gemma.cpp/gemma
Jan Wassenberg 12707ade80 Toward only using compressed weights:
CompressedLayer should all be f32 when weights are f32.

PiperOrigin-RevId: 640954519
2024-06-06 11:00:23 -07:00
..
evals Add MMLU eval to github 2024-05-20 10:20:53 -07:00
activations.h Simplifications: remove GemmaInterface and GemmaImpl 2024-06-06 05:54:21 -07:00
benchmark.cc Simplifications: remove GemmaInterface and GemmaImpl 2024-06-06 05:54:21 -07:00
common.cc Simplifications: remove GemmaInterface and GemmaImpl 2024-06-06 05:54:21 -07:00
common.h Simplifications: remove GemmaInterface and GemmaImpl 2024-06-06 05:54:21 -07:00
compress_weights.cc Simplifications: remove GemmaInterface and GemmaImpl 2024-06-06 05:54:21 -07:00
configs.h Simplifications: remove GemmaInterface and GemmaImpl 2024-06-06 05:54:21 -07:00
gemma.cc Simplifications: remove GemmaInterface and GemmaImpl 2024-06-06 05:54:21 -07:00
gemma.h Simplifications: remove GemmaInterface and GemmaImpl 2024-06-06 05:54:21 -07:00
gemma_test.cc Simplifications: remove GemmaInterface and GemmaImpl 2024-06-06 05:54:21 -07:00
ops.h Fix Softmax on SVE 2024-06-06 10:39:30 -07:00
ops_test.cc Fix Softmax on SVE 2024-06-06 10:39:30 -07:00
run.cc Simplifications: remove GemmaInterface and GemmaImpl 2024-06-06 05:54:21 -07:00
run_mmlu.cc Simplifications: remove GemmaInterface and GemmaImpl 2024-06-06 05:54:21 -07:00
weights.cc Simplifications: remove GemmaInterface and GemmaImpl 2024-06-06 05:54:21 -07:00
weights.h Toward only using compressed weights: 2024-06-06 11:00:23 -07:00