Skip to content

Fixed OpenCL offloading prints #2082

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Jul 5, 2023
Merged

Fixed OpenCL offloading prints #2082

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
15 changes: 12 additions & 3 deletions llama.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1158,6 +1158,7 @@ static void llama_model_load_internal(
}
}
#endif // GGML_USE_CUBLAS

#if defined(GGML_USE_CUBLAS) || defined(GGML_USE_CLBLAST)
const int n_gpu = std::min(n_gpu_layers, int(hparams.n_layer));

Expand All @@ -1166,6 +1167,10 @@ static void llama_model_load_internal(
fprintf(stderr, "%s: offloading non-repeating layers to GPU\n", __func__);
}
size_t vram_kv_cache = 0;

#ifdef GGML_USE_CUBLAS
const int max_backend_supported_layers = hparams.n_layer + 3;
const int max_offloadable_layers = low_vram ? hparams.n_layer + 1 : hparams.n_layer + 3;
if (n_gpu_layers > (int) hparams.n_layer + 1) {
if (low_vram) {
fprintf(stderr, "%s: cannot offload v cache to GPU due to low VRAM option\n", __func__);
Expand All @@ -1182,14 +1187,18 @@ static void llama_model_load_internal(
vram_kv_cache += MEM_REQ_KV_SELF().at(model.type) / 2;
}
}
const int max_offloadable_layers = low_vram ? hparams.n_layer + 1 : hparams.n_layer + 3;
#elif defined(GGML_USE_CLBLAST)
const int max_backend_supported_layers = hparams.n_layer + 1;
const int max_offloadable_layers = hparams.n_layer + 1;
#endif // GGML_USE_CUBLAS

fprintf(stderr, "%s: offloaded %d/%d layers to GPU\n",
__func__, std::min(n_gpu_layers, max_offloadable_layers), hparams.n_layer + 3);
__func__, std::min(n_gpu_layers, max_offloadable_layers), max_backend_supported_layers);
fprintf(stderr, "%s: total VRAM used: %zu MB\n",
__func__, (vram_weights + vram_scratch + vram_kv_cache + MB - 1) / MB); // round up
#else
(void) n_gpu_layers;
#endif
#endif // defined(GGML_USE_CUBLAS) || defined(GGML_USE_CLBLAST)
}

// populate `tensors_by_name`
Expand Down