| 26019 | low_row_elems * sizeof(float)); |
| 26020 | ds4_gpu_tensor *out_row = ds4_gpu_tensor_view( |
| 26021 | out, |
| 26022 | (uint64_t)t * out_dim * sizeof(float), |
| 26023 | out_dim * sizeof(float)); |
| 26024 | ok = heads_row && low_row && out_row && |
| 26025 | metal_graph_attention_output_dense_quant_low(low_row, |
| 26026 | g, |
| 26027 | model, |
| 26028 | out_a, |
| 26029 | group_dim, |
| 26030 | rank, |
| 26031 | 0, |
| 26032 | n_groups, |
| 26033 | heads_row); |
| 26034 | if (ok) ok = metal_graph_matmul_dense_quant_tensor(out_row, |
| 26035 | model, |
| 26036 | out_b, |
| 26037 | low_row_elems, |
| 26038 | out_dim, |
| 26039 | low_row, |
| 26040 | 1); |