From da1429535bf53ad51e94dafbd70dc773bc395383 Mon Sep 17 00:00:00 2001 From: EL HACHIMI Aymane Date: Fri, 14 Nov 2025 16:13:38 +0100 Subject: [PATCH 1/4] Add example conda environment file --- environment.yml | 16 ++++++++++++++++ 1 file changed, 16 insertions(+) create mode 100644 environment.yml diff --git a/environment.yml b/environment.yml new file mode 100644 index 0000000..bdb93a6 --- /dev/null +++ b/environment.yml @@ -0,0 +1,16 @@ +name: evo2 +channels: + - nvidia + - conda-forge + - pytorch + - defaults +dependencies: + - python=3.12 + - pip + - pytorch=2.5.1 + - pytorch-cuda=12.1 + - transformer-engine-torch=2.3.0 + - flash-attn=2.8.0.post2 + - pip: + - vtx>=0.0.8 + - evo2 From 212bba9b86e4f3d9781058d666d69c4bdffd0c2b Mon Sep 17 00:00:00 2001 From: EL HACHIMI Aymane Date: Fri, 14 Nov 2025 16:27:06 +0100 Subject: [PATCH 2/4] Update the README File --- README.md | 32 ++++++++++++++++++++++++++++++++ 1 file changed, 32 insertions(+) diff --git a/README.md b/README.md index c92902c..2abdae1 100644 --- a/README.md +++ b/README.md @@ -84,6 +84,16 @@ For the 40b model: ``` python -m evo2.test.test_evo2_generation --model_name evo2_40b ``` +### Example conda environment + +For users who prefer a reproducible setup, we provide an example conda environment file (`environment.yml`) that pins Python, CUDA, PyTorch, Transformer Engine, FlashAttention and Vortex. + +You can create and activate the environment with: + +```bash +conda env create -f environment.yml +conda activate evo2 +python -m evo2.test.test_evo2_generation --model_name evo2_7b ### Docker @@ -100,6 +110,28 @@ Once inside the container: ```bash python -m evo2.test.test_evo2_generation --model_name evo2_7b ``` +### Troubleshooting (installation) + +Some common installation errors and suggested fixes: + +- `ModuleNotFoundError: No module named 'transformer_engine'` + Make sure the PyTorch extension for Transformer Engine is installed, for example: + ```bash + pip install "transformer_engine[pytorch]" + or follow the installation instructions from the Transformer Engine documentation. + + - `ModuleNotFoundError: No module named 'vortex'` + Install the Vortex package (published as `vtx` on PyPI), for example: + + ```bash + pip install "vtx>=0.0.8" + + - Errors mentioning `libnvrtc` or `Could not find shared object file for Transformer Engine torch lib` + These errors typically indicate a mismatch between CUDA, PyTorch, and Transformer Engine versions, or that CUDA libraries are not available in your `LD_LIBRARY_PATH`. We recommend: + + - using the same CUDA and PyTorch versions as in `environment.yml`, + - checking that your NVIDIA drivers and CUDA toolkit satisfy the requirements listed in the README, + - consulting the Transformer Engine and Vortex documentation for environment-specific installation details. ## Usage From 09cbd906e76cc30cbdbb51b5d424671c288332ef Mon Sep 17 00:00:00 2001 From: EL HACHIMI Aymane Date: Fri, 14 Nov 2025 16:32:39 +0100 Subject: [PATCH 3/4] Update README.md Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> --- README.md | 14 -------------- 1 file changed, 14 deletions(-) diff --git a/README.md b/README.md index 2abdae1..01ddf3e 100644 --- a/README.md +++ b/README.md @@ -118,20 +118,6 @@ Some common installation errors and suggested fixes: Make sure the PyTorch extension for Transformer Engine is installed, for example: ```bash pip install "transformer_engine[pytorch]" - or follow the installation instructions from the Transformer Engine documentation. - - - `ModuleNotFoundError: No module named 'vortex'` - Install the Vortex package (published as `vtx` on PyPI), for example: - - ```bash - pip install "vtx>=0.0.8" - - - Errors mentioning `libnvrtc` or `Could not find shared object file for Transformer Engine torch lib` - These errors typically indicate a mismatch between CUDA, PyTorch, and Transformer Engine versions, or that CUDA libraries are not available in your `LD_LIBRARY_PATH`. We recommend: - - - using the same CUDA and PyTorch versions as in `environment.yml`, - - checking that your NVIDIA drivers and CUDA toolkit satisfy the requirements listed in the README, - - consulting the Transformer Engine and Vortex documentation for environment-specific installation details. ## Usage From 6b27ac216000e291e33c4fa421376a001d81b0c4 Mon Sep 17 00:00:00 2001 From: EL HACHIMI Aymane Date: Fri, 14 Nov 2025 16:40:17 +0100 Subject: [PATCH 4/4] Move flash-attn to pip section as recommended by Gemini --- environment.yml | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/environment.yml b/environment.yml index bdb93a6..5c1370d 100644 --- a/environment.yml +++ b/environment.yml @@ -10,7 +10,7 @@ dependencies: - pytorch=2.5.1 - pytorch-cuda=12.1 - transformer-engine-torch=2.3.0 - - flash-attn=2.8.0.post2 - pip: + - flash-attn==2.8.0.post2 --no-build-isolation - vtx>=0.0.8 - evo2