diff --git a/README.md b/README.md
index 3c5a614..18f1fae 100644
--- a/README.md
+++ b/README.md
@@ -1,42 +1,291 @@
-# FoundDiff
-Official implementation of "FoundDiff: Foundational Diffusion Model for Generalizable Low-Dose CT Denoising" 
+# FoundDiff - Environment Setup Guide (Working)
 
-## Approach
-![](figs/network.png)
+## Model Weights
 
+Download the required model weights from: [Google Drive](https://drive.google.com/drive/folders/1B33XyPqC9KkmzmfrCq20-7Xxuf-23PMc?usp=sharing)
 
-## Updates
-Feb, 2026: Upload DA-CLIP and FoundDiff model weight (https://drive.google.com/drive/folders/1B33XyPqC9KkmzmfrCq20-7Xxuf-23PMc?usp=sharing)   
-July, 2025: initial commit.  
+You will need:
+- `DA-CLIP.pth` - Place in `src/` (or as specified in DA-Diff.py)
+- `model-400.pt` - Place in `checkpoints/FoundDiff/sample/`
 
+## Quick Setup (Recommended)
 
-## Data Preparation
-The 2016 AAPM-Mayo dataset can be downloaded from: [CT Clinical Innovation Center](https://ctcicblog.mayo.edu/2016-low-dose-ct-grand-challenge/) (B30 kernel)  
-The 2020 AAPM-Mayo dataset can be downloaded from: [cancer imaging archive](https://wiki.cancerimagingarchive.net/pages/viewpage.action?pageId=52758026)   
+Use the provided `install.yaml` to create the environment:
 
+```bash
+conda env create -f install.yaml
+conda activate founddiff
 
-## Requirements
+# Install pre-built wheels for causal-conv1d and mamba-ssm
+pip install causal-conv1d==1.2.2.post1 --no-build-isolation
+pip install https://github.com/state-spaces/mamba/releases/download/v2.2.2/mamba_ssm-2.2.2+cu118torch2.1cxx11abiFalse-cp310-cp310-linux_x86_64.whl
+
+# Verify
+python -c "import selective_scan_cuda; print('CUDA kernels loaded successfully!')"
 ```
+
+## System Requirements
+
 - Linux Platform
-- torch==1.12.1+cu113 # depends on the CUDA version of your machine
-- torchvision==0.13.1+cu113
-- Python==3.8.0
-- numpy==1.22.3
+- NVIDIA GPU with CUDA 11.8+ support
+- Python 3.10 
+
+## Manual Step-by-Step Installation (Alternative)
+
+### 1. Create Conda Environment
+
+If you prefer to use the provided install.yaml, skip to step 4 after creating the environment:
+
+```bash
+conda create -n founddiff python=3.10 -y
+conda activate founddiff
+```
+
+### 2. Install PyTorch 2.1.0 with CUDA 11.8
+
+```bash
+pip install torch==2.1.0 torchvision==0.16.0 --index-url https://download.pytorch.org/whl/cu118
+pip install "numpy<2"
 ```
 
-## Traning and & Inference
+### 3. Install CUDA Toolkit 11.8
 
+**Critical**: Use the specific nvidia channel label to get CUDA 11.8 (not newer versions):
 
-#### Training:  
+```bash
+conda install -c "nvidia/label/cuda-11.8.0" cuda-toolkit -y
 ```
-CUDA_VISIBLE_DEVICES=1 python train.py --name FoundDiff --is_train --train_num_steps 400000
+
+Verify installation:
+```bash
+nvcc --version
+# Should show: CUDA compilation tools, release 11.8, V11.8.89
+```
+
+### 4. Install Prerequisites
+
+```bash
+pip install wheel packaging ninja
+pip install causal-conv1d==1.2.2.post1 --no-build-isolation
+```
+
+### 5. Install mamba-ssm (Pre-built Wheel)
+
+Download and install the pre-built wheel for CUDA 11.8 + Python 3.10:
+
+```bash
+pip install https://github.com/state-spaces/mamba/releases/download/v2.2.2/mamba_ssm-2.2.2+cu118torch2.1cxx11abiFalse-cp310-cp310-linux_x86_64.whl
+```
+
+### 6. Set Environment Variables
+
+Create activation script to ensure torch libraries are found:
+
+```bash
+mkdir -p $CONDA_PREFIX/etc/conda/activate.d/
+echo 'export LD_LIBRARY_PATH="$CONDA_PREFIX/lib/python3.10/site-packages/torch/lib:$CONDA_PREFIX/lib:$LD_LIBRARY_PATH"' > $CONDA_PREFIX/etc/conda/activate.d/env_vars.sh
+echo 'export CUDA_HOME="$CONDA_PREFIX"' >> $CONDA_PREFIX/etc/conda/activate.d/env_vars.sh
+echo 'export PATH="$CONDA_PREFIX/bin:$PATH"' >> $CONDA_PREFIX/etc/conda/activate.d/env_vars.sh
 ```
 
-#### Inference & testing:
-Put DA-CLIP.pth in src/DA-Diff.py and model-400.pt in checkpoints/FoundDiff/sample  
+### 7. Verify Installation
+
+```bash
+conda activate founddiff
+python -c "import selective_scan_cuda; print('CUDA kernels loaded successfully!')"
+```
+
+Should output: `CUDA kernels loaded successfully!`
+
+## Additional Dependencies
+
+The project requires several Python packages. Install them:
+
+```bash
+pip install kornia einops ipdb pywt accelerate ema-pytorch wandb Augmentor opencv-python lmdb scipy scikit-image matplotlib timm
+```
+
+Note: Some packages may require additional system dependencies or may have version conflicts. Install them one by one if needed.
+
+## Running the Project
+
+### Training
+
+```bash
+conda activate founddiff
+CUDA_VISIBLE_DEVICES=1 python train.py --name FoundDiff --is_train --train_num_steps 400000
 ```
+
+### Inference
+
+Download the model weights from the Google Drive link above, then run:
+
+```bash
+conda activate founddiff
 CUDA_VISIBLE_DEVICES=4 python train.py --name FoundDiff --epoch 400 --dataset 2020_seen
 ```
-Please refer to options files for more setting.
 
+Ensure:
+- `DA-CLIP.pth` is placed in `src/` (or as specified in DA-Diff.py)
+- `model-400.pt` is placed in `checkpoints/FoundDiff/sample/`
+
+## Environment Summary
+
+| Package | Version | Source |
+|---------|---------|--------|
+| Python | 3.10 | conda |
+| PyTorch | 2.1.0+cu118 | pytorch index |
+| torchvision | 0.16.0+cu118 | pytorch index |
+| numpy | 1.26.x | pip |
+| CUDA Toolkit | 11.8.0 | nvidia/label/cuda-11.8.0 |
+| causal-conv1d | 1.2.2.post1 | pip (pre-built) |
+| mamba-ssm | 2.2.2 | GitHub releases |
+
+## Data Format and Structure
+
+### Required Data Format
+
+The code expects data in **NumPy .npy format** (not PNG, JPG, or DICOM). Each file should be a 2D numpy array of shape (512, 512).
+
+```python
+# How data is loaded in the code (from pdf_dataset.py)
+q_data = np.load(self.q_path_list[index]).astype(np.float32)  # Low dose CT
+f_data = np.load(self.f_path_list[index]).astype(np.float32)  # Full dose CT
+```
+
+### Required Folder Structure
+
+The code expects data organized in this structure:
+
+#### For Mayo 2020 Dataset (Testing)
+
+```
+Mayo2020_ab_2d/test/full_1mm/          # High dose (full dose) - abdominal
+Mayo2020_ab_2d/test/quarter_1mm/       # Low dose (1/4 dose) - abdominal
+
+Mayo2020_lung_2d/test/full_1mm/       # High dose - lung
+Mayo2020_lung_2d/test/quarter_1mm/     # Low dose - lung
+
+Mayo2020_head_2d_2/test/full_1mm/       # High dose - head
+Mayo2020_head_2d_2/test/quarter_1mm/     # Low dose - head
+```
+
+#### For Mayo 2016 Dataset
+
+```
+Mayo2016_2d/test/full_1mm/           # High dose
+Mayo2016_2d/test/quarter_1mm/         # Low dose
+```
+
+Or alternatively:
+```
+Mayo2016_2d/train/full_1mm/          # High dose (training)
+Mayo2016_2d/quarter_1mm/            # Low dose
+```
+
+### Files to Modify for Custom Data Paths
+
+You need to modify paths in **2 files**:
+
+#### 1. data/pdf_dataset.py (Lines 331-399)
+
+Key paths to change:
+```python
+# Line 331-340: Abdominal data
+ab_ndct = sorted_list('/mnt/miah203/zhchen/Mayo2020_ab_2d/'+ phase+'/full_1mm/*')[:num]
+ab_dose_1_4_list = sorted_list('/mnt/miah203/zhchen/Mayo2020_ab_2d/'+phase+'/quarter_1mm/*')[start:num:stride]
+
+# Line 354-363: Lung data
+lung_ndct = sorted_list('/mnt/miah203/zhchen/Mayo2020_lung_2d/'+ phase+'/full_1mm/*')[:num]
+lung_dose_1_4_list = sorted_list('/mnt/miah203/zhchen/Mayo2020_lung_2d/'+phase+'/quarter_1mm/*')[start:num:stride]
+
+# Line 382-391: Head data
+head_ndct = sorted_list('/mnt/miah203/zhchen/Mayo2020_head_2d_2/'+ phase+'/full_1mm/*')[:num]
+head_dose_1_4_list = sorted_list('/mnt/miah203/zhchen/Mayo2020_head_2d_2/'+phase+'/quarter_1mm/*')[start:num:stride]
+
+# Line 398-399: Mayo2016 data
+self.mayo16_ldct = sorted_list('/mnt/miah203/zhchen/Mayo2016_2d/quarter_1mm/*')
+self.mayo16_ndct_path_list = sorted_list('/mnt/miah203/zhchen/Mayo2016_2d/train/full_1mm/*')
+```
+
+#### 2. data/mayo16_dataset.py (Lines 42-49)
+
+```python
+# Line 42-43: Test data paths
+self.f_path_list = sorted_list('/mnt/miah203/zhchen/CQ500_2d/test/full_1mm/*')
+self.q_path_list = sorted_list('/mnt/miah203/zhchen/CQ500_2d/test/sim-0.25/*')
+
+# Line 46-48: Training data paths
+self.q_path_list = sorted_list('/mnt/miah203/zhchen/Mayo2016_2d/test/quarter_1mm/*')
+self.f_path_list = sorted_list('/mnt/miah203/zhchen/Mayo2016_2d/test/full_1mm/*')
+```
+
+### Example: Setting Custom Data Path
+
+To use your own data at `/home/user/my_data/`, change the paths in pdf_dataset.py:
+
+```python
+# Before:
+ab_ndct = sorted_list('/mnt/miah203/zhchen/Mayo2020_ab_2d/'+ phase+'/full_1mm/*')[:num]
+
+# After:
+ab_ndct = sorted_list('/home/user/my_data/Mayo2020_ab_2d/'+ phase+'/full_1mm/*')[:num]
+```
+
+### Data Filename Matching
+
+The code matches low-dose and high-dose files by index:
+- Low dose file: `ab-001.npy` → loads from `quarter_1mm/`
+- High dose file: `001.npy` → loads from `full_1mm/` using index from filename
+
+Ensure your low-dose and high-dose files have matching indices in their filenames.
+
+### Converting DICOM to .npy
+
+If your data is in DICOM format (.IMA, .dcm), you need to convert it:
+
+```python
+import numpy as np
+import pydicom
+from PIL import Image
+import os
+
+def dicom_to_npy(dicom_path, output_path):
+    """Convert DICOM to .npy format"""
+    ds = pydicom.dcmread(dicom_path)
+    img = ds.pixel_array
+    
+    # Resize to 512x512 if needed
+    if img.shape != (512, 512):
+        img = Image.fromarray(img).resize((512, 512))
+        img = np.array(img)
+    
+    # Normalize to HU values if needed
+    # img = img * ds.RescaleSlope + ds.RescaleIntercept
+    
+    np.save(output_path, img.astype(np.float32))
+
+# Example usage:
+for f in os.listdir('/path/to/dicom/files'):
+    if f.endswith('.dcm') or f.endswith('.IMA'):
+        dicom_to_npy(f, f.replace('.dcm', '.npy'))
+```
+
+## Troubleshooting
+
+### "libc10.so: cannot open shared object file"
+- Ensure the activation script in Step 6 is created correctly
+- Or manually run: `export LD_LIBRARY_PATH="$CONDA_PREFIX/lib/python3.10/site-packages/torch/lib:$CONDA_PREFIX/lib:$LD_LIBRARY_PATH"`
+
+### "nvcc shows CUDA 13.2"
+- You installed cuda-toolkit without specifying the channel label
+- Recreate environment and use: `conda install -c "nvidia/label/cuda-11.8.0" cuda-toolkit -y`
+
+### mamba-ssm compilation fails
+- Always use pre-built wheels from GitHub releases
+- Building from source requires specific toolchain and is not recommended
+
+## Credits
 
+- FoundDiff: "Foundational Diffusion Model for Generalizable Low-Dose CT Denoising"
+- Mamba SSM: https://github.com/state-spaces/mamba
\ No newline at end of file
diff --git a/install.yaml b/install.yaml
index ca91772..da915e8 100644
--- a/install.yaml
+++ b/install.yaml
@@ -1,247 +1,40 @@
-name: diffusion
+name: founddiff
+channels:
+  - pytorch
+  - nvidia/label/cuda-11.8.0
+  - conda-forge
+  - defaults
 dependencies:
-  - _libgcc_mutex=0.1=main
-  - _openmp_mutex=5.1=1_gnu
-  - _pytorch_select=0.1=cpu_0
-  - autopep8=2.0.0=pyhd8ed1ab_0
-  - blas=1.0=mkl
-  - brotli=1.0.9=h5eee18b_7
-  - brotli-bin=1.0.9=h5eee18b_7
-  - brotlipy=0.7.0=py37h27cfd23_1003
-  - bzip2=1.0.8=h7b6447c_0
-  - ca-certificates=2022.9.24=ha878542_0
-  - certifi=2022.9.24=pyhd8ed1ab_0
-  - cffi=1.15.1=py37h74dc2b5_0
-  - cryptography=38.0.1=py37h9ce1e76_0
-  - cuda=11.6.2=0
-  - cuda-cccl=11.6.55=hf6102b2_0
-  - cuda-command-line-tools=11.6.2=0
-  - cuda-compiler=11.6.2=0
-  - cuda-cudart=11.6.55=he381448_0
-  - cuda-cudart-dev=11.6.55=h42ad0f4_0
-  - cuda-cuobjdump=11.6.124=h2eeebcb_0
-  - cuda-cupti=11.6.124=h86345e5_0
-  - cuda-cuxxfilt=11.6.124=hecbf4f6_0
-  - cuda-driver-dev=11.6.55=0
-  - cuda-gdb=11.8.86=0
-  - cuda-libraries=11.6.2=0
-  - cuda-libraries-dev=11.6.2=0
-  - cuda-memcheck=11.8.86=0
-  - cuda-nsight=11.8.86=0
-  - cuda-nsight-compute=11.8.0=0
-  - cuda-nvcc=11.6.124=hbba6d2d_0
-  - cuda-nvdisasm=11.8.86=0
-  - cuda-nvml-dev=11.6.55=haa9ef22_0
-  - cuda-nvprof=11.8.87=0
-  - cuda-nvprune=11.6.124=he22ec0a_0
-  - cuda-nvrtc=11.6.124=h020bade_0
-  - cuda-nvrtc-dev=11.6.124=h249d397_0
-  - cuda-nvtx=11.6.124=h0630a44_0
-  - cuda-nvvp=11.8.87=0
-  - cuda-runtime=11.6.2=0
-  - cuda-samples=11.6.101=h8efea70_0
-  - cuda-sanitizer-api=11.8.86=0
-  - cuda-toolkit=11.6.2=0
-  - cuda-tools=11.6.2=0
-  - cuda-visual-tools=11.6.2=0
-  - cudatoolkit=11.0.221=h6bb024c_0
-  - cycler=0.11.0=pyhd3eb1b0_0
-  - dbus=1.13.18=hb2f20db_0
-  - expat=2.4.9=h6a678d5_0
-  - ffmpeg=4.3=hf484d3e_0
-  - fontconfig=2.14.1=h52c9d5c_1
-  - fonttools=4.25.0=pyhd3eb1b0_0
-  - freetype=2.12.1=h4a9f257_0
-  - gds-tools=1.4.0.31=0
-  - giflib=5.2.1=h7b6447c_0
-  - glib=2.69.1=h4ff587b_1
-  - gmp=6.2.1=h295c915_3
-  - gnutls=3.6.15=he1e5248_0
-  - gst-plugins-base=1.14.0=h8213a91_2
-  - gstreamer=1.14.0=h28cd5cc_2
-  - icu=58.2=he6710b0_3
-  - idna=3.4=py37h06a4308_0
-  - intel-openmp=2021.4.0=h06a4308_3561
-  - jpeg=9e=h7f8727e_0
-  - kiwisolver=1.4.2=py37h295c915_0
-  - krb5=1.19.2=hac12032_0
-  - lame=3.100=h7b6447c_0
-  - lcms2=2.12=h3be6417_0
-  - ld_impl_linux-64=2.38=h1181459_1
-  - lerc=3.0=h295c915_0
-  - libbrotlicommon=1.0.9=h5eee18b_7
-  - libbrotlidec=1.0.9=h5eee18b_7
-  - libbrotlienc=1.0.9=h5eee18b_7
-  - libclang=10.0.1=default_hb85057a_2
-  - libcublas=11.11.3.6=0
-  - libcublas-dev=11.11.3.6=0
-  - libcufft=10.9.0.58=0
-  - libcufft-dev=10.9.0.58=0
-  - libcufile=1.4.0.31=0
-  - libcufile-dev=1.4.0.31=0
-  - libcurand=10.3.0.86=0
-  - libcurand-dev=10.3.0.86=0
-  - libcusolver=11.4.1.48=0
-  - libcusolver-dev=11.4.1.48=0
-  - libcusparse=11.7.5.86=0
-  - libcusparse-dev=11.7.5.86=0
-  - libdeflate=1.8=h7f8727e_5
-  - libedit=3.1.20221030=h5eee18b_0
-  - libevent=2.1.12=h8f2d780_0
-  - libffi=3.3=he6710b0_2
-  - libgcc-ng=11.2.0=h1234567_1
-  - libgomp=11.2.0=h1234567_1
-  - libiconv=1.16=h7f8727e_2
-  - libidn2=2.3.2=h7f8727e_0
-  - libllvm10=10.0.1=hbcb73fb_5
-  - libnpp=11.8.0.86=0
-  - libnpp-dev=11.8.0.86=0
-  - libnvjpeg=11.9.0.86=0
-  - libnvjpeg-dev=11.9.0.86=0
-  - libpng=1.6.37=hbc83047_0
-  - libpq=12.9=h16c4e8d_3
-  - libstdcxx-ng=11.2.0=h1234567_1
-  - libtasn1=4.16.0=h27cfd23_0
-  - libtiff=4.4.0=hecacb30_2
-  - libunistring=0.9.10=h27cfd23_0
-  - libuuid=1.41.5=h5eee18b_0
-  - libuv=1.40.0=h7b6447c_0
-  - libwebp=1.2.4=h11a3e52_0
-  - libwebp-base=1.2.4=h5eee18b_0
-  - libxcb=1.15=h7f8727e_0
-  - libxkbcommon=1.0.1=hfa300c1_0
-  - libxml2=2.9.14=h74e7548_0
-  - libxslt=1.1.35=h4e12654_0
-  - lz4-c=1.9.3=h295c915_1
-  - matplotlib=3.5.3=py37h06a4308_0
-  - matplotlib-base=3.5.3=py37hf590b9c_0
-  - mkl=2021.4.0=h06a4308_640
-  - mkl-service=2.4.0=py37h7f8727e_0
-  - mkl_fft=1.3.1=py37hd3c417c_0
-  - mkl_random=1.2.2=py37h51133e4_0
-  - munkres=1.1.4=py_0
-  - ncurses=6.3=h5eee18b_3
-  - nettle=3.7.3=hbbd107a_1
-  - ninja=1.10.2=h06a4308_5
-  - ninja-base=1.10.2=hd09550d_5
-  - nsight-compute=2022.3.0.22=0
-  - nspr=4.33=h295c915_0
-  - nss=3.74=h0370c37_0
-  - numpy=1.21.5=py37h6c91a56_3
-  - numpy-base=1.21.5=py37ha15fc14_3
-  - openh264=2.1.1=h4ff587b_0
-  - openssl=1.1.1o=h166bdaf_0
-  - packaging=21.3=pyhd3eb1b0_0
-  - pcre=8.45=h295c915_0
-  - pillow=9.2.0=py37hace64e9_1
-  - pip=22.2.2=py37h06a4308_0
-  - ply=3.11=py37_0
-  - pycodestyle=2.9.1=pyhd8ed1ab_0
-  - pycparser=2.21=pyhd3eb1b0_0
-  - pyopenssl=22.0.0=pyhd3eb1b0_0
-  - pyparsing=3.0.9=py37h06a4308_0
-  - pyqt=5.15.7=py37h6a678d5_1
-  - pyqt5-sip=12.11.0=py37h6a678d5_1
-  - pysocks=1.7.1=py37_1
-  - python=3.7.9=h7579374_0
-  - python-dateutil=2.8.2=pyhd3eb1b0_0
-  - pytorch=1.13.0=py3.7_cuda11.6_cudnn8.3.2_0
-  - pytorch-cuda=11.6=h867d48c_0
-  - pytorch-mutex=1.0=cuda
-  - qt-main=5.15.2=h327a75a_7
-  - qt-webengine=5.15.9=hd2b0992_4
-  - qtwebkit=5.212=h4eab89a_4
-  - readline=8.2=h5eee18b_0
-  - requests=2.28.1=py37h06a4308_0
-  - setuptools=65.5.0=py37h06a4308_0
-  - sip=6.6.2=py37h6a678d5_0
-  - six=1.16.0=pyhd3eb1b0_1
-  - sqlite=3.39.3=h5082296_0
-  - tk=8.6.12=h1ccaba5_0
-  - toml=0.10.2=pyhd3eb1b0_0
-  - tomli=2.0.1=pyhd8ed1ab_0
-  - torchaudio=0.13.0=py37_cu116
-  - torchvision=0.14.0=py37_cu116
-  - tornado=6.2=py37h5eee18b_0
-  - typing_extensions=4.3.0=py37h06a4308_0
-  - urllib3=1.26.12=py37h06a4308_0
-  - wheel=0.37.1=pyhd3eb1b0_0
-  - xz=5.2.6=h5eee18b_0
-  - zlib=1.2.13=h5eee18b_0
-  - zstd=1.5.2=ha4553b6_0
+  - python=3.10
+  - pytorch=2.1.0
+  - torchvision=0.16.0
+  - numpy<2
+  - scipy
+  - scikit-image
+  - matplotlib
+  - pillow
+  - pywavelets
+  - lmdb
+  - ninja
+  - wheel
+  - packaging
+  - pip
   - pip:
-    - absl-py==1.4.0
-    - accelerate==0.14.0
-    - addict==2.4.0
-    - astor==0.8.1
-    - augmentor==0.2.10
-    - cachetools==5.3.0
-    - charset-normalizer==2.1.1
-    - cloudpickle==2.2.0
-    - colorama==0.4.6
-    - contextlib2==21.6.0
-    - crc32c==2.3.post0
-    - diffusers==0.15.1
-    - einops==0.6.0
-    - ema-pytorch==0.0.10
-    - filelock==3.8.0
-    - flash-attn==1.0.2
-    - future==0.18.2
-    - fvcore==0.1.5.post20220512
-    - google-auth==2.16.2
-    - google-auth-oauthlib==0.4.6
-    - grpcio==1.51.3
-    - huggingface-hub==0.13.4
-    - imageio==2.22.4
-    - importlib-metadata==5.0.0
-    - iopath==0.1.10
-    - joblib==1.2.0
-    - json-tricks==3.16.1
-    - lmdb==1.4.0
-    - markdown==3.4.1
-    - markupsafe==2.1.2
-    - natsort==8.2.0
-    - networkx==2.6.3
-    - nni==2.10
-    - oauthlib==3.2.2
-    - opencv-python==4.7.0.68
-    - pandas==1.3.5
-    - portalocker==2.6.0
-    - prettytable==3.5.0
-    - protobuf==3.20.3
-    - psutil==5.9.4
-    - pyasn1==0.4.8
-    - pyasn1-modules==0.2.8
-    - pythonwebhdfs==0.2.3
-    - pytorch-fid==0.3.0
-    - pytorch-warmup==0.1.1
-    - pytz==2022.6
-    - pywavelets==1.3.0
-    - pyyaml==6.0
-    - regex==2022.10.31
-    - requests-oauthlib==1.3.1
-    - responses==0.22.0
-    - rsa==4.9
-    - schema==0.7.5
-    - scikit-image==0.19.3
-    - scikit-learn==1.0.2
-    - scipy==1.7.3
-    - simplejson==3.18.0
-    - tabulate==0.9.0
-    - tensorboard==2.11.2
-    - tensorboard-data-server==0.6.1
-    - tensorboard-plugin-wit==1.8.1
-    - termcolor==2.1.0
-    - tfrecord==1.14.1
-    - thop==0.1.1-2209072238
-    - threadpoolctl==3.1.0
-    - tifffile==2021.11.2
-    - timm==0.6.11
-    - tqdm==4.64.1
-    - typeguard==2.13.3
-    - types-toml==0.10.8.1
-    - wcwidth==0.2.5
-    - websockets==10.4
-    - werkzeug==2.2.3
-    - yacs==0.1.8
-    - zipp==3.10.0
+    - einops
+    - kornia
+    - timm
+    - accelerate
+    - ema-pytorch==0.3.3
+    - wandb
+    - augmentor
+    - opencv-python
+    - pydicom
+    - imageio
+    - ipdb
+
+# Note: causal-conv1d and mamba-ssm require pre-built wheels for your CUDA/Python version.
+# Install them manually after creating the environment:
+#   pip install causal-conv1d==1.2.2.post1 --no-build-isolation
+#   pip install https://github.com/state-spaces/mamba/releases/download/v2.2.2/mamba_ssm-2.2.2+cu118torch2.1cxx11abiFalse-cp310-cp310-linux_x86_64.whl
+#
+# Or check https://github.com/state-spaces/mamba/releases for the correct wheel for your setup.
\ No newline at end of file