open_clip

6f43e8fa · mashun1 · 6f43e8fa · 6f43e8fa · 6f43e8fa · 6f43e8fa
Commit 6f43e8fa authored Sep 14, 2024 by mashun1
20 changed files
--- a/.gitattributes
+++ b/.gitattributes
+*.py linguist-language=python
+*.ipynb linguist-documentation
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
+name: Continuous integration
+on:
+  push:
+    branches:
+    - main 
+    paths-ignore:
+    - '**.md'
+    - 'CITATION.cff'
+    - 'LICENSE'
+    - '.gitignore'
+    - 'docs/**'
+  pull_request:
+    branches:
+    - main 
+    paths-ignore:
+    - '**.md'
+    - 'CITATION.cff'
+    - 'LICENSE'
+    - '.gitignore'
+    - 'docs/**'
+  workflow_dispatch:
+    inputs:
+      manual_revision_reference:
+        required: false
+        type: string
+      manual_revision_test:
+        required: false
+        type: string
+env:
+  REVISION_REFERENCE: v2.8.2
+  #9d31b2ec4df6d8228f370ff20c8267ec6ba39383 earliest compatible v2.7.0 + pretrained_hf param
+jobs:
+  Tests:
+    strategy:
+      matrix:
+        os: [ ubuntu-latest ] #, macos-latest ]
+        python: [ 3.8 ]
+        job_num: [ 4 ]
+        job: [ 1, 2, 3, 4 ]
+    runs-on: ${{ matrix.os }}
+    steps:
+    - uses: actions/checkout@v3
+      with:
+        fetch-depth: 0
+        ref: ${{ inputs.manual_revision_test }}
+    - name: Set up Python ${{ matrix.python }}
+      id: pythonsetup
+      uses: actions/setup-python@v4
+      with:
+        python-version: ${{ matrix.python }}
+    - name: Venv cache
+      id: venv-cache
+      uses: actions/cache@v3
+      with:
+        path: .env
+        key: venv-${{ matrix.os }}-${{ steps.pythonsetup.outputs.python-version }}-${{ hashFiles('requirements*') }}
+    - name: Pytest durations cache
+      uses: actions/cache@v3
+      with:
+        path: .test_durations
+        key: test_durations-${{ matrix.os }}-${{ steps.pythonsetup.outputs.python-version }}-${{ matrix.job }}-${{ github.run_id }}
+        restore-keys: test_durations-0-
+    - name: Setup
+      if: steps.venv-cache.outputs.cache-hit != 'true'
+      run: |
+        python3 -m venv .env
+        source .env/bin/activate
+        pip install -e .[test]
+    - name: Prepare test data
+      run: |
+        source .env/bin/activate
+        python -m pytest \
+          --quiet --co \
+          --splitting-algorithm least_duration \
+          --splits ${{ matrix.job_num }} \
+          --group ${{ matrix.job }} \
+          -m regression_test \
+          tests \
+          | head -n -2 | grep -Po 'test_inference_with_data\[\K[^]]*(?=-False]|-True])' \
+          > models_gh_runner.txt
+        if [ -n "${{ inputs.manual_revision_reference }}" ]; then
+          REVISION_REFERENCE=${{ inputs.manual_revision_reference }}
+        fi
+        python tests/util_test.py \
+          --save_model_list models_gh_runner.txt \
+          --model_list models_gh_runner.txt \
+          --git_revision $REVISION_REFERENCE
+    - name: Unit tests
+      run: |
+        source .env/bin/activate
+        if [[ -f .test_durations ]]
+        then
+          cp .test_durations durations_1
+          mv .test_durations durations_2
+        fi
+        python -m pytest \
+          -x -s -v \
+          --splitting-algorithm least_duration \
+          --splits ${{ matrix.job_num }} \
+          --group ${{ matrix.job }} \
+          --store-durations \
+          --durations-path durations_1 \
+          --clean-durations \
+          -m "not regression_test" \
+          tests
+        OPEN_CLIP_TEST_REG_MODELS=models_gh_runner.txt python -m pytest \
+          -x -s -v \
+          --store-durations \
+          --durations-path durations_2 \
+          --clean-durations \
+          -m "regression_test" \
+          tests
+        jq -s -S 'add' durations_* > .test_durations
+    - name: Collect pytest durations
+      uses: actions/upload-artifact@v3
+      with:
+        name: pytest_durations_${{ matrix.os }}-${{ matrix.python }}-${{ matrix.job }}
+        path: .test_durations
+  Collect:
+    needs: Tests
+    runs-on: ubuntu-latest
+    steps:
+    - name: Cache
+      uses: actions/cache@v3
+      with:
+        path: .test_durations
+        key: test_durations-0-${{ github.run_id }}
+    - name: Collect
+      uses: actions/download-artifact@v3
+      with:
+        path: artifacts
+    - name: Consolidate
+      run: |
+        jq -n -S \
+          'reduce (inputs | to_entries[]) as {$key, $value} ({}; .[$key] += $value)' \
+          artifacts/pytest_durations_*/.test_durations > .test_durations
--- a/.github/workflows/clear-cache.yml
+++ b/.github/workflows/clear-cache.yml
+name: Clear cache
+on:
+  workflow_dispatch:
+permissions:
+  actions: write
+jobs:
+  clear-cache:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Clear cache
+        uses: actions/github-script@v6
+        with:
+          script: |
+            const caches = await github.rest.actions.getActionsCacheList({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+            })
+            for (const cache of caches.data.actions_caches) {
+              console.log(cache)
+              await github.rest.actions.deleteActionsCacheById({
+                owner: context.repo.owner,
+                repo: context.repo.repo,
+                cache_id: cache.id,
+              })
+            }
--- a/.github/workflows/python-publish.yml
+++ b/.github/workflows/python-publish.yml
+name: Release
+on:
+  push:
+    branches:
+    - main
+jobs:
+  deploy:
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/checkout@v2
+    - uses: actions-ecosystem/action-regex-match@v2
+      id: regex-match
+      with:
+        text: ${{ github.event.head_commit.message }}
+        regex: '^Release ([^ ]+)'
+    - name: Set up Python
+      uses: actions/setup-python@v2
+      with:
+        python-version: '3.8'
+    - name: Install dependencies
+      run: |
+        python -m pip install --upgrade pip
+        pip install setuptools wheel twine build
+    - name: Release
+      if: ${{ steps.regex-match.outputs.match != '' }}
+      uses: softprops/action-gh-release@v1
+      with:
+        tag_name: v${{ steps.regex-match.outputs.group1 }}
+    - name: Build and publish
+      if: ${{ steps.regex-match.outputs.match != '' }}
+      env:
+        TWINE_USERNAME: __token__
+        TWINE_PASSWORD: ${{ secrets.PYPI_PASSWORD }}
+      run: |
+        python -m build
+        twine upload dist/*
--- a/.gitignore
+++ b/.gitignore
+**/logs/
+**/wandb/
+models/
+features/
+results/
+tests/data/
+*.pt
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+.pdm-build/
+# C extensions
+*.so
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+pip-wheel-metadata/
+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/
+# 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
+target/
+# Jupyter Notebook
+.ipynb_checkpoints
+# IPython
+profile_default/
+ipython_config.py
+# pyenv
+.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
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow
+__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/
+sync.sh
+gpu1sync.sh
+.idea
+*.pdf
+**/._*
+**/*DS_*
+**.jsonl
+src/sbatch
+src/misc
+.vscode
+src/debug
+core.*
+# Allow
+!src/evaluation/misc/results_dbs/*
+# self
+scripts/single_train_test.sh
+pretrained_models/
+logs/
+dist/
+datasets/
\ No newline at end of file
--- a/CITATION.cff
+++ b/CITATION.cff
+cff-version: 1.1.0
+message: If you use this software, please cite it as below.
+authors:
+  - family-names: Ilharco
+    given-names: Gabriel
+  - family-names: Wortsman
+    given-names: Mitchell
+  - family-names: Wightman
+    given-names: Ross
+  - family-names: Gordon
+    given-names: Cade   
+  - family-names: Carlini
+    given-names: Nicholas
+  - family-names: Taori
+    given-names: Rohan
+  - family-names: Dave
+    given-names: Achal
+  - family-names: Shankar
+    given-names: Vaishaal
+  - family-names: Namkoong
+    given-names: Hongseok
+  - family-names: Miller
+    given-names: John
+  - family-names: Hajishirzi
+    given-names: Hannaneh
+  - family-names: Farhadi
+    given-names: Ali
+  - family-names: Schmidt
+    given-names: Ludwig
+title: OpenCLIP
+version: v0.1
+doi: 10.5281/zenodo.5143773
+date-released: 2021-07-28
--- a/Dockerfile
+++ b/Dockerfile
+FROM image.sourcefind.cn:5000/dcu/admin/base/pytorch:2.1.0-ubuntu20.04-dtk24.04.1-py3.10
\ No newline at end of file
--- a/HISTORY.md
+++ b/HISTORY.md
+## 2.24.0
+* Fix missing space in error message
+* use model flag for normalizing embeddings
+* init logit_bias for non siglip pretrained models
+* Fix logit_bias load_checkpoint addition 
+* Make CoCa model match CLIP models for logit scale/bias init
+* Fix missing return of "logit_bias" in CoCa.forward
+* Add NLLB-CLIP with SigLIP models
+* Add get_logits method and NLLB tokenizer
+* Remove the empty file src/open_clip/generation_utils.py
+* Update params.py: "BatchNorm" -> "LayerNorm" in the description string for "--lock-text-freeze-layer-norm"
+## 2.23.0
+* Add CLIPA-v2 models
+* Add SigLIP models
+* Add MetaCLIP models
+* Add NLLB-CLIP models
+* CLIPA train code
+* Minor changes/fixes
+    * Remove protobuf version limit
+    * Stop checking model name when loading CoCa models
+    * Log native wandb step
+    * Use bool instead of long masks
+## 2.21.0
+* Add SigLIP loss + training support
+* Add more DataComp models (B/16, B/32 and B/32@256)
+* Update default num workers
+* Update CoCa generation for `transformers>=4.31`
+* PyTorch 2.0 `state_dict()` compatibility fix for compiled models
+* Fix padding in `ResizeMaxSize`
+* Convert JIT model on state dict load for `pretrained='filename…'`
+* Other minor changes and fixes (typos, README, dependencies, CI)
+## 2.20.0
+* Add EVA models
+* Support serial worker training
+* Fix Python 3.7 compatibility 
+## 2.19.0
+* Add DataComp models
+## 2.18.0
+* Enable int8 inference without `.weight` attribute
+## 2.17.2
+* Update push_to_hf_hub
+## 2.17.0
+* Add int8 support
+* Update notebook demo
+* Refactor zero-shot classification code
+## 2.16.2
+* Fixes for context_length and vocab_size attributes 
+## 2.16.1
+* Fixes for context_length and vocab_size attributes 
+* Fix --train-num-samples logic
+* Add HF BERT configs for PubMed CLIP model
+## 2.16.0
+* Add improved g-14 weights
+* Update protobuf version
+## 2.15.0
+* Add convnext_xxlarge weights
+* Fixed import in readme
+* Add samples per second per gpu logging
+* Fix slurm example
+## 2.14.0
+* Move dataset mixtures logic to shard level
+* Fix CoCa accum-grad training
+* Safer transformers import guard
+* get_labels refactoring
+## 2.13.0
+* Add support for dataset mixtures with different sampling weights
+* Make transformers optional again 
+## 2.12.0
+* Updated convnext configs for consistency
+* Added input_patchnorm option
+* Clean and improve CoCa generation
+* Support model distillation
+* Add ConvNeXt-Large 320x320 fine-tune weights
+## 2.11.1
+* Make transformers optional
+* Add MSCOCO CoCa finetunes to pretrained models
+## 2.11.0
+* coca support and weights
+* ConvNeXt-Large weights
+## 2.10.1
+* `hf-hub:org/model_id` support for loading models w/ config and weights in Hugging Face Hub
+## 2.10.0
+* Added a ViT-bigG-14 model.
+* Added an up-to-date example slurm script for large training jobs.
+* Added a option to sync logs and checkpoints to S3 during training.
+* New options for LR schedulers, constant and constant with cooldown
+* Fix wandb autoresuming when resume is not set
+* ConvNeXt `base` & `base_w` pretrained models added
+* `timm-` model prefix removed from configs
+* `timm` augmentation + regularization (dropout / drop-path) supported
+## 2.9.3
+* Fix wandb collapsing multiple parallel runs into a single one
+## 2.9.2
+* Fix braceexpand memory explosion for complex webdataset urls
+## 2.9.1
+* Fix release
+## 2.9.0
+* Add training feature to auto-resume from the latest checkpoint on restart via `--resume latest`
+* Allow webp in webdataset
+* Fix logging for number of samples when using gradient accumulation
+* Add model configs for convnext xxlarge
+## 2.8.2
+* wrapped patchdropout in a torch.nn.Module
+## 2.8.1
+* relax protobuf dependency
+* override the default patch dropout value in 'vision_cfg'
+## 2.8.0
+* better support for HF models
+* add support for gradient accumulation
+* CI fixes
+* add support for patch dropout
+* add convnext configs
+## 2.7.0
+* add multilingual H/14 xlm roberta large
+## 2.6.1
+* fix setup.py _read_reqs
+## 2.6.0
+* Make openclip training usable from pypi.
+* Add xlm roberta large vit h 14 config.
+## 2.5.0
+* pretrained B/32 xlm roberta base: first multilingual clip trained on laion5B
+* pretrained B/32 roberta base: first clip trained using an HF text encoder
+## 2.4.1
+* Add missing hf_tokenizer_name in CLIPTextCfg.
+## 2.4.0
+* Fix #211, missing RN50x64 config. Fix type of dropout param for ResNet models
+* Bring back LayerNorm impl that casts to input for non bf16/fp16 
+* zero_shot.py: set correct tokenizer based on args
+* training/params.py: remove hf params and get them from model config
+## 2.3.1
+* Implement grad checkpointing for hf model.
+* custom_text: True if hf_model_name is set
+* Disable hf tokenizer parallelism 
+## 2.3.0
+* Generalizable Text Transformer with HuggingFace Models (@iejMac)
+## 2.2.0
+* Support for custom text tower
+* Add checksum verification for pretrained model weights 
+## 2.1.0
+* lot including sota models, bfloat16 option, better loading, better metrics
+## 1.2.0
+* ViT-B/32 trained on Laion2B-en
+* add missing openai RN50x64 model
+## 1.1.1
+* ViT-B/16+
+* Add grad checkpointing support
+* more robust data loader
--- a/LICENSE
+++ b/LICENSE
+Copyright (c) 2012-2021 Gabriel Ilharco, Mitchell Wortsman, 
+Nicholas Carlini, Rohan Taori, Achal Dave, Vaishaal Shankar, 
+John Miller, Hongseok Namkoong, Hannaneh Hajishirzi, Ali Farhadi, 
+Ludwig Schmidt
+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.
--- a/MANIFEST.in
+++ b/MANIFEST.in
+include src/open_clip/bpe_simple_vocab_16e6.txt.gz
+include src/open_clip/model_configs/*.json
--- a/Makefile
+++ b/Makefile
+install: ## [Local development] Upgrade pip, install requirements, install package.
+	python -m pip install -U pip
+	python -m pip install -e .
+install-training:
+	python -m pip install -r requirements-training.txt
+install-test: ## [Local development] Install test requirements
+	python -m pip install -r requirements-test.txt
+test: ## [Local development] Run unit tests
+	python -m pytest -x -s -v tests
--- a/README.md
+++ b/README.md
+# open-clip
+## 论文
+**Reproducible scaling laws for contrastive language-image learning**
+* https://arxiv.org/pdf/2212.07143
+## 模型结构
+CLIP 模型有两个主要组件，一个文本编码器和一个图像编码器。对于文本编码器，使用了`Transformer`；对于图像编码器采用了`ResNet`和`Vision Transformer（ViT）`。
+![alt text](readme_imgs/model.png)
+## 算法原理
+CLIP通过最大化`文本-图像`匹配得分同时训练一个图像编码器和文本编码器。
+![alt text](readme_imgs/alg.png)
+## 环境配置
+### Docker（方法一）
+    docker pull image.sourcefind.cn:5000/dcu/admin/base/pytorch:2.1.0-ubuntu20.04-dtk24.04.1-py3.10
+    docker run --shm-size 50g --network=host --name=clip --privileged --device=/dev/kfd --device=/dev/dri --group-add video --cap-add=SYS_PTRACE --security-opt seccomp=unconfined -v 项目地址(绝对路径):/home/ -v /opt/hyhal:/opt/hyhal:ro -it <your IMAGE ID> bash
+    pip install timm --no-deps
+    pip install -r requirements.txt
+    pip install build 
+    python -m build && pip install dist/open_clip_torch-2.26.1-py3-none-any.whl
+### Dockerfile（方法二）
+    docker build -t <IMAGE_NAME>:<TAG> .
+    docker run --shm-size 50g --network=host --name=clip --privileged --device=/dev/kfd --device=/dev/dri --group-add video --cap-add=SYS_PTRACE --security-opt seccomp=unconfined -v 项目地址(绝对路径):/home/ -v /opt/hyhal:/opt/hyhal:ro -it <your IMAGE ID> bash
+    pip install timm --no-deps
+    pip install -r requirements.txt
+    pip install build 
+    python -m build && pip install dist/open_clip_torch-2.26.1-py3-none-any.whl
+### Anaconda (方法三)
+1、关于本项目DCU显卡所需的特殊深度学习库可从光合开发者社区下载安装：
+https://developer.hpccube.com/tool/
+    DTK驱动：dtk24.04.1
+    python：python3.10
+    torch: 2.1.0
+    torchvision: 0.16.0
+Tips：以上dtk驱动、python、torch等DCU相关工具版本需要严格一一对应
+2、其它非特殊库参照requirements.txt安装
+    pip install timm --no-deps
+    pip install -r requirements.txt
+    pip install build 
+    python -m build && pip install dist/open_clip_torch-2.26.1-py3-none-any.whl
+## 数据集
+### CLIP
+#### 数据下载
+完整数据集下载参考[img2dataset](https://github.com/rom1504/img2dataset)。
+这里提供其他数据集，可通过[SCNet高速通道下载](http://113.200.138.88:18080/aidatasets/)。
+- 训练数据集：[cc3m wds](http://113.200.138.88:18080/aidatasets/pixparse/cc3m-wds)
+- 验证数据集：[imagenet2012-val](http://113.200.138.88:18080/aidatasets/project-dependency/imagenet-2012/)，数据下载完成后在数据目录下运行`bash valprep.sh`对数据进行分类(将extra_utils下的`valprep.sh`移动到数据目录下)。
+#### 数据处理
+在准备数据集时，可以使用`img2dataset`直接下载的数据集(`.tar`)，也可自行准备数据集，自行准备的数据集可以`image_path caption`的形式并以`csv`格式保存，本项目提供了相应的脚本，位于`extra_utils/convert_to_csv.py`。
+|filepath|title|
+|:-------:|:-------:|
+|<path/to/image>|a white cat|
+注意：更多数据处理及使用方式，可参考[官方readme](README_official.md)。
+### CoCa
+运行`extra_utils/coca_data.py`会自动下载并处理数据。也可通过[SCNet高速下载通道](http://113.200.138.88:18080/aidatasets/project-dependency/mscoco2014)下载数据集并修改代码中`root_path`后运行该脚本处理数据。
+## 训练
+### CLIP
+#### 单卡运行
+    python -m open_clip_train.main \
+    --save-frequency 1 \
+    --zeroshot-frequency 1 \
+    --report-to tensorboard \
+    --train-data="/path/to/train_data.csv"  \
+    --val-data="/path/to/validation_data.csv"  \
+    --csv-img-key filepath \
+    --csv-caption-key title \
+    --imagenet-val=/path/to/imagenet/root/val/ \
+    --warmup 10000 \
+    --batch-size=128 \
+    --lr=1e-3 \
+    --wd=0.1 \
+    --epochs=30 \
+    --workers=8 \
+    --model RN50 \
+    --csv-separator "," \
+    --precision amp_bf16 
+注意：`imagenet-val`表示使用imagenet作为验证集，若不需要可以删除。
+#### 多卡运行
+##### 单节点
+    cd open_clip/src
+    torchrun --nproc_per_node 4 -m open_clip_train.main \
+    --train-data '/data/cc12m/cc12m-train-{0000..2175}.tar' \
+    --train-num-samples 10968539 \
+    --dataset-type webdataset \
+    --batch-size 320 \
+    --precision amp_bf16 \
+    --workers 4 \
+    --imagenet-val /data/imagenet/validation/ 
+##### 多节点
+    cd open_clip/src
+    torchrun --nproc_per_node=4 \
+    --rdzv_endpoint=$HOSTE_NODE_ADDR \
+    -m open_clip_train.main \
+    --train-data '/data/cc12m/cc12m-train-{0000..2175}.tar' \
+    --train-num-samples 10968539 \
+    --dataset-type webdataset \
+    --batch-size 320 \
+    --precision amp_bf16 \
+    --workers 4 \
+    --imagenet-val /data/imagenet/validation/ \
+    --precision amp_bf16 
+注意：更多方式请参考[README_official.md](README_official.md)，使用`fp16`训练会导致`nan`，需要替换为`bf16`。
+### CoCa
+```python
+python -m open_clip_train.main \
+    --dataset-type "csv" \
+    --train-data /path/to/your.csv \
+    --warmup 1000 \
+    --batch-size 32 \
+    --lr 1e-5 \
+    --wd 0.1 \
+    --epochs 1 \
+    --workers 3 \
+    --model "coca_ViT-L-14" \
+    --coca-contrastive-loss-weight 0 \
+    --coca-caption-loss-weight 1 \
+    --log-every-n-steps 100
+```
+## 推理
+### CLIP
+    # 不提供参数会使用默认参数并自动下载模型
+    # export HF_ENDPOINT=https://hf-mirror.com
+    python custom_tests/simple_test.py --model_name <xxx> --pretrained <可以是本地路径/path/to/open_clip_pytorch_model.bin，也可以是远程仓库名(如：laion2b_s34b_b79k)>
+### CoCa
+    # 可以仅输入image_path，其他使用默认参数
+    # export HF_ENDPOINT=https://hf-mirror.com
+    python custom_tests/coca_simple_test.py <path/to/image>
+注意：可使用以下参数修改使用的模型`--model_name <xxx> --pretrained <可以是本地路径/path/to/open_clip_pytorch_model.bin，也可以是远程仓库名(如：mscoco_finetuned_laion2B-s13B-b90k)> `
+## result
+`python custom_test.py`
+CLIP: 
+|a diagram|a dog|a cat|
+|:---:|:---:|:---:|
+|0.995|0.004|0.001|
+CoCa:
+输入：<img src="readme_imgs/girl.png" style="zoom:50%;">
+输出：a beautiful young woman with a long black ponytail .
+### 精度
+K100_AI与L20精度保持一致。
+## 应用场景
+### 算法类别
+`图像分类`
+### 热点应用行业
+`电商,绘画,交通`
+## 预训练权重
+下载模型后放入`pretrained_models`文件夹中（需要自行创建）。
+完整模型列表可通过`python -c "import open_clip; print(open_clip.list_pretrained())"`查看，这里提供常用的预训练权重链接。
+|model|链接|
+|:---:|:------:|
+|CLIP-ViT-B-32-laion2B-s34B-b79K|[原始链接](https://huggingface.co/laion/CLIP-ViT-B-32-laion2B-s34B-b79K)/[SCNet高速下载通道](http://113.200.138.88:18080/aimodels/CLIP-ViT-B-32-laion2B-s34B-b79K)||
+|CLIP-ViT-B-16-laion2B-s34B-b88K|[原始链接](https://huggingface.co/laion/CLIP-ViT-B-16-laion2B-s34B-b88K)/[SCNet高速下载通道](http://113.200.138.88:18080/aimodels/CLIP-ViT-B-16-laion2B-s34B-b88K)|
+|CLIP-ViT-H-14-laion2B-s32B-b79K|[原始链接](https://huggingface.co/laion/CLIP-ViT-H-14-laion2B-s32B-b79K)/[SCNet高速下载通道](http://113.200.138.88:18080/aimodels/CLIP-ViT-H-14-laion2B-s32B-b79K)|
+|CLIP-ViT-bigG-14-laion2B-39B-b160k|[原始链接](https://huggingface.co/laion/CLIP-ViT-bigG-14-laion2B-39B-b160k)/[SCNet高速下载通道](http://113.200.138.88:18080/aimodels/CLIP-ViT-bigG-14-laion2B-39B-b160k)|
+|CLIP-convnext_large_d_320.laion2B-s29B-b131K-ft-soup|[原始链接](https://huggingface.co/laion/CLIP-convnext_large_d_320.laion2B-s29B-b131K-ft-soup)/[SCNet高速下载通道](http://113.200.138.88:18080/aimodels/CLIP-convnext_large_d_320.laion2B-s29B-b131K-ft-soup)|
+|bioclip|[原始链接](https://huggingface.co/imageomics/bioclip)/[SCNet高速下载通道](http://113.200.138.88:18080/aimodels/imageomics/bioclip)|
+|mscoco_finetuned_CoCa-ViT-L-14-laion2B-s13B-b90k |[原始链接](https://hf-mirror.com/laion/mscoco_finetuned_CoCa-ViT-L-14-laion2B-s13B-b90k/tree/main)/[SCNet高速下载通道](http://113.200.138.88:18080/aimodels/mscoco_finetuned_CoCa-ViT-L-14-laion2B-s13B-b90k)|
+## 源码仓库及问题反馈
+* https://developer.hpccube.com/codes/modelzoo/clip_pytorch
+## 参考资料
+* https://github.com/mlfoundations/open_clip/
--- a/README_official.md
+++ b/README_official.md
+# OpenCLIP
+[[Paper]](https://arxiv.org/abs/2212.07143) [[Citations]](#citing) [[Clip Colab]](https://colab.research.google.com/github/mlfoundations/open_clip/blob/master/docs/Interacting_with_open_clip.ipynb) [[Coca Colab]](https://colab.research.google.com/github/mlfoundations/open_clip/blob/master/docs/Interacting_with_open_coca.ipynb)
+[![pypi](https://img.shields.io/pypi/v/open_clip_torch.svg)](https://pypi.python.org/pypi/open_clip_torch)
+Welcome to an open source implementation of OpenAI's [CLIP](https://arxiv.org/abs/2103.00020) (Contrastive Language-Image Pre-training).
+Using this codebase, we have trained several models on a variety of data sources and compute budgets, ranging from [small-scale experiments](docs/LOW_ACC.md) to larger runs including models trained on datasets such as [LAION-400M](https://arxiv.org/abs/2111.02114), [LAION-2B](https://arxiv.org/abs/2210.08402) and [DataComp-1B](https://arxiv.org/abs/2304.14108).
+Many of our models and their scaling properties are studied in detail in the paper [reproducible scaling laws for contrastive language-image learning](https://arxiv.org/abs/2212.07143).
+Some of the best models we've trained and their zero-shot ImageNet-1k accuracy are shown below, along with the ViT-L model trained by OpenAI and other state-of-the-art open source alternatives (all can be loaded via OpenCLIP).
+We provide more details about our full collection of pretrained models [here](docs/PRETRAINED.md), and zero-shot results for 38 datasets [here](docs/openclip_results.csv).
+| Model    | Training data | Resolution | # of samples seen | ImageNet zero-shot acc. | 
+| -------- | ------- |  ------- |  ------- |  ------- |  
+| ConvNext-Base | LAION-2B  | 256px | 13B | 71.5% |
+| ConvNext-Large | LAION-2B  | 320px | 29B | 76.9% |
+| ConvNext-XXLarge | LAION-2B | 256px | 34B | 79.5% |
+| ViT-B/32  | DataComp-1B  | 256px | 34B | 72.8% |
+| ViT-B/16  | DataComp-1B  | 224px | 13B | 73.5% |
+| ViT-L/14  | LAION-2B  | 224px | 32B | 75.3% |
+| ViT-H/14  | LAION-2B  | 224px | 32B | 78.0% |
+| ViT-L/14  | DataComp-1B  | 224px | 13B | 79.2% |
+| ViT-G/14  | LAION-2B  | 224px | 34B | 80.1% |
+|  |  |   |   |  |
+| ViT-L/14 [(Original CLIP)](https://arxiv.org/abs/2103.00020) | WIT | 224px | 13B | 75.5% | 
+| ViT-SO400M/14 [(SigLIP)](https://arxiv.org/abs/2303.15343) | WebLI | 224px | 45B | 82.0% | 
+| ViT-SO400M-14-SigLIP-384 [(SigLIP)](https://arxiv.org/abs/2303.15343) |  WebLI | 384px | 45B | 83.1% |
+| ViT-H/14-quickgelu [(DFN)](https://arxiv.org/abs/2309.17425) | DFN-5B | 224px | 39B | 83.4% | 
+| ViT-H-14-378-quickgelu [(DFN)](https://arxiv.org/abs/2309.17425) | DFN-5B | 378px | 44B | 84.4% |
+Model cards with additional model specific details can be found on the Hugging Face Hub under the OpenCLIP library tag: https://huggingface.co/models?library=open_clip. 
+If you found this repository useful, please consider [citing](#citing).
+We welcome anyone to submit an issue or send an email if you have any other requests or suggestions.
+Note that portions of `src/open_clip/` modelling and tokenizer code are adaptations of OpenAI's official [repository](https://github.com/openai/CLIP).
+## Approach
+| ![CLIP](https://raw.githubusercontent.com/mlfoundations/open_clip/main/docs/CLIP.png) |
+|:--:|
+| Image Credit: https://github.com/openai/CLIP |
+## Usage
+```
+pip install open_clip_torch
+```
+```python
+import torch
+from PIL import Image
+import open_clip
+model, _, preprocess = open_clip.create_model_and_transforms('ViT-B-32', pretrained='laion2b_s34b_b79k')
+model.eval()  # model in train mode by default, impacts some models with BatchNorm or stochastic depth active
+tokenizer = open_clip.get_tokenizer('ViT-B-32')
+image = preprocess(Image.open("docs/CLIP.png")).unsqueeze(0)
+text = tokenizer(["a diagram", "a dog", "a cat"])
+with torch.no_grad(), torch.cuda.amp.autocast():
+    image_features = model.encode_image(image)
+    text_features = model.encode_text(text)
+    image_features /= image_features.norm(dim=-1, keepdim=True)
+    text_features /= text_features.norm(dim=-1, keepdim=True)
+    text_probs = (100.0 * image_features @ text_features.T).softmax(dim=-1)
+print("Label probs:", text_probs)  # prints: [[1., 0., 0.]]
+```
+See also this [[Clip Colab]](https://colab.research.google.com/github/mlfoundations/open_clip/blob/master/docs/Interacting_with_open_clip.ipynb).
+To compute billions of embeddings efficiently, you can use [clip-retrieval](https://github.com/rom1504/clip-retrieval) which has openclip support.
+### Pretrained models
+We offer a simple model interface to instantiate both pre-trained and untrained models.
+To see which pretrained models are available, use the following code snippet.
+More details about our pretrained models are available [here](docs/PRETRAINED.md).
+```python
+>>> import open_clip
+>>> open_clip.list_pretrained()
+```
+You can find more about the models we support (e.g. number of parameters, FLOPs) in [this table](docs/model_profile.csv).
+NOTE: Many existing checkpoints use the QuickGELU activation from the original OpenAI models. This activation is actually less efficient than native torch.nn.GELU in recent versions of PyTorch. The model defaults are now nn.GELU, so one should use model definitions with `-quickgelu` postfix for the OpenCLIP pretrained weights. All OpenAI pretrained weights will always default to QuickGELU. One can also use the non `-quickgelu` model definitions with pretrained weights using QuickGELU but there will be an accuracy drop, for fine-tune that will likely vanish for longer runs.
+Future trained models will use nn.GELU.
+### Loading models
+Models can be loaded with `open_clip.create_model_and_transforms`, as shown in the example below. The model name and corresponding `pretrained` keys are compatible with the outputs of `open_clip.list_pretrained()`. 
+The `pretrained` argument also accepts local paths, for example `/path/to/my/b32.pt`.
+You can also load checkpoints from huggingface this way. To do so, download the `open_clip_pytorch_model.bin` file (for example, [https://huggingface.co/laion/CLIP-ViT-L-14-DataComp.XL-s13B-b90K/tree/main](https://huggingface.co/laion/CLIP-ViT-L-14-DataComp.XL-s13B-b90K/blob/main/open_clip_pytorch_model.bin)), and use `pretrained=/path/to/open_clip_pytorch_model.bin`.
+```python
+# pretrained also accepts local paths
+model, _, preprocess = open_clip.create_model_and_transforms('ViT-B-32', pretrained='laion2b_s34b_b79k') 
+```
+## Fine-tuning on classification tasks
+This repository is focused on training CLIP models. To fine-tune a *trained* zero-shot model on a downstream classification task such as ImageNet, please see [our other repository: WiSE-FT](https://github.com/mlfoundations/wise-ft). The [WiSE-FT repository](https://github.com/mlfoundations/wise-ft) contains code for our paper on [Robust Fine-tuning of Zero-shot Models](https://arxiv.org/abs/2109.01903), in which we introduce a technique for fine-tuning zero-shot models while preserving robustness under distribution shift.
+## Data
+To download datasets as webdataset, we recommend [img2dataset](https://github.com/rom1504/img2dataset).
+### Conceptual Captions
+See [cc3m img2dataset example](https://github.com/rom1504/img2dataset/blob/main/dataset_examples/cc3m.md).
+### YFCC and other datasets
+In addition to specifying the training data via CSV files as mentioned above, our codebase also supports [webdataset](https://github.com/webdataset/webdataset), which is recommended for larger scale datasets. The expected format is a series of `.tar` files. Each of these `.tar` files should contain two files for each training example, one for the image and one for the corresponding text. Both files should have the same name but different extensions. For instance, `shard_001.tar` could contain files such as `abc.jpg` and `abc.txt`. You can learn more about `webdataset` at [https://github.com/webdataset/webdataset](https://github.com/webdataset/webdataset). We use `.tar` files with 1,000 data points each, which we create using [tarp](https://github.com/webdataset/tarp).
+You can download the YFCC dataset from [Multimedia Commons](http://mmcommons.org/).
+Similar to OpenAI, we used a subset of YFCC to reach the aforementioned accuracy numbers.
+The indices of images in this subset are in [OpenAI's CLIP repository](https://github.com/openai/CLIP/blob/main/data/yfcc100m.md).
+## Training CLIP
+### Install
+We advise you first create a virtual environment with:
+```
+python3 -m venv .env
+source .env/bin/activate
+pip install -U pip
+```
+You can then install openclip for training with `pip install 'open_clip_torch[training]'`.
+#### Development
+If you want to make changes to contribute code, you can clone openclip then run `make install` in openclip folder (after creating a virtualenv)
+Install pip PyTorch as per https://pytorch.org/get-started/locally/
+You may run `make install-training` to install training deps
+#### Testing
+Test can be run with `make install-test` then `make test`
+`python -m pytest -x -s -v tests -k "training"` to run a specific test
+Running regression tests against a specific git revision or tag:
+1. Generate testing data
+    ```sh
+    python tests/util_test.py --model RN50 RN101 --save_model_list models.txt --git_revision 9d31b2ec4df6d8228f370ff20c8267ec6ba39383
+    ```
+    **_WARNING_: This will invoke git and modify your working tree, but will reset it to the current state after data has been generated! \
+    Don't modify your working tree while test data is being generated this way.**
+2. Run regression tests
+    ```sh
+    OPEN_CLIP_TEST_REG_MODELS=models.txt python -m pytest -x -s -v -m regression_test
+    ```
+### Sample single-process running code:
+```bash
+python -m open_clip_train.main \
+    --save-frequency 1 \
+    --zeroshot-frequency 1 \
+    --report-to tensorboard \
+    --train-data="/path/to/train_data.csv"  \
+    --val-data="/path/to/validation_data.csv"  \
+    --csv-img-key filepath \
+    --csv-caption-key title \
+    --imagenet-val=/path/to/imagenet/root/val/ \
+    --warmup 10000 \
+    --batch-size=128 \
+    --lr=1e-3 \
+    --wd=0.1 \
+    --epochs=30 \
+    --workers=8 \
+    --model RN50
+```
+Note: `imagenet-val` is the path to the *validation* set of ImageNet for zero-shot evaluation, not the training set!
+You can remove this argument if you do not want to perform zero-shot evaluation on ImageNet throughout training. Note that the `val` folder should contain subfolders. If it does not, please use [this script](https://raw.githubusercontent.com/soumith/imagenetloader.torch/master/valprep.sh).
+### Multi-GPU and Beyond
+This code has been battle tested up to 1024 A100s and offers a variety of solutions
+for distributed training. We include native support for SLURM clusters.
+As the number of devices used to train increases, so does the space complexity of
+the the logit matrix. Using a naïve all-gather scheme, space complexity will be
+`O(n^2)`. Instead, complexity may become effectively linear if the flags
+`--gather-with-grad` and `--local-loss` are used. This alteration results in one-to-one
+numerical results as the naïve method.
+#### Epochs
+For larger datasets (eg Laion2B), we recommend setting `--train-num-samples` to a lower value than the full epoch, for example `--train-num-samples 135646078` to 1/16 of an epoch in conjunction with `--dataset-resampled` to do sampling with replacement. This allows having frequent checkpoints to evaluate more often.
+#### Patch Dropout
+<a href="https://arxiv.org/abs/2212.00794">Recent research</a> has shown that one can dropout half to three-quarters of the visual tokens, leading to up to 2-3x training speeds without loss of accuracy.
+You can set this on your visual transformer config with the key `patch_dropout`.
+In the paper, they also finetuned without the patch dropout at the end. You can do this with the command-line argument `--force-patch-dropout 0.`
+#### Multiple data sources
+OpenCLIP supports using multiple data sources, by separating different data paths with `::`.
+For instance, to train on CC12M and on LAION, one might use `--train-data "/data/cc12m/cc12m-train-{0000..2175}.tar::/data/LAION-400M/{00000..41455}.tar"`.
+Using `--dataset-resampled` is recommended for these cases.
+By default, on expectation the amount of times the model will see a sample from each source is proportional to the size of the source.
+For instance, when training on one data source with size 400M and one with size 10M, samples from the first source are 40x more likely to be seen in expectation.
+We also support different weighting of the data sources, by using the `--train-data-upsampling-factors` flag.
+For instance, using `--train-data-upsampling-factors=1::1` in the above scenario is equivalent to not using the flag, and `--train-data-upsampling-factors=1::2` is equivalent to upsampling the second data source twice.
+If you want to sample from data sources with the same frequency, the upsampling factors should be inversely proportional to the sizes of the data sources.
+For instance, if dataset `A` has 1000 samples and dataset `B` has 100 samples, you can use `--train-data-upsampling-factors=0.001::0.01` (or analogously, `--train-data-upsampling-factors=1::10`).
+#### Single-Node
+We make use of `torchrun` to launch distributed jobs. The following launches a
+a job on a node of 4 GPUs:
+```bash
+cd open_clip/src
+torchrun --nproc_per_node 4 -m open_clip_train.main \
+    --train-data '/data/cc12m/cc12m-train-{0000..2175}.tar' \
+    --train-num-samples 10968539 \
+    --dataset-type webdataset \
+    --batch-size 320 \
+    --precision amp \
+    --workers 4 \
+    --imagenet-val /data/imagenet/validation/
+```
+#### Multi-Node
+The same script above works, so long as users include information about the number
+of nodes and host node.
+```bash
+cd open_clip/src
+torchrun --nproc_per_node=4 \
+    --rdzv_endpoint=$HOSTE_NODE_ADDR \
+    -m open_clip_train.main \
+    --train-data '/data/cc12m/cc12m-train-{0000..2175}.tar' \
+    --train-num-samples 10968539 \
+    --dataset-type webdataset \
+    --batch-size 320 \
+    --precision amp \
+    --workers 4 \
+    --imagenet-val /data/imagenet/validation/
+```
+#### SLURM
+This is likely the easiest solution to utilize. The following script was used to
+train our largest models:
+```bash
+#!/bin/bash -x
+#SBATCH --nodes=32
+#SBATCH --gres=gpu:4
+#SBATCH --ntasks-per-node=4
+#SBATCH --cpus-per-task=6
+#SBATCH --wait-all-nodes=1
+#SBATCH --job-name=open_clip
+#SBATCH --account=ACCOUNT_NAME
+#SBATCH --partition PARTITION_NAME
+eval "$(/path/to/conda/bin/conda shell.bash hook)" # init conda
+conda activate open_clip
+export CUDA_VISIBLE_DEVICES=0,1,2,3
+export MASTER_PORT=12802
+master_addr=$(scontrol show hostnames "$SLURM_JOB_NODELIST" | head -n 1)
+export MASTER_ADDR=$master_addr
+cd /shared/open_clip
+export PYTHONPATH="$PYTHONPATH:$PWD/src"
+srun --cpu_bind=v --accel-bind=gn python -u src/open_clip_train/main.py \
+    --save-frequency 1 \
+    --report-to tensorboard \
+    --train-data="/data/LAION-400M/{00000..41455}.tar" \
+    --warmup 2000 \
+    --batch-size=256 \
+    --epochs=32 \
+    --workers=8 \
+    --model ViT-B-32 \
+    --name "ViT-B-32-Vanilla" \
+    --seed 0 \
+    --local-loss \
+    --gather-with-grad
+```
+### Resuming from a checkpoint:
+```bash
+python -m open_clip_train.main \
+    --train-data="/path/to/train_data.csv" \
+    --val-data="/path/to/validation_data.csv"  \
+    --resume /path/to/checkpoints/epoch_K.pt
+```
+### Training CoCa:
+Training [CoCa](https://arxiv.org/abs/2205.01917) models is enabled through specifying a CoCa config using the ```--model``` parameter of the training script. Currently available configs are "coca_base", "coca_ViT-B-32", and "coca_roberta-ViT-B-32" (which uses RoBERTa as the text encoder). CoCa configs are different from CLIP configs because they have an additional "multimodal_cfg" component which specifies parameters for the multimodal text decoder. Here's an example from the coca_ViT-B-32 config:
+```json
+"multimodal_cfg": {
+	"context_length": 76,
+	"vocab_size": 49408,
+	"width": 512,
+	"heads": 8,
+	"layers": 12,
+	"latent_dim": 512,
+	"attn_pooler_heads": 8
+}
+```
+Credit to [lucidrains](https://github.com/lucidrains) for [initial code](https://github.com/lucidrains/CoCa-pytorch), [gpucce](https://github.com/gpucce) for adapting the code to open_clip, and [iejMac](https://github.com/iejMac) for training the models.
+### Generating text with CoCa
+```python
+import open_clip
+import torch
+from PIL import Image
+model, _, transform = open_clip.create_model_and_transforms(
+  model_name="coca_ViT-L-14",
+  pretrained="mscoco_finetuned_laion2B-s13B-b90k"
+)
+im = Image.open("cat.jpg").convert("RGB")
+im = transform(im).unsqueeze(0)
+with torch.no_grad(), torch.cuda.amp.autocast():
+  generated = model.generate(im)
+print(open_clip.decode(generated[0]).split("<end_of_text>")[0].replace("<start_of_text>", ""))
+```
+See also this [[Coca Colab]](https://colab.research.google.com/github/mlfoundations/open_clip/blob/master/docs/Interacting_with_open_coca.ipynb)
+### Fine Tuning CoCa
+To fine-tune coca on mscoco, first create the dataset, one way is using a csvdataset and perhaps the simplest way to do it is using [CLIP_benchmark](https://github.com/LAION-AI/CLIP_benchmark) which in turn uses [pycocotools](https://github.com/cocodataset/cocoapi) (that can be used also by itself).
+```python
+from clip_benchmark.datasets.builder import build_dataset
+import pandas as pd
+import os
+root_path = "path/to/data/dir" # set this to smth meaningful
+ds = build_dataset("mscoco_captions", root=root_path, split="train", task="captioning") # this downloads the dataset if it is not there already
+coco = ds.coco
+imgs = coco.loadImgs(coco.getImgIds())
+future_df = {"filepath":[], "title":[]}
+for img in imgs:
+    caps = coco.imgToAnns[img["id"]]
+    for cap in caps:
+        future_df["filepath"].append(img["file_name"])
+        future_df["title"].append(cap["caption"])
+pd.DataFrame.from_dict(future_df).to_csv(
+  os.path.join(root_path, "train2014.csv"), index=False, sep="\t"
+)
+```
+This should create a csv dataset that one can use to fine-tune coca with open_clip
+```bash
+python -m open_clip_train.main \
+    --dataset-type "csv" \
+    --train-data "path/to/data/dir/train2014.csv" \
+    --warmup 1000 \
+    --batch-size 128 \
+    --lr 1e-5 \
+    --wd 0.1 \
+    --epochs 1 \
+    --workers 3 \
+    --model "coca_ViT-L-14" \
+    --report-to "wandb" \
+    --coca-contrastive-loss-weight 0 \
+    --coca-caption-loss-weight 1 \
+    --log-every-n-steps 100
+```
+This is a general setting, open_clip has very parameters that can be set, ```python -m open_clip_train.main --help``` should show them. The only relevant change compared to pre-training are the two arguments
+```bash
+--coca-contrastive-loss-weight 0
+--coca-caption-loss-weight 1
+```
+which make the model only train the generative side.
+### Training with pre-trained language models as text encoder:
+If you wish to use different language models as the text encoder for CLIP you can do so by using one of the Hugging Face model configs in ```src/open_clip/model_configs``` and passing in it's tokenizer as the ```--model``` and ```--hf-tokenizer-name``` parameters respectively. Currently we only support RoBERTa ("test-roberta" config), however adding new models should be trivial. You can also determine how many layers, from the end, to leave unfrozen with the ```--lock-text-unlocked-layers``` parameter. Here's an example command to train CLIP with the RoBERTa LM that has it's last 10 layers unfrozen:
+```bash
+python -m open_clip_train.main \
+         --train-data="pipe:aws s3 cp s3://s-mas/cc3m/{00000..00329}.tar -" \
+         --train-num-samples 3000000 \
+         --val-data="pipe:aws s3 cp s3://s-mas/cc3m/{00330..00331}.tar -" \
+         --val-num-samples 10000 \
+         --dataset-type webdataset \
+         --batch-size 256 \
+         --warmup 2000 \
+         --epochs 10 \
+         --lr 5e-4 \
+         --precision amp \
+         --workers 6 \
+         --model "roberta-ViT-B-32" \
+         --lock-text \
+         --lock-text-unlocked-layers 10 \
+         --name "10_unfrozen" \
+         --report-to "tensorboard" \
+```
+### Loss Curves
+When run on a machine with 8 GPUs the command should produce the following training curve for Conceptual Captions:
+![CLIP zero shot training curve](https://raw.githubusercontent.com/mlfoundations/open_clip/main/docs/clip_zeroshot.png)
+More detailed curves for Conceptual Captions are given at [/docs/clip_conceptual_captions.md](/docs/clip_conceptual_captions.md).
+When training a RN50 on YFCC the same hyperparameters as above are used, with the exception of `lr=5e-4` and `epochs=32`.
+Note that to use another model, like `ViT-B/32` or `RN50x4` or `RN50x16` or `ViT-B/16`, specify with `--model RN50x4`.
+### Logging
+For tensorboard logging, run:
+```bash
+tensorboard --logdir=logs/tensorboard/ --port=7777
+```
+For wandb logging, we recommend looking at the `step` variable instead of `Step`, since the later was not properly set in earlier versions of this codebase.
+For older runs with models trained before https://github.com/mlfoundations/open_clip/pull/613, the `Step` variable should be ignored.
+For newer runs, after that PR, the two variables are the same.
+## Evaluation / Zero-Shot
+We recommend https://github.com/LAION-AI/CLIP_benchmark#how-to-use for systematic evaluation on 40 datasets.
+### Evaluating local checkpoint:
+```bash
+python -m open_clip_train.main \
+    --val-data="/path/to/validation_data.csv"  \
+    --model RN101 \
+    --pretrained /path/to/checkpoints/epoch_K.pt
+```
+### Evaluating hosted pretrained checkpoint on ImageNet zero-shot prediction:
+```bash
+python -m open_clip_train.main \
+    --imagenet-val /path/to/imagenet/validation \
+    --model ViT-B-32-quickgelu \
+    --pretrained laion400m_e32
+```
+### Model distillation
+You can distill from a pre-trained by using `--distill-model` and `--distill-pretrained` to specify the model you'd like to distill from.
+For instance, to distill from OpenAI ViT-L/14 use `--distill-model ViT-L-14 --distill-pretrained openai`.
+### Gradient accumulation
+To simulate larger batches use `--accum-freq k`. If per gpu batch size, `--batch-size`, is `m`, then the effective batch size will be `k * m * num_gpus`.
+When increasing `--accum-freq` from its default of 1, samples/s will remain approximately constant (batch size will double, as will time-per-batch). It is recommended to use other features to reduce batch size such as `--grad-checkpointing --local-loss --gather-with-grad` before increasing `--accum-freq`. `--accum-freq` can be used in addition to these features.
+Instead of 1 forward pass per example, there are now 2 forward passes per-example. However, the first is done with `torch.no_grad`.
+There is some additional GPU memory required --- the features and data from all `m` batches are stored in memory.
+There are also `m` loss computations instead of the usual 1.
+For more information see Cui et al. (https://arxiv.org/abs/2112.09331) or Pham et al. (https://arxiv.org/abs/2111.10050).
+### Int8 Support
+We have beta support for int8 training and inference.
+You can enable int8 training with `--use-bnb-linear SwitchBackLinearGlobal` or `--use-bnb-linear SwitchBackLinearGlobalMemEfficient`.
+Please see the bitsandbytes library for definitions for these layers.
+For CLIP VIT-Huge this should currently correspond to a 10% training speedup with no accuracy loss.
+More speedups comin when the attention layer is refactored so that linear layers man be replaced there, too.
+See the tutorial https://github.com/mlfoundations/open_clip/blob/main/tutorials/int8_tutorial.ipynb or [paper](https://arxiv.org/abs/2304.13013).
+### Support for remote loading/training
+It is always possible to resume directly from a remote file, e.g., a file in an s3 bucket. Just set `--resume s3://<path-to-checkpoint> `.
+This will work with any filesystem supported by `fsspec`.
+It is also possible to train `open_clip` models while continuously backing up to s3. This can help to avoid slow local file systems.
+Say that your node has a local ssd `/scratch`, an s3 bucket `s3://<path-to-bucket>`.
+In that case, set `--logs /scratch` and `--remote-sync s3://<path-to-bucket>`. Then, a background process will sync `/scratch/<run-name>` to `s3://<path-to-bucket>/<run-name>`. After syncing, the background process will sleep for `--remote-sync-frequency` seconds, which defaults to 5 minutes.
+There is also experimental support for syncing to other remote file systems, not just s3. To do so, specify `--remote-sync-protocol fsspec`. However, this is currently very slow and not recommended.
+Also, to optionally avoid saving too many checkpoints locally when using these features, you can use `--delete-previous-checkpoint` which deletes the previous checkpoint after saving a new one.
+Note: if you are using this feature with `--resume latest`, there are a few warnings. First, use with `--save-most-recent` is not supported. Second, only `s3` is supported. Finally, since the sync happens in the background, it is possible that the most recent checkpoint may not be finished syncing to the remote.
+### Pushing Models to Hugging Face Hub
+The module `open_clip.push_to_hf_hub` includes helpers for pushing models /w weights and config to the HF Hub.
+The tool can be run from command line, ex:
+`python -m open_clip.push_to_hf_hub --model convnext_large_d_320 --pretrained /train/checkpoints/epoch_12.pt --repo-id laion/CLIP-convnext_large_d_320.laion2B-s29B-b131K-ft`
+## Acknowledgments
+We gratefully acknowledge the Gauss Centre for Supercomputing e.V. (www.gauss-centre.eu) for funding this part of work by providing computing time through the John von Neumann Institute for Computing (NIC) on the GCS Supercomputer JUWELS Booster at Jülich Supercomputing Centre (JSC).
+## The Team
+Current development of this repository is led by [Ross Wightman](https://rwightman.com/), [Romain Beaumont](https://github.com/rom1504), [Cade Gordon](http://cadegordon.io/), and [Vaishaal Shankar](http://vaishaal.com/).
+The original version of this repository is from a group of researchers at UW, Google, Stanford, Amazon, Columbia, and Berkeley.
+[Gabriel Ilharco*](http://gabrielilharco.com/), [Mitchell Wortsman*](https://mitchellnw.github.io/), [Nicholas Carlini](https://nicholas.carlini.com/), [Rohan Taori](https://www.rohantaori.com/), [Achal Dave](http://www.achaldave.com/), [Vaishaal Shankar](http://vaishaal.com/), [John Miller](https://people.eecs.berkeley.edu/~miller_john/), [Hongseok Namkoong](https://hsnamkoong.github.io/), [Hannaneh Hajishirzi](https://homes.cs.washington.edu/~hannaneh/), [Ali Farhadi](https://homes.cs.washington.edu/~ali/), [Ludwig Schmidt](https://people.csail.mit.edu/ludwigs/)
+Special thanks to [Jong Wook Kim](https://jongwook.kim/) and [Alec Radford](https://github.com/Newmu) for help with reproducing CLIP!
+## Citing
+If you found this repository useful, please consider citing:
+```bibtex
+@software{ilharco_gabriel_2021_5143773,
+  author       = {Ilharco, Gabriel and
+                  Wortsman, Mitchell and
+                  Wightman, Ross and
+                  Gordon, Cade and
+                  Carlini, Nicholas and
+                  Taori, Rohan and
+                  Dave, Achal and
+                  Shankar, Vaishaal and
+                  Namkoong, Hongseok and
+                  Miller, John and
+                  Hajishirzi, Hannaneh and
+                  Farhadi, Ali and
+                  Schmidt, Ludwig},
+  title        = {OpenCLIP},
+  month        = jul,
+  year         = 2021,
+  note         = {If you use this software, please cite it as below.},
+  publisher    = {Zenodo},
+  version      = {0.1},
+  doi          = {10.5281/zenodo.5143773},
+  url          = {https://doi.org/10.5281/zenodo.5143773}
+}
+```
+```bibtex
+@inproceedings{cherti2023reproducible,
+  title={Reproducible scaling laws for contrastive language-image learning},
+  author={Cherti, Mehdi and Beaumont, Romain and Wightman, Ross and Wortsman, Mitchell and Ilharco, Gabriel and Gordon, Cade and Schuhmann, Christoph and Schmidt, Ludwig and Jitsev, Jenia},
+  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
+  pages={2818--2829},
+  year={2023}
+}
+```
+```bibtex
+@inproceedings{Radford2021LearningTV,
+  title={Learning Transferable Visual Models From Natural Language Supervision},
+  author={Alec Radford and Jong Wook Kim and Chris Hallacy and A. Ramesh and Gabriel Goh and Sandhini Agarwal and Girish Sastry and Amanda Askell and Pamela Mishkin and Jack Clark and Gretchen Krueger and Ilya Sutskever},
+  booktitle={ICML},
+  year={2021}
+}
+```
+```bibtex
+@inproceedings{schuhmann2022laionb,
+  title={{LAION}-5B: An open large-scale dataset for training next generation image-text models},
+  author={Christoph Schuhmann and
+          Romain Beaumont and
+          Richard Vencu and
+          Cade W Gordon and
+          Ross Wightman and
+          Mehdi Cherti and
+          Theo Coombes and
+          Aarush Katta and
+          Clayton Mullis and
+          Mitchell Wortsman and
+          Patrick Schramowski and
+          Srivatsa R Kundurthy and
+          Katherine Crowson and
+          Ludwig Schmidt and
+          Robert Kaczmarczyk and
+          Jenia Jitsev},
+  booktitle={Thirty-sixth Conference on Neural Information Processing Systems Datasets and Benchmarks Track},
+  year={2022},
+  url={https://openreview.net/forum?id=M3Y74vmsMcY}
+}
+```
+[![DOI](https://zenodo.org/badge/390536799.svg)](https://zenodo.org/badge/latestdoi/390536799)
--- a/custom_tests/coca_simple_test.py
+++ b/custom_tests/coca_simple_test.py
+import open_clip
+import torch
+from PIL import Image
+if __name__ == "__main__":
+    from argparse import ArgumentParser
+    parser = ArgumentParser()
+    parser.add_argument("--model_name", type=str, default="coca_ViT-L-14")
+    parser.add_argument("--pretrained", type=str, default="mscoco_finetuned_laion2B-s13B-b90k")
+    parser.add_argument("--image_path", type=str)
+    args = parser.parse_args()
+    model, _, transform = open_clip.create_model_and_transforms(
+        model_name=args.model_name,
+        pretrained=args.pretrained
+    )
+    model.cuda()
+    im = Image.open(args.image_path).convert("RGB")
+    im = transform(im).unsqueeze(0).cuda()
+    with torch.no_grad(), torch.cuda.amp.autocast():
+        generated = model.generate(im)
+    print(open_clip.decode(generated[0]).split("<end_of_text>")[0].replace("<start_of_text>", ""))
\ No newline at end of file
--- a/custom_tests/simple_test.py
+++ b/custom_tests/simple_test.py
+import torch
+from PIL import Image
+import open_clip
+if __name__ == "__main__":
+    import argparse
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--model_name", default="ViT-B-32", type=str)
+    parser.add_argument("--pretrained", default="laion2b_s34b_b79k", type=str)
+    args = parser.parse_args()
+    model, _, preprocess = open_clip.create_model_and_transforms(args.model_name, pretrained=args.pretrained)
+    model.eval()  # model in train mode by default, impacts some models with BatchNorm or stochastic depth active
+    tokenizer = open_clip.get_tokenizer(args.model_name)
+    model.cuda()
+    image = preprocess(Image.open("docs/CLIP.png")).unsqueeze(0)
+    text = tokenizer(["a diagram", "a dog", "a cat"])
+    with torch.no_grad(), torch.cuda.amp.autocast():
+        image_features = model.encode_image(image)
+        text_features = model.encode_text(text)
+        image_features /= image_features.norm(dim=-1, keepdim=True)
+        text_features /= text_features.norm(dim=-1, keepdim=True)
+        text_probs = (100.0 * image_features @ text_features.T).softmax(dim=-1)
+    print("Label probs:", text_probs)  # prints: [[1., 0., 0.]]
--- a/docs/CLIP.png
+++ b/docs/CLIP.png
--- a/docs/Interacting_with_open_clip.ipynb
+++ b/docs/Interacting_with_open_clip.ipynb
--- a/docs/Interacting_with_open_coca.ipynb
+++ b/docs/Interacting_with_open_coca.ipynb
+{
+  "nbformat": 4,
+  "nbformat_minor": 0,
+  "metadata": {
+    "colab": {
+      "provenance": []
+    },
+    "kernelspec": {
+      "name": "python3",
+      "display_name": "Python 3"
+    },
+    "language_info": {
+      "name": "python"
+    },
+    "accelerator": "GPU",
+    "gpuClass": "standard"
+  },
+  "cells": [
+    {
+      "cell_type": "code",
+      "source": [
+        "!pip install open_clip_torch transformers"
+      ],
+      "metadata": {
+        "id": "JvaEkx8Cyvhg"
+      },
+      "execution_count": null,
+      "outputs": []
+    },
+    {
+      "cell_type": "code",
+      "execution_count": 18,
+      "metadata": {
+        "id": "vE4lFFkKyotX"
+      },
+      "outputs": [],
+      "source": [
+        "import open_clip\n",
+        "import torch\n",
+        "\n",
+        "model, _, transform = open_clip.create_model_and_transforms(\n",
+        "  model_name=\"coca_ViT-L-14\",\n",
+        "  pretrained=\"mscoco_finetuned_laion2B-s13B-b90k\"\n",
+        ")"
+      ]
+    },
+    {
+      "cell_type": "code",
+      "source": [
+        "!wget https://i.imgur.com/8H7XCH0.jpg -O cat.jpg"
+      ],
+      "metadata": {
+        "id": "oOaE1AmDyth_"
+      },
+      "execution_count": null,
+      "outputs": []
+    },
+    {
+      "cell_type": "code",
+      "source": [
+        "from IPython.display import Image\n",
+        "Image('cat.jpg')"
+      ],
+      "metadata": {
+        "colab": {
+          "base_uri": "https://localhost:8080/",
+          "height": 407
+        },
+        "id": "Y9Q6bhVA2L01",
+        "outputId": "1b920080-e8cd-4d2f-fb23-30e6f1b612f9"
+      },
+      "execution_count": 19,
+      "outputs": [
+        {
+          "output_type": "execute_result",
+          "data": {
+            "image/jpeg": "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\n",
+            "text/plain": [
+              "<IPython.core.display.Image object>"
+            ]
+          },
+          "metadata": {},
+          "execution_count": 19
+        }
+      ]
+    },
+    {
+      "cell_type": "code",
+      "source": [
+        "from PIL import Image\n",
+        "im = Image.open(\"cat.jpg\").convert(\"RGB\")\n",
+        "im = transform(im).unsqueeze(0)\n",
+        "\n",
+        "with torch.no_grad(), torch.cuda.amp.autocast():\n",
+        "  generated = model.generate(im)\n",
+        "\n",
+        "print(open_clip.decode(generated[0]).split(\"<end_of_text>\")[0].replace(\"<start_of_text>\", \"\"))"
+      ],
+      "metadata": {
+        "colab": {
+          "base_uri": "https://localhost:8080/"
+        },
+        "id": "byZKXMGzyr5Y",
+        "outputId": "122eb099-6704-4e3c-fa7c-a05dd87ce64f"
+      },
+      "execution_count": 22,
+      "outputs": [
+        {
+          "output_type": "stream",
+          "name": "stdout",
+          "text": [
+            "an orange and white cat on top of a turtle . \n"
+          ]
+        }
+      ]
+    }
+  ]
+}
\ No newline at end of file
--- a/docs/LOW_ACC.md
+++ b/docs/LOW_ACC.md
+As we describe in more detail below, CLIP models in a medium accuracy regime already allow us to draw conclusions about the robustness of larger CLIP models since the models follow reliable scaling laws.
+[Cherti et al., 2022](https://arxiv.org/abs/2212.07143) and [Gadre et al., 2023](https://arxiv.org/abs/2304.14108) show additional discussions about the scaling behavior of CLIP models.
+## Scaling trends
+The plot below shows how zero-shot performance of CLIP models varies as we scale the number of samples used for training. Zero-shot performance increases steadily for both ImageNet and [ImageNetV2](https://arxiv.org/abs/1902.10811), and is far from saturated at ~15M samples.
+<img src="https://raw.githubusercontent.com/mlfoundations/open_clip/main/docs/scaling.png" width="700">
+## Why are low-accuracy CLIP models interesting?
+**TL;DR:** CLIP models have high effective robustness, even at small scales.
+CLIP models are particularly intriguing because they are more robust to natural distribution shifts (see Section 3.3 in the [CLIP paper](https://arxiv.org/abs/2103.00020)).
+This phenomena is illustrated by the figure below, with ImageNet accuracy on the x-axis
+and [ImageNetV2](https://arxiv.org/abs/1902.10811) (a reproduction of the ImageNet validation set with distribution shift) accuracy on the y-axis.
+Standard training denotes training on the ImageNet train set and the CLIP zero-shot models
+are shown as stars.
+![CLIP scatter plot](https://raw.githubusercontent.com/mlfoundations/open_clip/main/docs/effective_robustness.png)
+As observed by [Taori et al., 2020](https://arxiv.org/abs/2007.00644) and [Miller et al., 2021](https://arxiv.org/abs/2107.04649), the in-distribution
+and out-of-distribution accuracies of models trained on ImageNet follow a predictable linear trend (the red line in the above plot). *Effective robustness*
+quantifies robustness as accuracy beyond this baseline, i.e., how far a model lies above the red line. Ideally a model would not suffer from distribution shift and fall on the y = x line ([trained human labelers are within a percentage point of the y = x line](http://proceedings.mlr.press/v119/shankar20c.html)).
+Even though the CLIP models trained with
+this codebase achieve much lower accuracy than those trained by OpenAI, our models still lie on the same
+trend of improved effective robustness (the purple line). Therefore, we can study what makes
+CLIP robust without requiring industrial-scale compute.
+For more information on effective robustness, please see:
+- [Recht et al., 2019](https://arxiv.org/abs/1902.10811).
+- [Taori et al., 2020](https://arxiv.org/abs/2007.00644).
+- [Miller et al., 2021](https://arxiv.org/abs/2107.04649).
+To know more about the factors that contribute to CLIP's robustness refer to [Fang et al., 2022](https://arxiv.org/abs/2205.01397).
\ No newline at end of file
--- a/docs/PRETRAINED.md
+++ b/docs/PRETRAINED.md
+## Pretrained model results
+We evaluate the full collection of available models on a suite of 38 datasets in a zero-shot setting (i.e., without fine-tuning), following [Gadre et al., 2023](https://arxiv.org/abs/2304.14108).
+Click below to see the full results.
+- [Full results (English)](openclip_results.csv)
+- [Classification-only results](openclip_classification_results.csv)
+- [Retrieval results](openclip_retrieval_results.csv)
+- [Multilingual retrieval results](openclip_multilingual_retrieval_results.csv)
+## Pretrained model details
+Below are details for several of our pretrained models.
+### LAION-400M - https://laion.ai/laion-400-open-dataset
+We ran experiments in an attempt to reproduce OpenAI's ViT results with the comparably sized (and open) LAION-400M dataset. Trained
+weights can be found in release [v0.2](https://github.com/mlfoundations/open_clip/releases/tag/v0.2-weights).
+The LAION400M weights have been trained on the JUWELS supercomputer (see acknowledgements section below).
+#### ViT-B/32 224x224
+We replicate OpenAI's results on ViT-B/32, reaching a top-1 ImageNet-1k zero-shot accuracy of 62.96%.
+<img src="https://raw.githubusercontent.com/mlfoundations/open_clip/main/docs/laion_clip_zeroshot.png" width="700">
+**Zero-shot comparison (courtesy of Andreas Fürst)**
+<img src="https://raw.githubusercontent.com/mlfoundations/open_clip/main/docs/laion_openai_compare_b32.jpg" width="700">
+ViT-B/32 was trained with 128 A100 (40 GB) GPUs for ~36 hours, 4600 GPU-hours. The per-GPU batch size was 256 for a global batch size of 32768. 256 is much lower than it could have been (~320-384) due to being sized initially before moving to 'local' contrastive loss.
+#### ViT-B/16 224x224
+The B/16 LAION400M training reached a top-1 ImageNet-1k zero-shot validation score of 67.07.
+<img src="https://raw.githubusercontent.com/mlfoundations/open_clip/main/docs/laion_clip_zeroshot_b16.png" width="700">
+This was the first major train session using the updated webdataset 0.2.x code. A bug was found that prevented shards from being shuffled properly between nodes/workers each epoch. This was fixed part way through training (epoch 26) but likely had an impact.
+ViT-B/16 was trained with 176 A100 (40 GB) GPUS for ~61 hours, 10700 GPU-hours. Batch size per GPU was 192 for a global batch size of 33792.
+#### ViT-B/16+ 240x240
+The B/16+ 240x240 LAION400M training reached a top-1 ImageNet-1k zero-shot validation score of 69.21.
+This model is the same depth as the B/16, but increases the
+- vision width from 768 -> 896
+- text width from 512 -> 640
+- the resolution 224x224 -> 240x240 (196 -> 225 tokens)
+<img src="https://raw.githubusercontent.com/mlfoundations/open_clip/main/docs/laion_clip_zeroshot_b16_plus_240.png" width="700">
+Unlike the B/16 run above, this model was a clean run with no dataset shuffling issues.
+ViT-B/16+ was trained with 224 A100 (40 GB) GPUS for ~61 hours, 13620 GPU-hours. Batch size per GPU was 160 for a global batch size of 35840.
+#### ViT-L/14 224x224
+The L/14 LAION-400M training reached a top-1 ImageNet-1k zero-shot validation score of 72.77.
+<img src="https://raw.githubusercontent.com/mlfoundations/open_clip/main/docs/laion_clip_zeroshot_l14.png" width="700">
+ViT-L/14 was trained with 400 A100 (40 GB) GPUS for ~127 hours, 50800 GPU-hours. Batch size per GPU was 96 for a global batch size of 38400. Grad checkpointing was enabled.
+### LAION-2B (en) - https://laion.ai/laion-5b-a-new-era-of-open-large-scale-multi-modal-datasets/
+A ~2B sample subset of LAION-5B with english captions (https://huggingface.co/datasets/laion/laion2B-en)
+#### ViT-B/32 224x224
+A ViT-B/32 trained on LAION-2B, reaching a top-1 ImageNet-1k zero-shot accuracy of 65.62%.
+<img src="https://raw.githubusercontent.com/mlfoundations/open_clip/main/docs/laion2b_clip_zeroshot_b32.png" width="700">
+ViT-B/32 was trained with 112 A100 (40 GB) GPUs. The per-GPU batch size was 416 for a global batch size of 46592. Compute generously provided by [stability.ai](https://stability.ai/).
+A second iteration of B/32 was trained on stability.ai cluster with a larger global batch size and learning rate, hitting 66.6% top-1. See https://huggingface.co/laion/CLIP-ViT-B-32-laion2B-s34B-b79K
+#### ViT-L/14 224x224
+A ViT-L/14 with a 75.3% top-1 ImageNet-1k zero-shot was trained on JUWELS Booster. See model details here https://huggingface.co/laion/CLIP-ViT-L-14-laion2B-s32B-b82K
+These weights use a different dataset mean and std than others. Instead of using the OpenAI mean & std, inception style normalization `[-1, 1]` is used via a mean and std of `[0.5, 0.5, 0.5]`. This is handled automatically if using `open_clip.create_model_and_transforms` from pretrained weights.
+#### ViT-H/14 224x224
+A ViT-H/14 with a 78.0% top-1 ImageNet-1k zero-shot was trained on JUWELS Booster. See model details here https://huggingface.co/laion/CLIP-ViT-H-14-laion2B-s32B-b79K
+#### ViT-g/14 224x224
+A ViT-g/14 with a 76.6% top-1 ImageNet-1k zero-shot was trained on JUWELS Booster. See model details here https://huggingface.co/laion/CLIP-ViT-g-14-laion2B-s12B-b42K
+This model was trained with a shorted schedule than other LAION-2B models with 12B samples seen instead of 32+B. It matches LAION-400M training in samples seen. Many zero-shot results are lower as a result, but despite this it performs very well in some OOD zero-shot and retrieval tasks.
+#### ViT-B/32 roberta base
+A ViT-B/32 with roberta base encoder with a 61.7% top-1 ImageNet-1k zero-shot was trained on stability. See model details here https://huggingface.co/laion/CLIP-ViT-B-32-roberta-base-laion2B-s12B-b32k
+This is the first openclip model using a HF text tower. It has better performance on a range of tasks compared to the standard text encoder, see [metrics](https://huggingface.co/laion/CLIP-ViT-B-32-roberta-base-laion2B-s12B-b32k/blob/main/unknown.png)
+#### ViT-B/32 xlm roberta base
+A ViT-B/32 with xlm roberta base encoder with a 62.33% top-1 ImageNet-1k zero-shot was trained on stability. See model details here https://huggingface.co/laion/CLIP-ViT-B-32-xlm-roberta-base-laion5B-s13B-b90k
+This is the first openclip model trained on the full laion5B dataset; hence the first multilingual clip trained with openclip. It has better performance on a range of tasks compared to the standard text encoder, see [metrics](https://huggingface.co/laion/CLIP-ViT-B-32-xlm-roberta-base-laion5B-s13B-b90k/blob/main/metrics.png)
+A preliminary multilingual evaluation was run: 43% on imagenet1k italian (vs 21% for english B/32), 37% for imagenet1k japanese (vs 1% for english B/32 and 50% for B/16 clip japanese). It shows the multilingual property is indeed there as expected. Larger models will get even better performance.
+#### ViT-H/14 xlm roberta large
+A ViT-H/14 with xlm roberta large encoder with a 77.0% (vs 78% for the english equivalent) top-1 ImageNet-1k zero-shot was trained on stability. See model details here https://huggingface.co/laion/CLIP-ViT-H-14-frozen-xlm-roberta-large-laion5B-s13B-b90k
+This model was trained following the [LiT](https://arxiv.org/abs/2111.07991) methodology: the image tower was frozen (initialized from english openclip ViT-H/14), the text tower was initialized from [xlm roberta large](https://huggingface.co/xlm-roberta-large) and unfrozen. This reduced training cost by a 3x factor.
+See full english [metrics](https://huggingface.co/laion/CLIP-ViT-H-14-frozen-xlm-roberta-large-laion5B-s13B-b90k/resolve/main/results_xlm_roberta_large.png)
+On zero shot classification on imagenet with translated prompts this model reaches:
+- 56% in italian (vs 21% for https://github.com/clip-italian/clip-italian)
+- 53% in japanese (vs 54.6% for https://github.com/rinnakk/japanese-clip)
+- 55.7% in chinese (to be compared with https://github.com/OFA-Sys/Chinese-CLIP)
+#### YFCC-15M
+Below are checkpoints of models trained on YFCC-15M, along with their zero-shot top-1 accuracies on ImageNet and ImageNetV2. These models were trained using 8 GPUs and the same hyperparameters described in the "Sample running code" section, with the exception of `lr=5e-4` and `epochs=32`.
+- [ResNet-50](https://github.com/mlfoundations/open_clip/releases/download/v0.2-weights/rn50-quickgelu-yfcc15m-455df137.pt) (32.7% / 27.9%)
+- [ResNet-101](https://github.com/mlfoundations/open_clip/releases/download/v0.2-weights/rn101-quickgelu-yfcc15m-3e04b30e.pt) (34.8% / 30.0%)
+#### CC12M - https://github.com/google-research-datasets/conceptual-12m
+- [ResNet-50](https://github.com/mlfoundations/open_clip/releases/download/v0.2-weights/rn50-quickgelu-cc12m-f000538c.pt) (36.45%)
+### CommonPool and DataComp models
+As part of [DataComp](https://github.com/mlfoundations/datacomp), we trained models on CommonPool using various data filtering strategies.
+The best performing models are specified below for the xlarge scale, see our paper [DataComp: In seearch of the next generation of multimodal datasets](https://arxiv.org/abs/2304.14108) for more details.
+Additional models and more information can be found at [/docs/datacomp_models.md](/docs/datacomp_models.md).
+- `datacomp_xl_s13b_b90k`: A ViT-L/14 trained on DataComp-1B for 12.8B steps and batch size 90k. Achieves 79.2% zero-shot accuracy on ImageNet. Available at https://huggingface.co/laion/CLIP-ViT-L-14-DataComp.XL-s13B-b90K.
+- `commonpool_xl_clip_s13b_b90k`: A ViT-L/14 trained on CommonPool-XL filtered using CLIP scores, for 12.8B steps and batch size 90k. Achieves 76.4% zero-shot accuracy on ImageNet. Available at https://huggingface.co/laion/CLIP-ViT-L-14-CommonPool.XL.clip-s13B-b90K.
+- `commonpool_xl_laion_s13b_b90k`: A ViT-L/14 trained on CommonPool-XL filtered using the LAION-2B filtering scheme, for 12.8B steps and batch size 90k. Achieves 75.5% zero-shot accuracy on ImageNet. Available at https://huggingface.co/laion/CLIP-ViT-L-14-CommonPool.XL.laion-s13B-b90K.
+- `commonpool_xl_s13b_b90k`: A ViT-L/14 trained on CommonPool-XL without any filtering, for 12.8B steps and batch size 90k. Achieves 72.3% zero-shot accuracy on ImageNet. Available at https://huggingface.co/laion/CLIP-ViT-L-14-CommonPool.XL-s13B-b90K.
+If you use models trained on DataComp-1B or CommonPool variations, please consider citing the following:
+```bibtex
+@article{datacomp,
+  title={DataComp: In search of the next generation of multimodal datasets},
+  author={Samir Yitzhak Gadre, Gabriel Ilharco, Alex Fang, Jonathan Hayase, Georgios Smyrnis, Thao Nguyen, Ryan Marten, Mitchell Wortsman, Dhruba Ghosh, Jieyu Zhang, Eyal Orgad, Rahim Entezari, Giannis Daras, Sarah Pratt, Vivek Ramanujan, Yonatan Bitton, Kalyani Marathe, Stephen Mussmann, Richard Vencu, Mehdi Cherti, Ranjay Krishna, Pang Wei Koh, Olga Saukh, Alexander Ratner, Shuran Song, Hannaneh Hajishirzi, Ali Farhadi, Romain Beaumont, Sewoong Oh, Alex Dimakis, Jenia Jitsev, Yair Carmon, Vaishaal Shankar, Ludwig Schmidt},
+  journal={arXiv preprint arXiv:2304.14108},
+  year={2023}
+}
+```
+### MetaCLIP
+MetaCLIP models are described in the paper [Demystifying CLIP Data](https://arxiv.org/abs/2309.16671).
+These models were developed by Hu Xu, Saining Xie, Xiaoqing Ellen Tan, Po-Yao Huang, Russell Howes, Vasu Sharma, Shang-Wen Li, Gargi Ghosh, Luke Zettlemoyer and Christoph Feichtenhofer from Meta, New York University and the University of Washington.
+Models are licensed under CC-BY-NC.
+More details are available at https://github.com/facebookresearch/MetaCLIP.
+If you use MetaCLIP models, please cite the following:
+```bibtex
+@inproceedings{xu2023metaclip,
+   title={Demystifying CLIP Data},
+   author={Hu Xu, Saining Xie, Xiaoqing Ellen Tan, Po-Yao Huang, Russell Howes, Vasu, Sharma, Shang-Wen Li, Gargi Ghosh, Luke Zettlemoyer and Christoph Feichtenhofer},
+   journal={arXiv preprint arXiv:2309.16671},
+   year={2023}
+}
+```
+### EVA-CLIP
+EVA-CLIP models are described in the paper [EVA-CLIP: Improved Training Techniques for CLIP at Scale](https://arxiv.org/abs/2303.15389).
+These models were developed by Quan Sun, Yuxin Fang, Ledell Wu, Xinlong Wang and Yue Cao from BAAI and HUST.
+Models are licensed under the MIT License.
+More details are available at https://github.com/baaivision/EVA/tree/master/EVA-CLIP.
+If you use EVA models, please cite the following:
+```bibtex
+@article{EVA-CLIP,
+  title={EVA-CLIP: Improved Training Techniques for CLIP at Scale},
+  author={Sun, Quan and Fang, Yuxin and Wu, Ledell and Wang, Xinlong and Cao, Yue},
+  journal={arXiv preprint arXiv:2303.15389},
+  year={2023}
+}
+```
+### NLLB-CLIP
+NLLB-CLIP models are described in the paper [NLLB-CLIP - train performant multilingual image retrieval model on a budget](https://arxiv.org/abs/2309.01859) by Alexander Visheratin.
+The model was trained following the [LiT](https://arxiv.org/abs/2111.07991) methodology: the image tower was frozen, the text tower was initialized from the [NLLB](https://arxiv.org/abs/2207.04672) encoder and unfrozen.
+The model was trained on the [LAION-COCO-NLLB](https://huggingface.co/datasets/visheratin/laion-coco-nllb) dataset.
+The first version of the model (`nllb-clip`) described in the paper was trained using the OpenAI CLIP image encoder.
+The second version of the model (`nllb-clip-siglip`) was trained using the [SigLIP](https://arxiv.org/abs/2303.15343) image encoder.
+Models are licensed under CC-BY-NC.
+If you use NLLB-CLIP models, please cite the following:
+```bibtex
+@article{visheratin2023nllb,
+  title={NLLB-CLIP--train performant multilingual image retrieval model on a budget},
+  author={Visheratin, Alexander},
+  journal={arXiv preprint arXiv:2309.01859},
+  year={2023}
+}
+```
+### CLIPA
+CLIPA models are described in the following papers by Xianhang Li, Zeyu Wang, Cihang Xie from UC Santa Cruz:
+- [An Inverse Scaling Law for CLIP Training](https://arxiv.org/abs/2305.07017)
+- [CLIPA-v2: Scaling CLIP Training with 81.1% Zero-shot ImageNet Accuracy within a $10,000 Budget; An Extra $4,000 Unlocks 81.8% Accuracy](https://arxiv.org/abs/2306.15658)
+Models are licensed under Apache 2.0.
+More details are available at https://github.com/UCSC-VLAA/CLIPA and [here](clipa.md).
+If you use CLIPA models, please cite the following:
+```bibtex
+@inproceedings{li2023clipa,
+      title={An Inverse Scaling Law for CLIP Training},
+      author={Xianhang Li and Zeyu Wang and Cihang Xie},
+      booktitle={NeurIPS},
+      year={2023},
+}
+```
+```bibtex
+@article{li2023clipav2,
+      title={CLIPA-v2: Scaling CLIP Training with 81.1% Zero-shot ImageNet Accuracy within a $10,000 Budget; An Extra $4,000 Unlocks 81.8% Accuracy},
+      author={Xianhang Li and Zeyu Wang and Cihang Xie},
+      journal={arXiv preprint arXiv:2306.15658},
+      year={2023},
+}
+```
+### SigLIP
+SigLIP models are described in the paper [Sigmoid Loss for Language Image Pre-Training](https://arxiv.org/abs/2303.15343).
+These models were developed by Xiaohua Zhai, Basil Mustafa, Alexander Kolesnikov, Lucas Beyer from Google Deepmind.
+Models are licensed under the Apache 2 license.
+More details are available at hhttps://github.com/google-research/big_vision.
+If you use SigLIP models, please cite the following:
+```bibtex
+@article{zhai2023sigmoid,
+  title={Sigmoid loss for language image pre-training},
+  author={Zhai, Xiaohua and Mustafa, Basil and Kolesnikov, Alexander and Beyer, Lucas},
+  journal={arXiv preprint arXiv:2303.15343},
+  year={2023}
+}
+```
+### DFN
+Data Filtering Network models are described in https://arxiv.org/abs/2309.17425.
+These models were developed by Alex Fang, Albin Madappally Jose, Amit Jain, Ludwig Schmidt, Alexander Toshev and Vaishaal Shankar from Apple and the University of Washington.
+Models are licensed under the following: https://huggingface.co/apple/DFN5B-CLIP-ViT-H-14-384/blob/main/LICENSE.
+If you use DFN models, please cite the following:
+```bibtext
+@article{fang2023data,
+  title={Data Filtering Networks},
+  author={Fang, Alex and Jose, Albin Madappally and Jain, Amit and Schmidt, Ludwig and Toshev, Alexander and Shankar, Vaishaal},
+  journal={arXiv preprint arXiv:2309.17425},
+  year={2023}
+}
+```