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) [0x7fcdb1288565] /usr/local/lib64/libggml-base.so.0(ggml_print_backtrace+0x1eb) [0x7fcdb128892b] /usr/local/lib64/libggml-base.so.0(ggml_abort+0x11f) [0x7fcdb1288aaf] /usr/local/lib64/libggml-hip.so.0(+0x2fd2ef2) [0x7fcdb4317ef2] /usr/local/lib64/libggml-hip.so.0(+0x2fe7757) [0x7fcdb432c757] /usr/local/lib64/libggml-hip.so.0(+0x2fe5d2c) [0x7fcdb432ad2c] /usr/local/lib64/libggml-hip.so.0(+0x2fe4ef7) [0x7fcdb4329ef7] /usr/local/lib64/libggml-hip.so.0(+0x2fdf9db) [0x7fcdb43249db] /usr/local/lib64/libggml-hip.so.0(+0x2fdbcab) [0x7fcdb4320cab] /usr/local/lib64/libggml-hip.so.0(+0x2fd816f) [0x7fcdb431d16f] /usr/local/lib64/libggml-base.so.0(ggml_backend_sched_graph_compute_async+0x7f3) [0x7fcdb12a2fe3] /usr/local/lib64/libllama.so.0(_ZN13llama_context13graph_computeEP11ggml_cgraphb+0xa0) [0x7fcdb49e5a10] /usr/local/lib64/libllama.so.0(_ZN13llama_context14process_ubatchERK12llama_ubatch14llm_graph_typeP22llama_memory_context_iR11ggml_status+0xe2) [0x7fcdb49e76a2] /usr/local/lib64/libllama.so.0(_ZN13llama_context6decodeERK11llama_batch+0x3bf) [0x7fcdb49ec55f] /usr/local/lib64/libllama.so.0(llama_decode+0xe) [0x7fcdb49ed37e] /usr/local/bin/llama-bench() [0x40acdb] /usr/local/bin/llama-bench() [0x4087ec] /lib64/libc.so.6(+0x35b5) [0x7fcdb0c1e5b5] /lib64/libc.so.6(__libc_start_main+0x88) [0x7fcdb0c1e668] /usr/local/bin/llama-bench() [0x409b65] ✖ ! [rocm7.1.1] gemma-3-27b-it-BF16-00001-of-00002__hblt0__fa1 __longctx16384 failed (exit 0)