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

fix error: wrong result in 658746bb26702e50f2c59c0e4ada8e9da6010481

Browse source
This commit is contained in:
jianyuzh 2024-01-13 19:55:30 +08:00 • committed by Meng, Hengyu
parent bd38129aeb
commit 5b5389941e
2 changed files with 4 additions and 4 deletions
Download patch file Download diff file Expand all files Collapse all files

2
ggml-sycl.cpp
View file

@ -9712,7 +9712,7 @@ static void ggml_sycl_op_flatten(const ggml_tensor *src0,
src1_ddf = (float *) src1_extra->data_device[g_main_device_index];
} else {
src1_ddf = src1_f.alloc(ggml_nelements(src1));
// SYCL_CHECK(ggml_sycl_cpy_tensor_2d(src1_ddf, src1, 0, 0, 0, nrows1, main_stream));
SYCL_CHECK(ggml_sycl_cpy_tensor_2d(src1_ddf, src1, 0, 0, 0, nrows1, main_stream));
}
}
if (dst_on_device) {

6
run.sh
View file

@ -11,9 +11,9 @@ else
export GGML_SYCL_DEVICE=0
fi
echo GGML_SYCL_DEVICE=$GGML_SYCL_DEVICE
export GGML_SYCL_DEBUG=1
#export GGML_SYCL_DEBUG=1
#export GGML_SYCL_LIST_DEVICE=1
#./build/bin/main -m models/llama-2-7b.Q4_0.gguf -p "${INPUT1}" -e -n 400 -ngl 33 -c 2048
#./build/bin/main -m models/llama-2-7b.Q4_0.gguf -p "${INPUT2}" -n 400 -e -ngl 33
./build/bin/main -m models/llama-2-7b.Q4_0.gguf -p "${INPUT2}" -n 5 -e -ngl 33 -t 1 -s 0
./build/bin/main -m models/llama-2-7b.Q4_0.gguf -p "${INPUT2}" -n 400 -e -ngl 33
#./build/bin/main -m models/llama-2-7b.Q4_0.gguf -p "${INPUT2}" -n 5 -e -ngl 33 -t 1 -s 0