vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

metal : pad n_ctx by 32 (#6177)

Browse source 
* metal : require ne00 >= 128 for mat-mat kernels

ggml-ci

* llama : pad n_ctx by 32

ggml-ci
This commit is contained in:
Georgi Gerganov 2024-03-22 09:36:03 +02:00 • committed by GitHub
parent 59c17f02de
commit 95d576b48e
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
4 changed files with 14 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

4
examples/batched/batched.cpp
View file

@ -48,6 +48,8 @@ int main(int argc, char ** argv) {
params.prompt = "Hello my name is";
}
process_escapes(params.prompt);
// init LLM
llama_backend_init();
@ -78,7 +80,7 @@ int main(int argc, char ** argv) {
llama_context_params ctx_params = llama_context_default_params();
ctx_params.seed = 1234;
ctx_params.n_ctx = n_kv_req;
ctx_params.n_ctx = n_kv_req;
ctx_params.n_batch = std::max(n_len, n_parallel);
ctx_params.n_seq_max = n_parallel;
ctx_params.n_threads = params.n_threads;