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

Use std::runtime_error exceptions.

Browse source
This commit is contained in:
KerfuffleV2 2023-06-05 14:33:53 -06:00
parent 7ed5aca9ca
commit b3d605dc75
1 changed files with 3 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

6
llama.cpp
View file

@ -2133,10 +2133,10 @@ static void llama_convert_tensor_internal(const llama_load_tensor & tensor, llam
if (ggml_is_quantized(tensor.type)) { if (ggml_is_quantized(tensor.type)) {
qtype = ggml_internal_get_quantize_fn(tensor.type); qtype = ggml_internal_get_quantize_fn(tensor.type);
if (qtype.dequantize_row_q == NULL) { if (qtype.dequantize_row_q == NULL) {
throw format("type %s unsupported for integer quantization: no dequantization available", ggml_type_name(tensor.type)); throw std::runtime_error(format("type %s unsupported for integer quantization: no dequantization available", ggml_type_name(tensor.type)));
} }
} else if (tensor.type != GGML_TYPE_F16) { } else if (tensor.type != GGML_TYPE_F16) {
throw format("cannot dequantize/convert tensor type %s", ggml_type_name(tensor.type)); throw std::runtime_error(format("cannot dequantize/convert tensor type %s", ggml_type_name(tensor.type)));
} }
if (nthread < 2) { if (nthread < 2) {
@ -2299,7 +2299,7 @@ static void llama_model_quantize_internal(const std::string & fname_inp, const s
if (tensor.type == GGML_TYPE_F32) { if (tensor.type == GGML_TYPE_F32) {
f32_data = (float *) tensor.data; f32_data = (float *) tensor.data;
} else if (ggml_is_quantized(tensor.type) && !params->allow_requantize) { } else if (ggml_is_quantized(tensor.type) && !params->allow_requantize) {
throw format("requantizing from type %s is disabled", ggml_type_name(tensor.type)); throw std::runtime_error(format("requantizing from type %s is disabled", ggml_type_name(tensor.type)));
} else { } else {
llama_convert_tensor_internal(tensor, f32_conv_buf, nelements, nthread); llama_convert_tensor_internal(tensor, f32_conv_buf, nelements, nthread);
f32_data = (float *) f32_conv_buf.addr; f32_data = (float *) f32_conv_buf.addr;