pertTF 0.1.1__tar.gz

perttf-0.1.1/.github/workflows/publish-pypi.yml

@@ -0,0 +1,48 @@
+name: Publish to PyPI
+
+on:
+  workflow_dispatch:
+
+permissions:
+  contents: read
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.11'
+
+      - name: Build package
+        run: |
+          python -m pip install --upgrade pip build
+          python -m build
+
+      - name: Upload artifacts
+        uses: actions/upload-artifact@v4
+        with:
+          name: python-package-distributions
+          path: dist/
+
+  publish-pypi:
+    runs-on: ubuntu-latest
+    needs: build
+    environment:
+      name: pypi
+      url: https://pypi.org/p/pertTF
+    permissions:
+      id-token: write
+    steps:
+      - name: Download artifacts
+        uses: actions/download-artifact@v4
+        with:
+          name: python-package-distributions
+          path: dist/
+
+      - name: Publish to PyPI
+        uses: pypa/gh-action-pypi-publish@release/v1

perttf-0.1.1/.github/workflows/publish-testpypi.yml

@@ -0,0 +1,50 @@
+name: Publish to TestPyPI
+
+on:
+  workflow_dispatch:
+
+permissions:
+  contents: read
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.11'
+
+      - name: Build package
+        run: |
+          python -m pip install --upgrade pip build
+          python -m build
+
+      - name: Upload artifacts
+        uses: actions/upload-artifact@v4
+        with:
+          name: python-package-distributions
+          path: dist/
+
+  publish-testpypi:
+    runs-on: ubuntu-latest
+    needs: build
+    environment:
+      name: testpypi
+      url: https://test.pypi.org/p/pertTF
+    permissions:
+      id-token: write
+    steps:
+      - name: Download artifacts
+        uses: actions/download-artifact@v4
+        with:
+          name: python-package-distributions
+          path: dist/
+
+      - name: Publish to TestPyPI
+        uses: pypa/gh-action-pypi-publish@release/v1
+        with:
+          repository-url: https://test.pypi.org/legacy/

perttf-0.1.1/.gitignore

@@ -0,0 +1,11 @@
+scgpt
+.DS_Store
+*.h5ad
+wandb
+save
+*__pycache__*
+demos/run_on_hpc/.env
+data/*.h5ad
+runs/*
+demos/run_on_hpc/slurm.sh
+

perttf-0.1.1/INSTALL.md

@@ -0,0 +1,45 @@
+
+## Install Newest PertTF version 
+**Newest Version of PertTF can use Flash Attention for at least > 2x speed up, works on GPU enabled machines**
+
+This version still requires torchtext to be installed, because scgpt imports it, but it will not throw errors for torch > 2.3.0 version, this allows flash-attention to be utilized. pertTF no longer imports torchtext componenets, thus will work fine
+
+### Colab Install
+**NOTE for GOOGLE COLAB installation, check python, cuda and torch versions**
+ - Tesla T4s can only use flash v1
+ - V100s have no flash attn support (hopefully if SDPA layers are integrated into pertTF.py it can bring better performance)
+ - A100/L4 can use flash v2
+
+When installing flash v2 directly via `pip install flash-attn=xxx` as that will force compilation which takes forever, always install wheels
+currently colab comes with pytorch 2.8 and cuda 12.6
+The following might work (probably needs some tweeking)
+```
+pip install orbax==0.1.7
+pip install https://github.com/Dao-AILab/flash-attention/releases/download/v2.8.3/flash_attn-2.8.3+cu12torch2.8cxx11abiFALSE-cp312-cp312-linux_x86_64.whl
+pip install pandas scanpy "scvi-tools<1.0" numba datasets transformers==4.33.2 wandb cell-gears==0.1.2 torch_geometric pyarrow sentencepiece
+```
+
+### Clean Install on HPC
+```bash
+mamba create -n torch_flashv2 python=3.10 cuda-toolkit=12.8 'gxx>=6.0.0,<12.0' cudnn jupyter ipykernel ca-certificates matplotlib -y -c pytorch -c nvidia -c conda-forge
+pip install torch==2.6.0 torchvision orbax==0.1.7 torchdata torchmetrics 
+pip install pandas scanpy "scvi-tools<1.0" numba --upgrade "numpy<1.24" scib datasets transformers==4.33.2 wandb cell-gears==0.1.2 torch_geometric pyarrow sentencepiece
+
+# Install Flash attention v1 with the following (2x speed up) uncommented
+# pip install flash-attn==1.0.5 --no-build-isolation
+
+## Install Flash attention v2 (1.5x fasters than v1) with the following (or find a wheel that fits your python, cuda and torch version in github)
+pip install https://github.com/Dao-AILab/flash-attention/releases/download/v2.7.4.post1/flash_attn-2.7.4.post1+cu12torch2.6cxx11abiFALSE-cp310-cp310-linux_x86_64.whl 
+
+```
+
+#### INSTALL with FLASH ATTENTION v3 Beta (only on H200 architecture, 1.5-2x speed up over flash v2), 
+```bash
+mamba create -n pertTF_flashv3 python=3.10 'gxx>=6.0.0,<12.0' cudnn jupyter ipykernel ca-certificates matplotlib -y -c pytorch -c nvidia -c conda-forge
+pip install torch==2.8.0 torchvision orbax==0.1.7 torchdata torchmetrics 
+pip install pandas scanpy "scvi-tools<1.0" numba --upgrade "numpy<1.24" scib datasets==2.14.5 transformers==4.33.2 wandb cell-gears==0.0.2 torch_geometric pyarrow==15.0.2 ninja packages sentencepiece
+# To install flash attention v3 (1.5-2x speed up over v2) git required > 30mins and > 400GB RAM with 32 CPUS
+git clone https://github.com/Dao-AILab/flash-attention.git
+cd flash-attention
+python setup.py install 
+```

perttf-0.1.1/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2025 Wei Li
+
+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.

perttf-0.1.1/PKG-INFO

@@ -0,0 +1,79 @@
+Metadata-Version: 2.4
+Name: pertTF
+Version: 0.1.1
+Summary: Transformer model with single-cell perturbation predictions
+Author-email: Wei Li <wei@weililab.org>
+License: MIT
+Project-URL: Homepage, https://github.com/davidliwei/pertTF
+Project-URL: Repository, https://github.com/davidliwei/pertTF
+Keywords: single-cell,transformer,perturbation,bioinformatics
+Classifier: Development Status :: 3 - Alpha
+Classifier: Intended Audience :: Science/Research
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.7
+Classifier: Programming Language :: Python :: 3.8
+Classifier: Programming Language :: Python :: 3.9
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Requires-Python: >=3.7
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: scanpy>=1.10.0
+Dynamic: license-file
+
+<img src="assets/img/LOGO.png" alt="" width="800"/>
+
+**pertTF is a transformer model designed for single-cell perturbation predictions.**
+-----
+# Installation
+## Prerequisite environment
+pertTF require `torch > 2.3.0` and `cuda > 12.0` 
+
+best way to install is to set up a seperate envrionment with conda or mamba
+```bash
+# create independent environment (recommonded)
+mamba create -n pertTF_env python=3.10 cuda-toolkit=12.8 'gxx>=6.0.0,<12.0' cudnn ca-certificates -y -c pytorch -c nvidia -c conda-forge
+
+# pip install required packages
+# it is best to install torch == 2.6.0 to match the flash attention compiled wheel below
+# higher versions of torch may present difficulties for installing flash attention 2 
+pip install torch==2.6.0 torchvision orbax==0.1.7 torchdata torchmetrics pandas scanpy numba --upgrade "numpy<1.24" datasets transformers==4.33.2 wandb torch_geometric pyarrow sentencepiece huggingface_hub omegaconf
+```
+flash attention is strongly recommended for training or finetuning
+
+```bash
+# flash attention 2 installation
+#check ABI true/false first
+python -c "import torch;print(torch._C._GLIBCXX_USE_CXX11_ABI)"
+# install appropraite version (the example below is for ABI=FALSE)
+pip install https://github.com/Dao-AILab/flash-attention/releases/download/v2.7.4.post1/flash_attn-2.7.4.post1+cu12torch2.6cxx11abiFALSE-cp310-cp310-linux_x86_64.whl 
+
+# flash attention 3 installation (recommended for torch > 2.6.0 and hopper GPUs)
+# To install flash attention v3 (1.5-2x speed up over v2) requires > 30mins, > 400GB RAM, 32 CPUS (aim for more than this)
+git clone https://github.com/Dao-AILab/flash-attention.git
+cd flash-attention
+python setup.py install 
+```
+## pertTF installation
+You can install and use pertTF in two ways.
+
+The first way, pertTF is avaiable on PyPI. Use the following command to install pertTF:
+```bash
+pip install -i https://test.pypi.org/simple/ pertTF
+```
+The second way is suitable for you to run the most recent pertTF source code. First, fork our pertTF GitHub repository:
+
+```bash
+git clone https://github.com/davidliwei/pertTF.git
+```
+Then, in your python code, you can directly use the pertTF package:
+```python
+import sys
+sys.path.insert(0, '/content/pertTF/')
+```
+-----------------------
+
+## Tutorials
+
+- [Inference Tutorial](demos/tutorials/INFERENCE.md)

perttf-0.1.1/README.md

@@ -0,0 +1,55 @@
+<img src="assets/img/LOGO.png" alt="" width="800"/>
+
+**pertTF is a transformer model designed for single-cell perturbation predictions.**
+-----
+# Installation
+## Prerequisite environment
+pertTF require `torch > 2.3.0` and `cuda > 12.0` 
+
+best way to install is to set up a seperate envrionment with conda or mamba
+```bash
+# create independent environment (recommonded)
+mamba create -n pertTF_env python=3.10 cuda-toolkit=12.8 'gxx>=6.0.0,<12.0' cudnn ca-certificates -y -c pytorch -c nvidia -c conda-forge
+
+# pip install required packages
+# it is best to install torch == 2.6.0 to match the flash attention compiled wheel below
+# higher versions of torch may present difficulties for installing flash attention 2 
+pip install torch==2.6.0 torchvision orbax==0.1.7 torchdata torchmetrics pandas scanpy numba --upgrade "numpy<1.24" datasets transformers==4.33.2 wandb torch_geometric pyarrow sentencepiece huggingface_hub omegaconf
+```
+flash attention is strongly recommended for training or finetuning
+
+```bash
+# flash attention 2 installation
+#check ABI true/false first
+python -c "import torch;print(torch._C._GLIBCXX_USE_CXX11_ABI)"
+# install appropraite version (the example below is for ABI=FALSE)
+pip install https://github.com/Dao-AILab/flash-attention/releases/download/v2.7.4.post1/flash_attn-2.7.4.post1+cu12torch2.6cxx11abiFALSE-cp310-cp310-linux_x86_64.whl 
+
+# flash attention 3 installation (recommended for torch > 2.6.0 and hopper GPUs)
+# To install flash attention v3 (1.5-2x speed up over v2) requires > 30mins, > 400GB RAM, 32 CPUS (aim for more than this)
+git clone https://github.com/Dao-AILab/flash-attention.git
+cd flash-attention
+python setup.py install 
+```
+## pertTF installation
+You can install and use pertTF in two ways.
+
+The first way, pertTF is avaiable on PyPI. Use the following command to install pertTF:
+```bash
+pip install -i https://test.pypi.org/simple/ pertTF
+```
+The second way is suitable for you to run the most recent pertTF source code. First, fork our pertTF GitHub repository:
+
+```bash
+git clone https://github.com/davidliwei/pertTF.git
+```
+Then, in your python code, you can directly use the pertTF package:
+```python
+import sys
+sys.path.insert(0, '/content/pertTF/')
+```
+-----------------------
+
+## Tutorials
+
+- [Inference Tutorial](demos/tutorials/INFERENCE.md)

perttf-0.1.1/demos/run_on_hpc/.env_example

@@ -0,0 +1,4 @@
+WANDB_API_KEY=<WANDB_API_KEY_GOES_HERE>
+WANDB_SILENT=true
+WANDB__EXECUTABLE=/usr/local/bin/python
+KMP_WARNINGS=off

perttf-0.1.1/demos/run_on_hpc/README.md

@@ -0,0 +1,21 @@
+### scGPT Slurm Job Submission on HPC
+
+This repository contains the necessary scripts to run scGPT on a High Performance Computing (HPC) cluster using Slurm. The scripts are designed to be used with the [scGPT](https://github.com/bowang-lab/scGPT) repository.
+
+### Usage
+
+1. Clone this repository to your HPC cluster.
+2. Update the `.env_example` file with your wandb API key and save it as `.env` in the same directory.
+3. Create a `data` directory within the repository directory. Place the `.h5ad` file you want to use with scGPT into the `data` directory.
+4. Adjust the `slurm_example.sh` script to match the correct paths and parameters for your HPC cluster. Ensure you update `$DATA_PATH` in the same script to point to your `.h5ad` file.
+
+You can then use the following command to submit a job to the HPC cluster:
+```bash
+sbatch slurm.sh
+```
+
+The `slurm.sh` file will check whether the required software is installed and will proceed to set up the working directory if it is not. This includes: 
+
+- Cloning the scGPT repository
+- Downloading the necessary data
+- Setting up the Python environment (using UV)