diff --git a/examples/finetune/finetune.cpp b/examples/finetune/finetune.cpp index 8884656ab..2e368fd9b 100644 --- a/examples/finetune/finetune.cpp +++ b/examples/finetune/finetune.cpp @@ -386,7 +386,7 @@ void init_model(struct llama_model * input, struct my_llama_model * model, uint3 snprintf(name, GGML_MAX_NAME, "layers.%d.feed_forward.w2.weight", i); layer.w2 = llama_get_model_tensor(input, name); - + snprintf(name, GGML_MAX_NAME, "layers.%d.feed_forward.w3.weight", i); layer.w3 = llama_get_model_tensor(input, name); } diff --git a/ggml.c b/ggml.c index 1e2ba6de0..52efe0d0f 100644 --- a/ggml.c +++ b/ggml.c @@ -10815,7 +10815,7 @@ static void ggml_compute_forward_out_prod_q_f32( float * s0 = (float *) ((char *) src0->data + ( i01*nb01 + i02*nb02 + i03*nb03)); float * s1 = (float *) ((char *) src1->data + (i1*nb10 + i11*nb11 + i12*nb12 + i13*nb13)); float * d = (float *) ((char *) dst->data + ( i1*nb1 + i2*nb2 + i3*nb3)); - + dequantize_row_q(s0, wdata, ne0); ggml_vec_mad_f32(ne0, d, wdata, *s1); }