Remove the Graphcore IPU integration (#19405)

Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>

Author: awaelchli

.github/workflows/README.md

@@ -27,7 +27,6 @@ Brief description of all our automation tools used for boosting development perf
 
   - GPU: 2 x NVIDIA RTX 3090
   - TPU: [Google TPU v4-8](https://cloud.google.com/tpu/docs)
-  - IPU: [Colossus MK1 IPU](https://www.graphcore.ai/products/ipu)
 
 - To check which versions of Python or PyTorch are used for testing in our CI, see the corresponding workflow files or checkgroup config file at [`.github/checkgroup.yml`](../checkgroup.yml).
 

.gitignore

@@ -22,9 +22,7 @@ docs/source-pytorch/notebooks
 docs/source-pytorch/_static/images/course_UvA-DL
 docs/source-pytorch/_static/images/lightning_examples
 docs/source-pytorch/_static/fetched-s3-assets
-docs/source-pytorch/_static/images/ipu/
 docs/source-pytorch/integrations/hpu
-docs/source-pytorch/integrations/ipu
 
 docs/source-fabric/*/generated
 

docs/source-app/quickstart.rst

@@ -53,7 +53,6 @@ And that's it!
 
         GPU available: True (mps), used: False
         TPU available: False, using: 0 TPU cores
-        IPU available: False, using: 0 IPUs
 
           | Name    | Type     | Params | In sizes       | Out sizes
         ------------------------------------------------------------------

docs/source-pytorch/advanced/speed.rst

@@ -20,7 +20,7 @@ Training on Accelerators
 
 **Use when:** Whenever possible!
 
-With Lightning, running on GPUs, TPUs, IPUs on multiple nodes is a simple switch of a flag.
+With Lightning, running on GPUs, TPUs, HPUs on multiple nodes is a simple switch of a flag.
 
 GPU Training
 ============

docs/source-pytorch/common/index.rst

@@ -17,7 +17,6 @@
    ../advanced/model_parallel
    Train on single or multiple GPUs <../accelerators/gpu>
    Train on single or multiple HPUs <../integrations/hpu/index>
-   Train on single or multiple IPUs <../integrations/ipu/index>
    Train on single or multiple TPUs <../accelerators/tpu>
    Train on MPS <../accelerators/mps>
    Use a pretrained model <../advanced/pretrained>
@@ -168,13 +167,6 @@ How-to Guides
     :col_css: col-md-4
     :height: 180
 
-.. displayitem::
-    :header: Train on single or multiple IPUs
-    :description: Train models faster with IPU accelerators
-    :button_link: ../integrations/ipu/index.html
-    :col_css: col-md-4
-    :height: 180
-
 .. displayitem::
     :header: Train on single or multiple TPUs
     :description: TTrain models faster with TPU accelerators

docs/source-pytorch/common/precision_basic.rst

@@ -103,31 +103,26 @@ Precision support by accelerator
 ********************************
 
 .. list-table:: Precision with Accelerators
-   :widths: 20 20 20 20 20
+   :widths: 20 20 20 20
    :header-rows: 1
 
    * - Precision
      - CPU
      - GPU
      - TPU
-     - IPU
    * - 16 Mixed
      - No
      - Yes
      - No
-     - Yes
    * - BFloat16 Mixed
      - Yes
      - Yes
      - Yes
-     - No
    * - 32 True
      - Yes
      - Yes
      - Yes
-     - Yes
    * - 64 True
      - Yes
      - Yes
      - No
-     - No

docs/source-pytorch/common/trainer.rst

@@ -175,7 +175,7 @@ Trainer flags
 accelerator
 ^^^^^^^^^^^
 
-Supports passing different accelerator types (``"cpu", "gpu", "tpu", "ipu", "auto"``)
+Supports passing different accelerator types (``"cpu", "gpu", "tpu", "hpu", "auto"``)
 as well as custom accelerator instances.
 
 .. code-block:: python
@@ -393,9 +393,6 @@ Number of devices to train on (``int``), which devices to train on (``list`` or
     # Training with TPU Accelerator using 8 tpu cores
     trainer = Trainer(devices="auto", accelerator="tpu")
 
-    # Training with IPU Accelerator using 4 ipus
-    trainer = Trainer(devices="auto", accelerator="ipu")
-
 .. note::
 
     If the ``devices`` flag is not defined, it will assume ``devices`` to be ``"auto"`` and fetch the ``auto_device_count``

docs/source-pytorch/common_usecases.rst

@@ -133,13 +133,6 @@ Customize and extend Lightning for things like custom hardware or distributed st
    :button_link: integrations/hpu/index.html
    :height: 100
 
-.. displayitem::
-   :header: Train on single or multiple IPUs
-   :description: Train models faster with IPUs.
-   :col_css: col-md-12
-   :button_link: integrations/ipu/index.html
-   :height: 100
-
 .. displayitem::
    :header: Train on single or multiple TPUs
    :description: Train models faster with TPUs.

docs/source-pytorch/conf.py

@@ -94,18 +94,6 @@ def _load_py_module(name: str, location: str) -> ModuleType:
     target_dir="docs/source-pytorch/integrations/hpu",
     checkout="refs/tags/1.3.0",
 )
-assist_local.AssistantCLI.pull_docs_files(
-    gh_user_repo="Lightning-AI/lightning-Graphcore",
-    target_dir="docs/source-pytorch/integrations/ipu",
-    checkout="refs/tags/v0.1.0",
-    as_orphan=True,  # todo: this can be dropped after new IPU release
-)
-# the IPU also need one image
-URL_RAW_DOCS_GRAPHCORE = "https://raw.githubusercontent.com/Lightning-AI/lightning-Graphcore/v0.1.0/docs/source"
-for img in ["_static/images/ipu/profiler.png"]:
-    img_ = os.path.join(_PATH_HERE, "integrations", "ipu", img)
-    os.makedirs(os.path.dirname(img_), exist_ok=True)
-    urllib.request.urlretrieve(f"{URL_RAW_DOCS_GRAPHCORE}/{img}", img_)
 
 # Copy strategies docs as single pages
 assist_local.AssistantCLI.pull_docs_files(
@@ -340,7 +328,6 @@ def _load_py_module(name: str, location: str) -> ModuleType:
     "numpy": ("https://numpy.org/doc/stable/", None),
     "PIL": ("https://pillow.readthedocs.io/en/stable/", None),
     "torchmetrics": ("https://torchmetrics.readthedocs.io/en/stable/", None),
-    "graphcore": ("https://docs.graphcore.ai/en/latest/", None),
     "lightning_habana": ("https://lightning-ai.github.io/lightning-Habana/", None),
     "tensorboardX": ("https://tensorboardx.readthedocs.io/en/stable/", None),
     # needed for referencing App from lightning scope

docs/source-pytorch/expertise_levels.rst

@@ -190,34 +190,26 @@ Configure all aspects of Lightning for advanced usecases.
    :tag: advanced
 
 .. displayitem::
-   :header: Level 18: Explore IPUs
-   :description: Explore Intelligence Processing Unit (IPU) for model scaling.
-   :col_css: col-md-6
-   :button_link: levels/advanced_level_19.html
-   :height: 150
-   :tag: advanced
-
-.. displayitem::
-   :header: Level 19: Explore HPUs
+   :header: Level 18: Explore HPUs
    :description: Explore Havana Gaudi Processing Unit (HPU) for model scaling.
    :col_css: col-md-6
-   :button_link: levels/advanced_level_20.html
+   :button_link: levels/advanced_level_19.html
    :height: 150
    :tag: advanced
 
 .. displayitem::
-   :header: Level 20: Master TPUs
+   :header: Level 19: Master TPUs
    :description: Master TPUs and run on cloud TPUs.
    :col_css: col-md-6
-   :button_link: levels/advanced_level_21.html
+   :button_link: levels/advanced_level_20.html
    :height: 150
    :tag: advanced
 
 .. displayitem::
-   :header: Level 21: Train models with billions of parameters
+   :header: Level 20: Train models with billions of parameters
    :description: Scale GPU training to models with billions of parameters
    :col_css: col-md-6
-   :button_link: levels/advanced_level_22.html
+   :button_link: levels/advanced_level_21.html
    :height: 150
    :tag: advanced
 
@@ -240,34 +232,34 @@ Customize and extend Lightning for things like custom hardware or distributed st
 .. Add callout items below this line
 
 .. displayitem::
-   :header: Level 22: Extend the Lightning CLI
+   :header: Level 21: Extend the Lightning CLI
    :description: Extend the functionality of the Lightning CLI.
    :col_css: col-md-6
    :button_link: levels/expert_level_23.html
    :height: 150
    :tag: expert
 
 .. displayitem::
-   :header: Level 23: Integrate a custom cluster
+   :header: Level 22: Integrate a custom cluster
    :description: Integrate a custom cluster into Lightning.
    :col_css: col-md-6
    :button_link: levels/expert_level_24.html
    :height: 150
    :tag: expert
 
 .. displayitem::
-   :header: Level 24: Make your own profiler
+   :header: Level 23: Make your own profiler
    :description: Make your own profiler.
    :col_css: col-md-6
    :button_link: tuning/profiler_expert.html
    :height: 150
    :tag: expert
 
 .. displayitem::
-   :header: Level 25: Add a new accelerator or Strategy
+   :header: Level 24: Add a new accelerator or Strategy
    :description: Integrate a new accelerator or distributed strategy.
    :col_css: col-md-6
-   :button_link: levels/expert_level_27.html
+   :button_link: levels/expert_level_25.html
    :height: 150
    :tag: expert
 

docs/source-pytorch/extensions/accelerator.rst

@@ -4,13 +4,12 @@
 Accelerator
 ###########
 
-The Accelerator connects a Lightning Trainer to arbitrary hardware (CPUs, GPUs, TPUs, IPUs, MPS, ...).
+The Accelerator connects a Lightning Trainer to arbitrary hardware (CPUs, GPUs, TPUs, HPUs, MPS, ...).
 Currently there are accelerators for:
 
 - CPU
 - :doc:`GPU <../accelerators/gpu>`
 - :doc:`TPU <../accelerators/tpu>`
-- :doc:`IPU <../integrations/ipu/index>`
 - :doc:`HPU <../integrations/hpu/index>`
 - :doc:`MPS <../accelerators/mps>`
 

docs/source-pytorch/extensions/strategy.rst

@@ -57,9 +57,6 @@ Here are some examples:
     # Training with the DDP Spawn strategy on 8 TPU cores
     trainer = Trainer(strategy="ddp_spawn", accelerator="tpu", devices=8)
 
-    # Training with the default IPU strategy on 8 IPUs
-    trainer = Trainer(accelerator="ipu", devices=8)
-
 The below table lists all relevant strategies available in Lightning with their corresponding short-hand name:
 
 .. list-table:: Strategy Classes and Nicknames
@@ -87,9 +84,6 @@ The below table lists all relevant strategies available in Lightning with their
    * - hpu_single
      - ``SingleHPUStrategy``
      - Strategy for training on a single HPU device. :doc:`Learn more. <../integrations/hpu/index>`
-   * - ipu_strategy
-     - ``IPUStrategy``
-     - Plugin for training on IPU devices. :doc:`Learn more. <../integrations/ipu/index>`
    * - xla
      - :class:`~lightning.pytorch.strategies.XLAStrategy`
      - Strategy for training on multiple TPU devices using the :func:`torch_xla.distributed.xla_multiprocessing.spawn` method. :doc:`Learn more. <../accelerators/tpu>`

docs/source-pytorch/glossary/index.rst

@@ -20,7 +20,6 @@
    Half precision <../common/precision>
    HPU <../integrations/hpu/index>
    Inference <../deploy/production_intermediate>
-   IPU <../integrations/ipu/index>
    Lightning CLI <../cli/lightning_cli>
    LightningDataModule <../data/datamodule>
    LightningModule <../common/lightning_module>
@@ -177,13 +176,6 @@ Glossary
    :button_link: ../deploy/production_intermediate.html
    :height: 100
 
-.. displayitem::
-   :header: IPU
-   :description: Graphcore Intelligence Processing Unit for faster training
-   :col_css: col-md-12
-   :button_link: ../integrations/ipu/index.html
-   :height: 100
-
 .. displayitem::
    :header: Lightning CLI
    :description: A Command-line Interface (CLI) to interact with Lightning code via a terminal

docs/source-pytorch/integrations/ipu/index.rst

@@ -46,34 +46,26 @@ Configure all aspects of Lightning for advanced usecases.
    :tag: advanced
 
 .. displayitem::
-   :header: Level 18: Explore IPUs
-   :description: Explore Intelligence Processing Unit (IPU) for model scaling.
-   :col_css: col-md-6
-   :button_link: advanced_level_19.html
-   :height: 150
-   :tag: advanced
-
-.. displayitem::
-   :header: Level 19: Explore HPUs
+   :header: Level 18: Explore HPUs
    :description: Explore Habana Gaudi Processing Unit (HPU) for model scaling.
    :col_css: col-md-6
-   :button_link: advanced_level_20.html
+   :button_link: advanced_level_19.html
    :height: 150
    :tag: advanced
 
 .. displayitem::
-   :header: Level 20: Master TPUs
+   :header: Level 19: Master TPUs
    :description: Master TPUs and run on cloud TPUs.
    :col_css: col-md-6
-   :button_link: advanced_level_21.html
+   :button_link: advanced_level_20.html
    :height: 150
    :tag: advanced
 
 .. displayitem::
-   :header: Level 21: Train models with billions of parameters
+   :header: Level 20: Train models with billions of parameters
    :description: Scale GPU training to models with billions of parameters
    :col_css: col-md-6
-   :button_link: advanced_level_22.html
+   :button_link: advanced_level_21.html
    :height: 150
    :tag: advanced