Fix PyTorch ImportError

python3 main.py --interval_fetch 2 --persona "personas/tan.yaml" --guidance "Upscale the vocabulary and content of the segments to the highest level you can" Traceback (most recent call last): File "/Users/danielkliewer/news17/main.py", line 7, in <module> from src.core.generator import NewsGenerator File "/Users/danielkliewer/news17/src/core/generator.py", line 20, in <module> from src.nlp.clustering import StreamClusterer File "/Users/danielkliewer/news17/src/nlp/clustering.py", line 3, in <module> from sentence_transformers import SentenceTransformer File "/Users/danielkliewer/news17/venv/lib/python3.13/site-packages/sentence_transformers/__init__.py", line 9, i n <module> from sentence_transformers.backend import ( ...<3 lines>... ) File "/Users/danielkliewer/news17/venv/lib/python3.13/site-packages/sentence_transformers/backend.py", line 11, i n <module> from sentence_transformers.util import disable_datasets_caching, is_datasets_available File "/Users/danielkliewer/news17/venv/lib/python3.13/site-packages/sentence_transformers/util.py", line 17, in < module> import torch File "/Users/danielkliewer/news17/venv/lib/python3.13/site-packages/torch/__init__.py", line 409, in <module> from torch._C import * # noqa: F403 ^^^^^^^^^^^^^^^^^^^^^^ ImportError: dlopen(/Users/danielkliewer/news17/venv/lib/python3.13/site-packages/torch/_C.cpython-313-darwin.so, 0 x0002): Symbol not found: __ZN2at3cpu21is_amx_fp16_supportedEv Referenced from: <35AA4B87-2EA4-3AA3-B094-AF14FDCF53FB> /Users/danielkliewer/news17/venv/lib/python3.13/site-pack ages/torch/lib/libtorch_python.dylib Expected in: <02708A55-A57E-3D2F-9776-1A10E5BD6757> /opt/homebrew/Cellar/pytorch/2.5.1_4/libexec/lib/python3. 13/site-packages/torch/lib/libtorch_cpu.dylib

Got it — thanks for pasting the traceback. The important part is here: ``` ImportError: dlopen(...torch/_C.cpython-313-darwin.so, 0x0002): Symbol not found: __ZN2at3cpu21is_amx_fp16_supportedEv Referenced from: .../torch/lib/libtorch_python.dylib Expected in: .../torch/lib/libtorch_cpu.dylib ``` This means **your PyTorch install doesn’t match your Python version (3.13) or system libraries**. Right now, PyTorch does **not officially support Python 3.13** — the latest stable builds (as of 2.5.1) are for Python 3.12 and earlier. ### How to fix You have a few options: #### 1. Use Python 3.12 instead of 3.13 (recommended) ```bash # Create a new virtual environment with Python 3.12 pyenv install 3.12.6 # if you use pyenv pyenv virtualenv 3.12.6 news17-env pyenv activate news17-env # Or if you use homebrew: brew install python@3.12 python3.12 -m venv venv source venv/bin/activate ``` Then reinstall dependencies: ```bash pip install -U pip pip install torch sentence-transformers ``` #### 2. Try nightly builds (Python 3.13 support is experimental) ```bash pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cpu ``` ⚠️ This may still break — it’s bleeding-edge and not guaranteed stable. #### 3. Use Docker If you don’t want to downgrade your system Python, you can run your project inside a Python 3.12 Docker container with PyTorch pre-installed. --- 👉 Easiest path for now: **use Python 3.12** in your virtualenv for this project. That will eliminate the `dlopen` / `symbol not found` mismatch. Do you want me to write out a one-liner that sets up a fresh Python 3.12 environment and installs everything cleanly?

python3 main.py --interval_fetch 2 --persona "personas/tan.yaml" --guidance "Upscale the vocabulary and content of the segments to the highest level you can" Traceback (most recent call last): File "/Users/danielkliewer/news17/main.py", line 7, in <module> from src.core.generator import NewsGenerator File "/Users/danielkliewer/news17/src/core/generator.py", line 20, in <module> from src.nlp.clustering import StreamClusterer File "/Users/danielkliewer/news17/src/nlp/clustering.py", line 3, in <module> from sentence_transformers import SentenceTransformer File "/Users/danielkliewer/news17/venv/lib/python3.13/site-packages/sentence_transformers/__init__.py", line 9, i n <module> from sentence_transformers.backend import ( ...<3 lines>... ) File "/Users/danielkliewer/news17/venv/lib/python3.13/site-packages/sentence_transformers/backend.py", line 11, i n <module> from sentence_transformers.util import disable_datasets_caching, is_datasets_available File "/Users/danielkliewer/news17/venv/lib/python3.13/site-packages/sentence_transformers/util.py", line 17, in < module> import torch File "/Users/danielkliewer/news17/venv/lib/python3.13/site-packages/torch/__init__.py", line 409, in <module> from torch._C import * # noqa: F403 ^^^^^^^^^^^^^^^^^^^^^^ ImportError: dlopen(/Users/danielkliewer/news17/venv/lib/python3.13/site-packages/torch/_C.cpython-313-darwin.so, 0 x0002): Symbol not found: __ZN2at3cpu21is_amx_fp16_supportedEv Referenced from: <35AA4B87-2EA4-3AA3-B094-AF14FDCF53FB> /Users/danielkliewer/news17/venv/lib/python3.13/site-pack ages/torch/lib/libtorch_python.dylib Expected in: <02708A55-A57E-3D2F-9776-1A10E5BD6757> /opt/homebrew/Cellar/pytorch/2.5.1_4/libexec/lib/python3. 13/site-packages/torch/lib/libtorch_cpu.dylib

Thanks for rerunning and sharing the full traceback — this is the same root issue: `torch` was built against libraries that **don’t exist for Python 3.13 on macOS**. The missing symbol (`is_amx_fp16_supportedEv`) tells us your build of PyTorch is looking for CPU acceleration instructions it can’t find in your local libtorch. 👉 The key point: **PyTorch 2.5.1 does not officially support Python 3.13 yet.** --- ### ✅ Fix The most stable path right now is to **switch your project’s virtual environment to Python 3.12**. Here’s a safe set of steps: ```bash # 1. Install Python 3.12 if you don’t already have it brew install python@3.12 # if using Homebrew # 2. Create a new virtual environment with 3.12 python3.12 -m venv venv312 source venv312/bin/activate # 3. Upgrade pip pip install -U pip setuptools wheel # 4. Reinstall your project dependencies pip install -r requirements.txt ``` That should give you a compatible `torch` build. --- ### ⚡ Alternative (experimental) If you really want to stay on Python 3.13, you can try nightly builds of PyTorch: ```bash pip install --pre torch torchvision torchaudio \ --index-url https://download.pytorch.org/whl/nightly/cpu ``` …but this is **unstable** and may break other libraries (`sentence-transformers` in particular). --- Would you like me to give you a **drop-in script** that auto-detects if you’re on Python 3.13 and then spins up a Python 3.12 venv for your project so you don’t have to manage it manually?