jax-image-models 0.3.3__tar.gz

jax_image_models-0.3.3/.gitattributes

@@ -0,0 +1,2 @@
+# SCM syntax highlighting & preventing 3-way merges
+pixi.lock merge=binary linguist-language=YAML linguist-generated=true

jax_image_models-0.3.3/.github/workflows/ci.yml

@@ -0,0 +1,32 @@
+name: ci
+on:
+  push:
+    branches:
+      - master
+      - main
+permissions:
+  contents: write
+jobs:
+  deploy:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - name: Configure Git Credentials
+        run: |
+          git config user.name github-actions[bot]
+          git config user.email 41898282+github-actions[bot]@users.noreply.github.com
+      - uses: actions/setup-python@v5
+        with:
+          python-version: 3.x
+      - run: echo "cache_id=$(date --utc '+%V')" >> $GITHUB_ENV
+      - uses: actions/cache@v4
+        with:
+          key: mkdocs-material-${{ env.cache_id }}
+          path: .cache
+          restore-keys: |
+            mkdocs-material-
+      - name: Install uv
+        uses: astral-sh/setup-uv@v5
+      - name: Install the project
+        run: uv sync --locked --all-extras --dev && uv pip install -e .
+      - run: uv run mkdocs gh-deploy --force

jax_image_models-0.3.3/.github/workflows/claude-code-review.yml

@@ -0,0 +1,57 @@
+name: Claude Code Review
+
+on:
+  pull_request:
+    types: [opened, synchronize]
+    # Optional: Only run on specific file changes
+    # paths:
+    #   - "src/**/*.ts"
+    #   - "src/**/*.tsx"
+    #   - "src/**/*.js"
+    #   - "src/**/*.jsx"
+
+jobs:
+  claude-review:
+    # Optional: Filter by PR author
+    # if: |
+    #   github.event.pull_request.user.login == 'external-contributor' ||
+    #   github.event.pull_request.user.login == 'new-developer' ||
+    #   github.event.pull_request.author_association == 'FIRST_TIME_CONTRIBUTOR'
+
+    runs-on: ubuntu-latest
+    permissions:
+      contents: read
+      pull-requests: read
+      issues: read
+      id-token: write
+
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 1
+
+      - name: Run Claude Code Review
+        id: claude-review
+        uses: anthropics/claude-code-action@v1
+        with:
+          claude_code_oauth_token: ${{ secrets.CLAUDE_CODE_OAUTH_TOKEN }}
+          prompt: |
+            REPO: ${{ github.repository }}
+            PR NUMBER: ${{ github.event.pull_request.number }}
+
+            Please review this pull request and provide feedback on:
+            - Code quality and best practices
+            - Potential bugs or issues
+            - Performance considerations
+            - Security concerns
+            - Test coverage
+
+            Use the repository's CLAUDE.md for guidance on style and conventions. Be constructive and helpful in your feedback.
+
+            Use `gh pr comment` with your Bash tool to leave your review as a comment on the PR.
+
+          # See https://github.com/anthropics/claude-code-action/blob/main/docs/usage.md
+          # or https://docs.claude.com/en/docs/claude-code/cli-reference for available options
+          claude_args: '--allowed-tools "Bash(gh issue view:*),Bash(gh search:*),Bash(gh issue list:*),Bash(gh pr comment:*),Bash(gh pr diff:*),Bash(gh pr view:*),Bash(gh pr list:*)"'
+

jax_image_models-0.3.3/.github/workflows/claude.yml

@@ -0,0 +1,50 @@
+name: Claude Code
+
+on:
+  issue_comment:
+    types: [created]
+  pull_request_review_comment:
+    types: [created]
+  issues:
+    types: [opened, assigned]
+  pull_request_review:
+    types: [submitted]
+
+jobs:
+  claude:
+    if: |
+      (github.event_name == 'issue_comment' && contains(github.event.comment.body, '@claude')) ||
+      (github.event_name == 'pull_request_review_comment' && contains(github.event.comment.body, '@claude')) ||
+      (github.event_name == 'pull_request_review' && contains(github.event.review.body, '@claude')) ||
+      (github.event_name == 'issues' && (contains(github.event.issue.body, '@claude') || contains(github.event.issue.title, '@claude')))
+    runs-on: ubuntu-latest
+    permissions:
+      contents: read
+      pull-requests: read
+      issues: read
+      id-token: write
+      actions: read # Required for Claude to read CI results on PRs
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 1
+
+      - name: Run Claude Code
+        id: claude
+        uses: anthropics/claude-code-action@v1
+        with:
+          claude_code_oauth_token: ${{ secrets.CLAUDE_CODE_OAUTH_TOKEN }}
+
+          # This is an optional setting that allows Claude to read CI results on PRs
+          additional_permissions: |
+            actions: read
+
+          # Optional: Give a custom prompt to Claude. If this is not specified, Claude will perform the instructions specified in the comment that tagged it.
+          # prompt: 'Update the pull request description to include a summary of changes.'
+
+          # Optional: Add claude_args to customize behavior and configuration
+          # See https://github.com/anthropics/claude-code-action/blob/main/docs/usage.md
+          # or https://docs.claude.com/en/docs/claude-code/cli-reference for available options
+          # claude_args: '--allowed-tools Bash(gh pr:*)'
+

jax_image_models-0.3.3/.github/workflows/release.yaml

@@ -0,0 +1,26 @@
+name: "Publish release to PyPI"
+
+on:
+  push:
+    tags:
+      - v*
+
+jobs:
+  run:
+    runs-on: ubuntu-latest
+    environment:
+      name: pypi
+    permissions:
+      id-token: write
+      contents: read
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+      - name: Install uv
+        uses: astral-sh/setup-uv@08807647e7069bb48b6ef5acd8ec9567f424441b # v8.1.0
+      - name: Install Python
+        run: uv python install 3.13
+      - name: Build
+        run: uv build
+      - name: Publish
+        run: uv publish

jax_image_models-0.3.3/.gitignore

@@ -0,0 +1,205 @@
+bin
+tmp
+weights
+.claude
+tests/tokamax_cache
+# pixi environments
+.pixi
+*.egg-info
+.aider*
+
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# UV
+#   Similar to Pipfile.lock, it is generally recommended to include uv.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#uv.lock
+
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+#poetry.lock
+#poetry.toml
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#pdm.lock
+#   pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it
+#   in version control.
+#   https://pdm.fming.dev/latest/usage/project/#working-with-version-control
+.pdm.toml
+.pdm-python
+.pdm-build/
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# PyCharm
+#  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
+#  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
+#  and can be added to the global gitignore or merged into this file.  For a more nuclear
+#  option (not recommended) you can uncomment the following to ignore the entire idea folder.
+#.idea/
+
+# Abstra
+# Abstra is an AI-powered process automation framework.
+# Ignore directories containing user credentials, local state, and settings.
+# Learn more at https://abstra.io/docs
+.abstra/
+
+# Visual Studio Code
+#  Visual Studio Code specific template is maintained in a separate VisualStudioCode.gitignore
+#  that can be found at https://github.com/github/gitignore/blob/main/Global/VisualStudioCode.gitignore
+#  and can be added to the global gitignore or merged into this file. However, if you prefer,
+#  you could uncomment the following to ignore the entire vscode folder
+# .vscode/
+
+# Ruff stuff:
+.ruff_cache/
+
+# PyPI configuration file
+.pypirc
+
+# Cursor
+#  Cursor is an AI-powered code editor. `.cursorignore` specifies files/directories to
+#  exclude from AI features like autocomplete and code analysis. Recommended for sensitive data
+#  refer to https://docs.cursor.com/context/ignore-files
+.cursorignore
+.cursorindexingignore

jax_image_models-0.3.3/.pre-commit-config.yaml

@@ -0,0 +1,27 @@
+repos:
+# Standard hooks
+- repo: https://github.com/pre-commit/pre-commit-hooks
+  rev: v6.0.0
+  hooks:
+  - id: check-case-conflict
+  - id: check-docstring-first
+  - id: check-merge-conflict
+  - id: check-symlinks
+  - id: check-toml
+  - id: debug-statements
+  - id: mixed-line-ending
+  - id: requirements-txt-fixer
+  - id: trailing-whitespace
+
+- repo: https://github.com/astral-sh/ruff-pre-commit
+  # Ruff version.
+  rev: v0.14.1
+  hooks:
+    # Run the linter.
+    - id: ruff
+      args: [ --fix ]
+    # Sort imports
+    - id: ruff
+      args: ["check", "--select", "I", "--fix"]
+    # Run the formatter.
+    - id: ruff-format

jax_image_models-0.3.3/.python-version

@@ -0,0 +1 @@
+3.11

jax_image_models-0.3.3/.vscode/settings.json

@@ -0,0 +1,3 @@
+{
+    "python.languageServer": "None"
+}

jax_image_models-0.3.3/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2025 Pinak Paliwal
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

jax_image_models-0.3.3/PKG-INFO

@@ -0,0 +1,36 @@
+Metadata-Version: 2.4
+Name: jax-image-models
+Version: 0.3.3
+Summary: Jax Image Modeling of Models
+License-File: LICENSE
+Requires-Python: >=3.11
+Requires-Dist: flax>=0.10.6
+Requires-Dist: jax>=0.6.2
+Requires-Dist: jaxtyping>=0.3.2
+Requires-Dist: safetensors>=0.5.3
+Requires-Dist: tokamax>=0.0.12
+Description-Content-Type: text/markdown
+
+# Jax Image Modeling of Models (jimm)
+Docs are at: [https://pythoncrazy.github.io/jimm](https://pythoncrazy.github.io/jimm)
+- This aims to be the jax counterpart to timm, with the exception that for image-text models (CLIP, SigLIP, etc), we support the text model entirely.
+- Made with flax nnx, supports weight loading from pytorch_model.bin and safetensors (as well as both methods from huggingface).
+
+Models Supported:
+- Vision Transformers
+    - Both with a classification linear layer, or not
+    - Using a CLS Token for pooling, or using Multihead Attention Pooling
+    - Can load any standard variant of Vision Transformers of any size/resolution(e.g. "google/vit-base-patch16-224" or "google/vit-large-patch16-384")
+- CLIP
+    - Can load from any checkpoints of the clip model on github (such as "openai/clip-vit-base-patch32" or "geolocal/StreetCLIP")
+- SigLIP
+    - Can load any non-naflex version of the SigLIP model, from both siglipv1 and siglipv2 (eg "google/siglip-base-patch16-256" or "google/siglip2-large-patch16-512" from huggingface or locally)
+## Installation
+### Using pixi.sh:
+`pixi add jimm@https://github.com/pythoncrazy/jimm.git --pypi`
+### Using uv
+`uv add --dev git+https://github.com/pythoncrazy/jimm.git`
+or if you prefer to not add as a direct dependency:
+`uv pip install git+https://github.com/pythoncrazy/jimm.git`
+### Using pip/conda
+`pip install git+https://github.com/pythoncrazy/jimm.git`

jax_image_models-0.3.3/README.md

@@ -0,0 +1,23 @@
+# Jax Image Modeling of Models (jimm)
+Docs are at: [https://pythoncrazy.github.io/jimm](https://pythoncrazy.github.io/jimm)
+- This aims to be the jax counterpart to timm, with the exception that for image-text models (CLIP, SigLIP, etc), we support the text model entirely.
+- Made with flax nnx, supports weight loading from pytorch_model.bin and safetensors (as well as both methods from huggingface).
+
+Models Supported:
+- Vision Transformers
+    - Both with a classification linear layer, or not
+    - Using a CLS Token for pooling, or using Multihead Attention Pooling
+    - Can load any standard variant of Vision Transformers of any size/resolution(e.g. "google/vit-base-patch16-224" or "google/vit-large-patch16-384")
+- CLIP
+    - Can load from any checkpoints of the clip model on github (such as "openai/clip-vit-base-patch32" or "geolocal/StreetCLIP")
+- SigLIP
+    - Can load any non-naflex version of the SigLIP model, from both siglipv1 and siglipv2 (eg "google/siglip-base-patch16-256" or "google/siglip2-large-patch16-512" from huggingface or locally)
+## Installation
+### Using pixi.sh:
+`pixi add jimm@https://github.com/pythoncrazy/jimm.git --pypi`
+### Using uv
+`uv add --dev git+https://github.com/pythoncrazy/jimm.git`
+or if you prefer to not add as a direct dependency:
+`uv pip install git+https://github.com/pythoncrazy/jimm.git`
+### Using pip/conda
+`pip install git+https://github.com/pythoncrazy/jimm.git`

jax_image_models-0.3.3/docs/index.md

@@ -0,0 +1,74 @@
+# jimm docs
+
+jimm is a JAX image-model library built on Flax NNX. It supports loading pretrained weights from HuggingFace, hardware-accelerated attention, and FSDP-style explicit sharding.
+
+## Models Implemented
+
+* Vision Transformers (CLS pooling or Multihead Attention Pooling)
+* CLIP
+* SigLIP
+* more tbd — contribute, it's open source!
+
+## Flash / Splash Attention (via Tokamax)
+
+All models accept an `attention_fn` argument for hardware-accelerated attention using [Tokamax](https://github.com/openxla/tokamax):
+
+| Backend | Hardware | Notes |
+|---------|----------|-------|
+| `"mosaic"` | NVIDIA H100 (SM90) / B100 (SM100) | Pallas Mosaic GPU kernel |
+| `"triton"` | Any NVIDIA GPU | Pallas Triton kernel |
+| `"cudnn"` | NVIDIA GPU | Via JAX-NN / cuDNN |
+| `"mosaic_tpu"` | TPU v5 / v7 | Splash attention (block-sparse) |
+| `"xla_chunked"` | GPU / TPU | Flash-style chunked XLA |
+| `"xla"` | Any | Standard XLA fallback |
+
+Pass a list for automatic fallback:
+
+```python
+import jimm
+
+# GPU: try H100 Mosaic kernel, fall back to Triton, then XLA
+model = jimm.CLIP.from_pretrained("openai/clip-vit-large-patch14",
+                                   attention_fn=jimm.make_tokamax_attention(["mosaic", "triton", "xla"]))
+
+# TPU: try Splash attention, fall back to chunked XLA
+model = jimm.CLIP.from_pretrained("openai/clip-vit-large-patch14",
+                                   attention_fn=jimm.make_tokamax_attention(["mosaic_tpu", "xla_chunked"]))
+```
+
+> **Note:** Flash/Splash attention does not provide a speedup at typical vision/text context lengths (e.g. 256 image tokens, 77 text tokens). The primary benefit is memory reduction at longer context lengths.
+
+## FSDP / Explicit Sharding
+
+All models support JAX explicit sharding (FSDP-style) out of the box. Set up a mesh before model creation:
+
+```python
+from jax.experimental import mesh_utils
+from jax.sharding import AxisType, Mesh
+import jax
+
+n_devices = jax.device_count()
+mesh = Mesh(
+    mesh_utils.create_device_mesh((1, n_devices)),
+    ("data", "fsdp"),
+    axis_types=(AxisType.Explicit, AxisType.Explicit),
+)
+jax.set_mesh(mesh)
+
+model = jimm.CLIP.from_pretrained("openai/clip-vit-large-patch14")
+# params are automatically sharded across fsdp axis
+```
+
+Each model ships with a default sharding config (`CLIPSharding`, `SigLIPSharding`, `ViTSharding`) that shards large weight matrices on the contracting (`in_features`) dimension, keeping activations batch-sharded only. Specs are per-layer shapes; the Transformer stack patches Variable metadata after `nnx.vmap` so the optimizer (e.g. AdamW via `nnx.Optimizer`) initialises its state with the correct stacked spec — no manual fixups needed.
+
+Pass `sharding=jimm.common.sharding.NoSharding()` to disable all sharding.
+
+## Installation
+### Using pixi.sh:
+`pixi add jimm@https://github.com/pythoncrazy/jimm.git --pypi`
+### Using uv
+`uv add git+https://github.com/pythoncrazy/jimm.git`
+or if you prefer to not add as a direct dependency:
+`uv pip install git+https://github.com/pythoncrazy/jimm.git`
+### Using pip/conda
+`pip install git+https://github.com/pythoncrazy/jimm.git`

jax_image_models-0.3.3/docs/models/CLIP.md

@@ -0,0 +1,82 @@
+# CLIP (Contrastive Language–Image Pre-training)
+
+CLIP (Contrastive Language–Image Pre-training) is a neural network architecture that learns visual concepts from natural language supervision. It is trained on a large dataset of image-text pairs to create a unified vision-language model that can understand both images and text in a shared semantic space.
+
+CLIP consists of two main components:
+1. A vision encoder (Vision Transformer) that processes images into visual features
+2. A text encoder (Transformer) that processes text into textual features
+
+The model is trained using contrastive learning, where it learns to maximize the cosine similarity between the embeddings of matching image-text pairs while minimizing it for non-matching pairs. This allows CLIP to perform zero-shot classification by comparing image embeddings with text embeddings of potential labels.
+
+CLIP was introduced in the paper ["Learning Transferable Visual Models From Natural Language Supervision"](https://arxiv.org/abs/2103.00020) and has shown remarkable zero-shot generalization capabilities across a wide range of visual classification tasks. The CLIP model combines a Vision Transformer and a Text Transformer to learn joint representations of images and text. It is trained to maximize the similarity between matching image-text pairs while minimizing similarity between non-matching pairs.
+
+## Flash / Splash Attention
+
+CLIP supports hardware-accelerated attention via [Tokamax](https://github.com/openxla/tokamax). Pass an `attention_fn` at construction time:
+
+| Backend | Hardware | Notes |
+|---------|----------|-------|
+| `"mosaic"` | NVIDIA H100 (SM90) / B100 (SM100) | Pallas Mosaic GPU kernel |
+| `"triton"` | Any NVIDIA GPU | Pallas Triton kernel |
+| `"cudnn"` | NVIDIA GPU | Via JAX-NN / cuDNN |
+| `"mosaic_tpu"` | TPU v5 / v7 | Splash attention (block-sparse) |
+| `"xla_chunked"` | GPU / TPU | Flash-style chunked XLA |
+| `"xla"` | Any | Standard XLA fallback |
+
+```python
+import jimm
+
+# GPU: try H100 Mosaic kernel, fall back to Triton, then XLA
+model = jimm.CLIP.from_pretrained("openai/clip-vit-large-patch14",
+                                   attention_fn=jimm.make_tokamax_attention(["mosaic", "triton", "xla"]))
+
+# TPU: try Splash attention, fall back to chunked XLA
+model = jimm.CLIP.from_pretrained("openai/clip-vit-large-patch14",
+                                   attention_fn=jimm.make_tokamax_attention(["mosaic_tpu", "xla_chunked"]))
+```
+
+You can also apply different kernels to each encoder via `vision_attention_fn` and `text_attention_fn`.
+
+> **Note:** Flash/Splash attention does not provide a speedup at typical CLIP context lengths (256 image tokens, 77 text tokens). The primary benefit is memory reduction at longer sequence lengths.
+
+## FSDP / Explicit Sharding
+
+CLIP supports JAX explicit sharding (FSDP-style) out of the box via `CLIPSharding`. Large weight matrices are sharded on the contracting (`in_features`) dimension so that activations carry only the batch-axis sharding, avoiding duplicate-axis conflicts.
+
+```python
+from jax.experimental import mesh_utils
+from jax.sharding import AxisType, Mesh, NamedSharding, PartitionSpec as P
+import jax
+
+n_devices = jax.device_count()
+mesh = Mesh(
+    mesh_utils.create_device_mesh((1, n_devices)),
+    ("data", "fsdp"),
+    axis_types=(AxisType.Explicit, AxisType.Explicit),
+)
+jax.set_mesh(mesh)
+
+model = jimm.CLIP.from_pretrained("openai/clip-vit-large-patch14")
+# model params are automatically sharded across fsdp axis
+```
+
+`CLIPSharding` specs represent **per-layer** shapes. The `Transformer` stack prepends `None` for the scan axis to the Variable metadata after `nnx.vmap`, so the optimizer (e.g. `nnx.Optimizer` with AdamW) receives the correct stacked spec and initialises its state without any manual fixups.
+
+To disable sharding, pass `sharding=jimm.common.sharding.NoSharding()`.
+
+::: jimm.models.clip.CLIPVisionModel
+    options:
+        show_root_heading: true
+        show_source: true
+
+
+::: jimm.models.clip.CLIPTextModel
+    options:
+        show_root_heading: true
+        show_source: true
+
+
+::: jimm.models.clip.CLIP
+    options:
+        show_root_heading: true
+        show_source: true