diff --git a/crates/ctranslate2-bindings/src/ctranslate2.cc b/crates/ctranslate2-bindings/src/ctranslate2.cc index 3b71b19..62dacf3 100644 --- a/crates/ctranslate2-bindings/src/ctranslate2.cc +++ b/crates/ctranslate2-bindings/src/ctranslate2.cc @@ -118,7 +118,7 @@ std::shared_ptr create_engine( if (loader.device == ctranslate2::Device::CPU) { loader.compute_type = ctranslate2::ComputeType::INT8; } else if (loader.device == ctranslate2::Device::CUDA) { - loader.compute_type = ctranslate2::ComputeType::FLOAT16; + loader.compute_type = ctranslate2::ComputeType::INT8_FLOAT16; } if (model_type_str == "AutoModelForCausalLM") {