mcale6

Why is there a MAX_TEMPLATE =4 in the alphafold code? So even I submit 100 custom templates only 4 will be used?

Apple M1 Max cannot load llama3-8b-sfr weights (because no bfloat support?)

5

Mac OS, Metal, M1 MAX Macbook Pro, Python 3.11.6 in a _conda-env_ Running `python3 examples/llama3.py --model downloads/llama3-8b-sfr/ --no_api ` ``` loaded weights in 33.01 ms, 0.03 GB loaded at 1.02...