Pulse · ggml-org/ggml · GitHub

February 18, 2025 – February 25, 2025

Overview

3 Active pull requests

3 Active issues

2 Pull requests merged by 2 people

Support pure float16 add/sub/mul/div operations in the CUDA (and CPU) backend
#1121 merged Feb 25, 2025
sync : llama.cpp
#1122 merged Feb 25, 2025

1 Pull request opened by 1 person

Skip metal backend tests for fp16 add/sub/mul/div operations (which don't have metal implementations)
#1123 opened Feb 25, 2025

1 Issue closed by 1 person

Why use the NT computation method in `mul mat` operator?
#1108 closed Feb 25, 2025

2 Issues opened by 2 people

Tests, examples fail to build with GGML_BACKEND_DL=ON
#1120 opened Feb 23, 2025
Can't add two float16 tensors on CUDA?
#1117 opened Feb 19, 2025