| 1333 | std::vector<uint64_t> samples_ns; |
| 1334 | |
| 1335 | test(const cmd_params_instance & inst, const llama_model * lmodel, const llama_context * ctx) : |
| 1336 | cpu_info(get_cpu_info()), |
| 1337 | gpu_info(get_gpu_info()) { |
| 1338 | |
| 1339 | model_filename = inst.model; |
| 1340 | char buf[128]; |
| 1341 | llama_model_desc(lmodel, buf, sizeof(buf)); |
| 1342 | model_type = buf; |
| 1343 | model_size = llama_model_size(lmodel); |
| 1344 | model_n_params = llama_model_n_params(lmodel); |
| 1345 | n_batch = inst.n_batch; |
| 1346 | n_ubatch = inst.n_ubatch; |
| 1347 | n_threads = inst.n_threads; |
| 1348 | cpu_mask = inst.cpu_mask; |
| 1349 | cpu_strict = inst.cpu_strict; |
| 1350 | poll = inst.poll; |
| 1351 | type_k = inst.type_k; |
| 1352 | type_v = inst.type_v; |
| 1353 | n_gpu_layers = inst.n_gpu_layers; |
| 1354 | n_cpu_moe = inst.n_cpu_moe; |
| 1355 | split_mode = inst.split_mode; |
| 1356 | main_gpu = inst.main_gpu; |
| 1357 | no_kv_offload = inst.no_kv_offload; |
| 1358 | flash_attn = inst.flash_attn; |
| 1359 | devices = inst.devices; |
| 1360 | tensor_split = inst.tensor_split; |
| 1361 | tensor_buft_overrides = inst.tensor_buft_overrides; |
| 1362 | use_mmap = inst.use_mmap; |
| 1363 | use_direct_io = inst.use_direct_io; |
| 1364 | embeddings = inst.embeddings; |
| 1365 | no_op_offload = inst.no_op_offload; |
| 1366 | no_host = inst.no_host; |
| 1367 | n_prompt = inst.n_prompt; |
| 1368 | n_gen = inst.n_gen; |
| 1369 | n_depth = inst.n_depth; |
| 1370 | // RFC 3339 date-time format |
| 1371 | time_t t = time(NULL); |
| 1372 | std::strftime(buf, sizeof(buf), "%FT%TZ", gmtime(&t)); |
| 1373 | test_time = buf; |
| 1374 | |
| 1375 | (void) ctx; |
| 1376 | } |
| 1377 | |
| 1378 | uint64_t avg_ns() const { return ::avg(samples_ns); } |
| 1379 |
nothing calls this directly
no test coverage detected