diff options
| author | Kawrakow <48489457+ikawrakow@users.noreply.github.com> | 2024-07-27 07:55:01 +0200 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2024-07-27 07:55:01 +0200 |
| commit | 154e0d75fccf1784fe9ff6fd76a630b66563da3d (patch) | |
| tree | 81ce6dbb5b1900c1aa78a879f0593c694cab9d27 /examples/lookup | |
| parent | 0684c3e9c70d49323b4fc517128cbe222cab7f96 (diff) | |
Merge mainline llama.cpp (#3)
* Merging mainline - WIP
* Merging mainline - WIP
AVX2 and CUDA appear to work.
CUDA performance seems slightly (~1-2%) lower as it is so often
the case with llama.cpp/ggml after some "improvements" have been made.
* Merging mainline - fix Metal
* Remove check
---------
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
Diffstat (limited to 'examples/lookup')
| -rw-r--r-- | examples/lookup/README.md | 1 | ||||
| -rw-r--r-- | examples/lookup/lookup-stats.cpp | 3 | ||||
| -rw-r--r-- | examples/lookup/lookup.cpp | 1 |
3 files changed, 1 insertions, 4 deletions
diff --git a/examples/lookup/README.md b/examples/lookup/README.md index 5bfb0de9..71c345c0 100644 --- a/examples/lookup/README.md +++ b/examples/lookup/README.md @@ -10,4 +10,3 @@ More info: https://github.com/ggerganov/llama.cpp/pull/4484 https://github.com/ggerganov/llama.cpp/issues/4226 - diff --git a/examples/lookup/lookup-stats.cpp b/examples/lookup/lookup-stats.cpp index 0b171c87..2fe67100 100644 --- a/examples/lookup/lookup-stats.cpp +++ b/examples/lookup/lookup-stats.cpp @@ -31,7 +31,6 @@ int main(int argc, char ** argv){ // load the model std::tie(model, ctx) = llama_init_from_gpt_params(params); - GGML_ASSERT(llama_n_vocab(model) < (1 << 16)); // tokenize the prompt std::vector<llama_token> inp; @@ -65,7 +64,7 @@ int main(int argc, char ** argv){ } const int n_input = inp.size(); - const int n_ctx = params.n_ctx; + const int n_ctx = llama_n_ctx(ctx); int n_drafted = 0; int n_accept = 0; diff --git a/examples/lookup/lookup.cpp b/examples/lookup/lookup.cpp index 80ecd925..bb571bac 100644 --- a/examples/lookup/lookup.cpp +++ b/examples/lookup/lookup.cpp @@ -39,7 +39,6 @@ int main(int argc, char ** argv){ // load the model std::tie(model, ctx) = llama_init_from_gpt_params(params); - GGML_ASSERT(llama_n_vocab(model) < (1 << 16)); // tokenize the prompt std::vector<llama_token> inp; |