|
|
|
@ -325,10 +325,9 @@ bool gpt2_model_load(const std::string & fname, gpt2_model & model, gpt_vocab &
|
|
|
|
|
|
|
|
|
|
// create the ggml context
|
|
|
|
|
{
|
|
|
|
|
struct ggml_init_params params = {
|
|
|
|
|
.mem_size = ctx_size,
|
|
|
|
|
.mem_buffer = NULL,
|
|
|
|
|
};
|
|
|
|
|
struct ggml_init_params params;
|
|
|
|
|
params.mem_size = ctx_size;
|
|
|
|
|
params.mem_buffer = NULL;
|
|
|
|
|
|
|
|
|
|
model.ctx = ggml_init(params);
|
|
|
|
|
if (!model.ctx) {
|
|
|
|
@ -529,10 +528,9 @@ bool gpt2_eval(
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
struct ggml_init_params params = {
|
|
|
|
|
.mem_size = buf_size,
|
|
|
|
|
.mem_buffer = buf,
|
|
|
|
|
};
|
|
|
|
|
struct ggml_init_params params;
|
|
|
|
|
params.mem_size = buf_size;
|
|
|
|
|
params.mem_buffer = buf;
|
|
|
|
|
|
|
|
|
|
struct ggml_context * ctx0 = ggml_init(params);
|
|
|
|
|
struct ggml_cgraph gf = { .n_threads = n_threads };
|
|
|
|
|