add documentation for --mp_context

README.md

@@ -28,9 +28,9 @@ python -m venv venv
 source venv/bin/activate
 ```
 
-Install gridfm-graphkit in editable mode
+Install gridfm-graphkit from PyPI
 ```bash
-pip install -e .
+pip install gridfm-graphkit
 ```
 
 **`torch-scatter` is a required dependency.** It cannot be bundled in `pyproject.toml` because the correct wheel depends on your PyTorch and CUDA versions, so it must be installed separately.
@@ -49,7 +49,7 @@ pip install torch-scatter -f https://data.pyg.org/whl/torch-${TORCH_CUDA_VERSION
 For documentation generation and unit testing, install with the optional `dev` and `test` extras:
 
 ```bash
-pip install -e .[dev,test]
+pip install "gridfm-graphkit[dev,test]"
 ```
 
 
@@ -94,6 +94,7 @@ gridfm_graphkit train --config path/to/config.yaml
 | `--dataset_wrapper_cache_dir` | `str` | Disk cache directory for dataset wrapper; cache is loaded from here when present and saved after first population. | `None` |
 | `--profiler` | `str` | Enable Lightning profiler (`simple`, `advanced`, `pytorch`). | `None` |
 | `--compute_dc_ac_metrics` | `flag` | Compute ground-truth AC/DC power balance metrics on the test split. | `False` |
+| `--mp_context` | `str` | DataLoader multiprocessing start method (`spawn`, `fork`, `forkserver`). Defaults to PyTorch's automatic choice. On Linux, `spawn` is recommended for safety (CUDA + fork is unsafe); other choices emit a warning. | `None` |
 
 ### Examples
 
@@ -130,6 +131,7 @@ gridfm_graphkit finetune --config path/to/config.yaml --model_path path/to/model
 | `--dataset_wrapper_cache_dir` | `str` | Disk cache directory for dataset wrapper; cache is loaded from here when present and saved after first population. | `None` |
 | `--profiler` | `str` | Enable Lightning profiler (`simple`, `advanced`, `pytorch`). | `None` |
 | `--compute_dc_ac_metrics` | `flag` | Compute ground-truth AC/DC power balance metrics on the test split. | `False` |
+| `--mp_context` | `str` | DataLoader multiprocessing start method (`spawn`, `fork`, `forkserver`). Defaults to PyTorch's automatic choice. On Linux, `spawn` is recommended for safety (CUDA + fork is unsafe); other choices emit a warning. | `None` |
 
 
 ---
@@ -161,6 +163,7 @@ gridfm_graphkit evaluate --config path/to/eval.yaml --model_path path/to/model.p
 | `--profiler` | `str` | Enable Lightning profiler (`simple`, `advanced`, `pytorch`). | `None` |
 | `--compute_dc_ac_metrics` | `flag` | Compute ground-truth AC/DC power balance metrics on the test split. | `False` |
 | `--save_output` | `flag` | Save predictions as `<grid_name>_predictions.parquet` under MLflow artifacts (`.../artifacts/test`). | `False` |
+| `--mp_context` | `str` | DataLoader multiprocessing start method (`spawn`, `fork`, `forkserver`). Defaults to PyTorch's automatic choice. On Linux, `spawn` is recommended for safety (CUDA + fork is unsafe); other choices emit a warning. | `None` |
 
 ### Example with saved normalizer stats
 
@@ -204,6 +207,7 @@ gridfm_graphkit predict --config path/to/config.yaml --model_path path/to/model.
 | `--bfloat16` | `flag` | Cast model to `torch.bfloat16` (`model.to(torch.bfloat16)`). | `False` |
 | `--tf32` | `flag` | Enable TF32 on Ampere+ GPUs via `torch.set_float32_matmul_precision("high")`. | `False` |
 | `--profiler` | `str` | Enable Lightning profiler (`simple`, `advanced`, `pytorch`). | `None` |
+| `--mp_context` | `str` | DataLoader multiprocessing start method (`spawn`, `fork`, `forkserver`). Defaults to PyTorch's automatic choice. On Linux, `spawn` is recommended for safety (CUDA + fork is unsafe); other choices emit a warning. | `None` |
 
 ---
 
@@ -224,6 +228,7 @@ gridfm_graphkit benchmark --config path/to/config.yaml
 | `--dataset_wrapper_cache_dir` | `str` | Directory for dataset wrapper disk cache. | `None` |
 | `--num_workers` | `int` | Override `data.workers` from YAML. | `None` |
 | `--plugins` | `list[str]` | Python packages to import for plugin registration. | `[]` |
+| `--mp_context` | `str` | DataLoader multiprocessing start method (`spawn`, `fork`, `forkserver`). Defaults to PyTorch's automatic choice. On Linux, `spawn` is recommended for safety (CUDA + fork is unsafe); other choices emit a warning. | `None` |
 
 Use built-in help for full command details:
 

docs/install/installation.md

@@ -1,6 +1,6 @@
 # Installation
 
-The steps below mirror the [README](https://github.com/gridfm/gridfm-graphkit/blob/main/README.md#installation). Run them from the root of a local clone or source checkout of the repository.
+The steps below mirror the [README](https://github.com/gridfm/gridfm-graphkit/blob/main/README.md#installation).
 
 Create and activate a virtual environment (make sure you use the right python version = 3.10, 3.11 or 3.12. I highly recommend 3.12)
 
@@ -9,10 +9,10 @@ python -m venv venv
 source venv/bin/activate
 ```
 
-Install gridfm-graphkit in editable mode
+Install gridfm-graphkit from PyPI
 
 ```bash
-pip install -e .
+pip install gridfm-graphkit
 ```
 
 **`torch-scatter` is a required dependency.** It cannot be bundled in `pyproject.toml` because the correct wheel depends on your PyTorch and CUDA versions, so it must be installed separately.
@@ -33,5 +33,5 @@ pip install torch-scatter -f https://data.pyg.org/whl/torch-${TORCH_CUDA_VERSION
 For documentation generation and unit testing, install with the optional `dev` and `test` extras:
 
 ```bash
-pip install -e .[dev,test]
+pip install "gridfm-graphkit[dev,test]"
 ```

docs/quick_start/quick_start.md

@@ -40,6 +40,7 @@ gridfm_graphkit train --config path/to/config.yaml
 | `--dataset_wrapper_cache_dir` | `str` | Disk cache directory for dataset wrapper; cache is loaded from here when present and saved after first population. | `None` |
 | `--profiler` | `str` | Enable Lightning profiler (`simple`, `advanced`, `pytorch`). | `None` |
 | `--compute_dc_ac_metrics` | `flag` | Compute ground-truth AC/DC power balance metrics on the test split. | `False` |
+| `--mp_context` | `str` | DataLoader multiprocessing start method (`spawn`, `fork`, `forkserver`). Defaults to PyTorch's automatic choice. On Linux, `spawn` is recommended for safety (CUDA + fork is unsafe); other choices emit a warning. | `None` |
 
 ### Examples
 
@@ -76,6 +77,7 @@ gridfm_graphkit finetune --config path/to/config.yaml --model_path path/to/model
 | `--dataset_wrapper_cache_dir` | `str` | Disk cache directory for dataset wrapper; cache is loaded from here when present and saved after first population. | `None` |
 | `--profiler` | `str` | Enable Lightning profiler (`simple`, `advanced`, `pytorch`). | `None` |
 | `--compute_dc_ac_metrics` | `flag` | Compute ground-truth AC/DC power balance metrics on the test split. | `False` |
+| `--mp_context` | `str` | DataLoader multiprocessing start method (`spawn`, `fork`, `forkserver`). Defaults to PyTorch's automatic choice. On Linux, `spawn` is recommended for safety (CUDA + fork is unsafe); other choices emit a warning. | `None` |
 
 
 ---
@@ -107,6 +109,7 @@ gridfm_graphkit evaluate --config path/to/eval.yaml --model_path path/to/model.p
 | `--profiler` | `str` | Enable Lightning profiler (`simple`, `advanced`, `pytorch`). | `None` |
 | `--compute_dc_ac_metrics` | `flag` | Compute ground-truth AC/DC power balance metrics on the test split. | `False` |
 | `--save_output` | `flag` | Save predictions as `<grid_name>_predictions.parquet` under MLflow artifacts (`.../artifacts/test`). | `False` |
+| `--mp_context` | `str` | DataLoader multiprocessing start method (`spawn`, `fork`, `forkserver`). Defaults to PyTorch's automatic choice. On Linux, `spawn` is recommended for safety (CUDA + fork is unsafe); other choices emit a warning. | `None` |
 
 ### Example with saved normalizer stats
 
@@ -150,6 +153,7 @@ gridfm_graphkit predict --config path/to/config.yaml --model_path path/to/model.
 | `--bfloat16` | `flag` | Cast model to `torch.bfloat16` (`model.to(torch.bfloat16)`). | `False` |
 | `--tf32` | `flag` | Enable TF32 on Ampere+ GPUs via `torch.set_float32_matmul_precision("high")`. | `False` |
 | `--profiler` | `str` | Enable Lightning profiler (`simple`, `advanced`, `pytorch`). | `None` |
+| `--mp_context` | `str` | DataLoader multiprocessing start method (`spawn`, `fork`, `forkserver`). Defaults to PyTorch's automatic choice. On Linux, `spawn` is recommended for safety (CUDA + fork is unsafe); other choices emit a warning. | `None` |
 
 ---
 
@@ -170,6 +174,7 @@ gridfm_graphkit benchmark --config path/to/config.yaml
 | `--dataset_wrapper_cache_dir` | `str` | Directory for dataset wrapper disk cache. | `None` |
 | `--num_workers` | `int` | Override `data.workers` from YAML. | `None` |
 | `--plugins` | `list[str]` | Python packages to import for plugin registration. | `[]` |
+| `--mp_context` | `str` | DataLoader multiprocessing start method (`spawn`, `fork`, `forkserver`). Defaults to PyTorch's automatic choice. On Linux, `spawn` is recommended for safety (CUDA + fork is unsafe); other choices emit a warning. | `None` |
 
 Use built-in help for full command details: