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

llama : make tensor_split ptr instead of array (#2272)

Browse source
This commit is contained in:
Georgi Gerganov 2023-07-21 13:10:51 +03:00 • committed by GitHub
parent 54e3bc76fe
commit ae178ab46b
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
4 changed files with 8 additions and 4 deletions
Download patch file Download diff file Expand all files Collapse all files

2
examples/common.cpp
View file

@ -586,7 +586,7 @@ struct llama_context_params llama_context_params_from_gpt_params(const gpt_param
lparams.n_batch = params.n_batch;
lparams.n_gpu_layers = params.n_gpu_layers;
lparams.main_gpu = params.main_gpu;
memcpy(lparams.tensor_split, params.tensor_split, LLAMA_MAX_DEVICES*sizeof(float));
lparams.tensor_split = params.tensor_split;
lparams.low_vram = params.low_vram;
lparams.seed = params.seed;
lparams.f16_kv = params.memory_f16;