feat: add device name display (for example: A100 not just cuda) (#5146)

…ot just cuda)

<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit

* **New Features**
* Automatic GPU device detection added for PyTorch, TensorFlow, and
Paddle; detected GPU names are now shown in system diagnostics when
available.
* Detection is conditional and safe: if no GPU is present or detection
fails, diagnostics simply omit the device name without causing errors.

<sub>✏️ Tip: You can customize this high-level summary in your review
settings.</sub>
<!-- end of auto-generated comment: release notes by coderabbit.ai -->

Author: OutisLi

deepmd/pd/entrypoints/main.py

@@ -224,6 +224,20 @@ def get_backend_info(self) -> dict:
             **op_info,
         }
 
+    def get_device_name(self) -> str | None:
+        """Get the underlying GPU name.
+
+        Returns
+        -------
+        str or None
+            The device name if available, otherwise None.
+        """
+        if paddle.device.is_compiled_with_cuda():
+            cuda = paddle.device.cuda
+            if cuda.device_count() > 0:
+                return cuda.get_device_name()
+        return None
+
 
 def train(
     input_file: str,

deepmd/pt/entrypoints/main.py

@@ -252,6 +252,18 @@ def get_backend_info(self) -> dict:
             **op_info,
         }
 
+    def get_device_name(self) -> str | None:
+        """Use PyTorch's current device name as the device identifier.
+
+        Returns
+        -------
+        str or None
+            The device name if available, otherwise None.
+        """
+        if torch.cuda.is_available():
+            return torch.cuda.get_device_name(torch.cuda.current_device())
+        return None
+
 
 def train(
     input_file: str,

deepmd/tf/train/run_options.py

@@ -73,6 +73,24 @@ def get_backend_info(self) -> dict:
             "build with TF lib": GLOBAL_CONFIG["tf_libs"].replace(";", "\n"),
         }
 
+    def get_device_name(self) -> str | None:
+        """Get the hardware device name if available.
+
+        Returns
+        -------
+        str or None
+            The device name (e.g., NVIDIA A100) if available, otherwise None.
+        """
+        try:
+            gpus = tf.config.get_visible_devices("GPU")
+            if gpus:
+                details = tf.config.experimental.get_device_details(gpus[0])
+                return details.get("device_name")
+        except (AttributeError, RuntimeError):
+            # Experimental API may not exist or fail in some TF versions
+            pass
+        return None
+
 
 class RunOptions:
     """Class with info on how to run training (cluster, MPI and GPU config).

deepmd/utils/summary.py

@@ -74,6 +74,9 @@ def __call__(self) -> None:
                 "computing device": self.get_compute_device(),
             }
         )
+        device_name = self.get_device_name()
+        if device_name:
+            build_info["Device Name"] = device_name
         if self.is_built_with_cuda():
             env_value = os.environ.get("CUDA_VISIBLE_DEVICES", "unset")
             build_info["CUDA_VISIBLE_DEVICES"] = env_value
@@ -123,6 +126,10 @@ def get_compute_device(self) -> str:
     def get_ngpus(self) -> int:
         """Get the number of GPUs."""
 
+    @abstractmethod
+    def get_device_name(self) -> str | None:
+        """Get the device name (e.g., NVIDIA A800-SXM4-80GB) if available."""
+
     def get_backend_info(self) -> dict:
         """Get backend information."""
         return {}