update

Author: Borda

src/lightning/pytorch/callbacks/device_stats_monitor.py

@@ -34,21 +34,68 @@ class DeviceStatsMonitor(Callback):
     r"""Automatically monitors and logs device stats during training, validation and testing stage.
     ``DeviceStatsMonitor`` is a special callback as it requires a ``logger`` to passed as argument to the ``Trainer``.
 
-    Device statistics are logged with keys prefixed as ``DeviceStatsMonitor.{hook_name}/{base_metric_name}`` (e.g.,
-    ``DeviceStatsMonitor.on_train_batch_start/cpu_percent``).
-    The source of these metrics depends on the ``cpu_stats`` flag and the active accelerator.
 
-        CPU (via ``psutil``): Logs ``cpu_percent``, ``cpu_vm_percent``, ``cpu_swap_percent``.
-        All are percentages (%).
-        CUDA GPU (via :func:`torch.cuda.memory_stats`): Logs detailed memory statistics from
-        PyTorch's allocator (e.g., ``allocated_bytes.all.current``, ``num_ooms``; all in Bytes).
-        GPU compute utilization is not logged by default.
-        Other Accelerators (e.g., TPU, MPS): Logs device-specific stats:
+    **Logged Metrics**
 
-        - TPU example: ``avg. free memory (MB)``.
-        - MPS example: ``mps.current_allocated_bytes``.
+    Logs device statistics with keys prefixed as ``DeviceStatsMonitor.{hook_name}/{base_metric_name}``.
 
-        Observe logs or check accelerator documentation for details.
+    The actual metrics depend on the active accelerator and the ``cpu_stats`` flag.
+
+    **CPU (via `psutil`)**
+
+    - ``cpu_percent``: System-wide CPU utilization (%)
+    - ``cpu_vm_percent``: System-wide virtual memory (RAM) utilization (%)
+    - ``cpu_swap_percent``: System-wide swap memory utilization (%)
+
+    **CUDA GPU (via `torch.cuda.memory_stats`)**
+
+    Logs memory statistics from PyTorch caching allocator (all in Bytes). 
+    GPU compute utilization is not logged by default.
+
+    *General Memory Usage:*
+
+    - ``allocated_bytes.all.current``: Current allocated GPU memory
+    - ``allocated_bytes.all.peak``: Peak allocated GPU memory
+    - ``reserved_bytes.all.current``: Current reserved GPU memory (allocated + cached)
+    - ``reserved_bytes.all.peak``: Peak reserved GPU memory
+    - ``active_bytes.all.current``: Current GPU memory in active use
+    - ``active_bytes.all.peak``: Peak GPU memory in active use
+    - ``inactive_split_bytes.all.current``: Memory in inactive, splittable blocks
+
+    *Allocator Pool Statistics* (for ``small_pool`` and ``large_pool``):
+
+    - ``allocated_bytes.{pool_type}.current`` / ``.peak``
+    - ``reserved_bytes.{pool_type}.current`` / ``.peak``
+    - ``active_bytes.{pool_type}.current`` / ``.peak``
+
+    *Allocator Events:*
+
+    - ``num_ooms``: Cumulative out-of-memory errors
+    - ``num_alloc_retries``: Number of allocation retries
+    - ``num_device_alloc``: Number of device allocations
+    - ``num_device_free``: Number of device deallocations
+
+    For a full list of CUDA memory stats, see:
+    https://pytorch.org/docs/stable/generated/torch.cuda.memory_stats.html
+
+    **TPU (via `torch_xla`)**
+
+    *Memory Metrics* (per device, e.g. ``xla:0``):
+
+    - ``memory.free.xla:0``: Free HBM memory (MB)
+    - ``memory.used.xla:0``: Used HBM memory (MB)
+    - ``memory.percent.xla:0``: Percentage of HBM memory used (%)
+
+    *XLA Operation Counters:*
+
+    - ``CachedCompile.xla``
+    - ``CreateXlaTensor.xla``
+    - ``DeviceDataCacheMiss.xla``
+    - ``UncachedCompile.xla``
+    - ``xla::add.xla``, ``xla::addmm.xla``, etc.
+
+    These counters can be retrieved using:
+    ``torch_xla.debug.metrics.counter_names()``
 
     Args:
         cpu_stats: if ``None``, it will log CPU stats only if the accelerator is CPU.