Alex Renda b061ba9e2a llama : fix top-p sampling to match the canonical definition (#1953) ...

* Fix top-p sampling to match the standard definition (smallest set that has probability mass at least p, not largest set with probability mass less than p)

* top-p: correct gt to gte

* add test for correct top-p behavior

2023-06-24 13:15:01 +03:00

..

CMakeLists.txt

ggml : implement backward pass for llama + small training-llama-from-scratch example (#1360)

2023-05-13 15:56:40 +03:00

test-double-float.c

all : be more strict about converting float to double (#458)

2023-03-28 19:48:20 +03:00

test-grad0.c

train : improved training-from-scratch example (#1652)

2023-06-13 22:04:40 +03:00

test-opt.c

ggml : implement backward pass for llama + small training-llama-from-scratch example (#1360)

2023-05-13 15:56:40 +03:00

test-quantize-fns.cpp

build : fix and ignore MSVC warnings (#1889)

2023-06-16 21:23:53 +03:00

test-quantize-perf.cpp

build : fix and ignore MSVC warnings (#1889)

2023-06-16 21:23:53 +03:00

test-sampling.cpp

llama : fix top-p sampling to match the canonical definition (#1953)

2023-06-24 13:15:01 +03:00

test-tokenizer-0.cpp

llama : make model stateless and context stateful (llama_state) (#1797)

2023-06-24 11:47:58 +03:00