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

Add fprintf in ggml_cuda_op_add

Browse source 
When I tried CUDA offloading during finetuning following the readme, I got an assert here.
This probably isn't an important case because inference later gives a warning saying you should use f16 or f32 instead when using lora
This commit is contained in:
Andrew Godfrey 2023-10-21 22:40:40 -07:00
parent facb1a3e0f
commit 4d452dbc10
1 changed files with 1 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

1
ggml-cuda.cu
View file

@ -5910,6 +5910,7 @@ inline void ggml_cuda_op_add(
} else if (src0->type == GGML_TYPE_F16 && dst->type == GGML_TYPE_F16) { } else if (src0->type == GGML_TYPE_F16 && dst->type == GGML_TYPE_F16) {
add_f16_f32_f16_cuda((const half *) src0_dd, src1_dd, (half *) dst_dd, ggml_nelements(src0), main_stream); add_f16_f32_f16_cuda((const half *) src0_dd, src1_dd, (half *) dst_dd, ggml_nelements(src0), main_stream);
} else { } else {
fprintf(stderr, "%d, %d\n", src0->type, dst->type);
GGML_ASSERT(false); GGML_ASSERT(false);
} }