ggml_cuda_init: GGML_CUDA_FORCE_MMQ: no ggml_cuda_init: GGML_CUDA_FORCE_CUBLAS: no ggml_cuda_init: found 2 ROCm devices: Device 0: AMD Radeon AI PRO R9700, gfx1201 (0x1201), VMM: no, Wave Size: 32 Device 1: AMD Radeon AI PRO R9700, gfx1201 (0x1201), VMM: no, Wave Size: 32 | model | size | params | backend | ngl | n_ubatch | fa | test | t/s | | ------------------------------ | ---------: | ---------: | ---------- | --: | -------: | -: | --------------: | -------------------: | /opt/llama.cpp/ggml/src/ggml-cuda/ggml-cuda.cu:92: ROCm error /usr/local/lib64/libggml-base.so.0(+0x3565) [0x7f08f345f565] /usr/local/lib64/libggml-base.so.0(ggml_print_backtrace+0x1eb) [0x7f08f345f92b] /usr/local/lib64/libggml-base.so.0(ggml_abort+0x11f) [0x7f08f345faaf] /usr/local/lib64/libggml-hip.so.0(+0x1b3df2) [0x7f08f36cfdf2] /usr/local/lib64/libggml-hip.so.0(+0x1c8677) [0x7f08f36e4677] /usr/local/lib64/libggml-hip.so.0(+0x1c6c28) [0x7f08f36e2c28] /usr/local/lib64/libggml-hip.so.0(+0x1c5dda) [0x7f08f36e1dda] /usr/local/lib64/libggml-hip.so.0(+0x1c081a) [0x7f08f36dc81a] /usr/local/lib64/libggml-hip.so.0(+0x1bcae6) [0x7f08f36d8ae6] /usr/local/lib64/libggml-hip.so.0(+0x1b905f) [0x7f08f36d505f] /usr/local/lib64/libggml-base.so.0(ggml_backend_sched_graph_compute_async+0x7f3) [0x7f08f3479f63] /usr/local/lib64/libllama.so.0(_ZN13llama_context13graph_computeEP11ggml_cgraphb+0xa0) [0x7f08f66cfa90] /usr/local/lib64/libllama.so.0(_ZN13llama_context14process_ubatchERK12llama_ubatch14llm_graph_typeP22llama_memory_context_iR11ggml_status+0xe2) [0x7f08f66d1722] /usr/local/lib64/libllama.so.0(_ZN13llama_context6decodeERK11llama_batch+0x3bf) [0x7f08f66d65df] /usr/local/lib64/libllama.so.0(llama_decode+0xe) [0x7f08f66d73fe] /usr/local/bin/llama-bench() [0x40acdb] /usr/local/bin/llama-bench() [0x4087ec] /lib64/libc.so.6(+0x35b5) [0x7f08f2df55b5] /lib64/libc.so.6(__libc_start_main+0x88) [0x7f08f2df5668] /usr/local/bin/llama-bench() [0x409b65] ✖ ! [rocm6_4_4-rocwmma] gemma-3-27b-it-BF16-00001-of-00002__fa1 __longctx16384 failed (exit 0)