releases to [PyPI](https://pypi.org/project/deepspeed/) and encourage users to install from there in most cases.

accelerator/real_accelerator.py

@@ -20,7 +20,7 @@
 except ImportError as e:
     dsa2 = None
 
-SUPPORTED_ACCELERATOR_LIST = ['cuda', 'cpu', 'xpu', 'xpu.external', 'npu', 'mps', 'hpu', 'mlu']
+SUPPORTED_ACCELERATOR_LIST = ['cuda', 'cpu', 'xpu', 'xpu.external', 'npu', 'mps', 'hpu', 'mlu', 'sdaa']
 
 ds_accelerator = None
 
@@ -80,6 +80,12 @@ def get_accelerator():
             except ImportError as e:
                 raise ValueError(f"NPU_Accelerator requires torch_npu, which is not installed on this system.")
             pass
+        elif accelerator_name == "sdaa":
+            try:
+                import torch_sdaa  # noqa: F401 # type: ignore
+            except ImportError as e:
+                raise ValueError(f"SDAA_Accelerator requires torch_sdaa, which is not installed on this system.")
+            pass
         elif accelerator_name == "mps":
             try:
                 import torch.mps
@@ -137,6 +143,13 @@ def get_accelerator():
                 accelerator_name = "npu"
             except ImportError as e:
                 pass
+        if accelerator_name is None:
+            try:
+                import torch_sdaa  # noqa: F401,F811 # type: ignore
+
+                accelerator_name = "sdaa"
+            except ImportError as e:
+                pass
         if accelerator_name is None:
             try:
                 import torch.mps
@@ -205,6 +218,10 @@ def get_accelerator():
         from .npu_accelerator import NPU_Accelerator
 
         ds_accelerator = NPU_Accelerator()
+    elif accelerator_name == "sdaa":
+        from .sdaa_accelerator import SDAA_Accelerator
+
+        ds_accelerator = SDAA_Accelerator()
     elif accelerator_name == "mps":
         from .mps_accelerator import MPS_Accelerator