core dumped

[MarsMeng1994]

when i use cmd like:
./build/bin/main --model quantized.bin --prompt \"{}\"".format(prompt)

it going to core dumped like:
<AI> ggml_new_tensor_impl: not enough space in the scratch memory pool (needed 292106304, available 150994944) Segmentation fault (core dumped)

how to fix?

[foldl]

Which model are you using?

[foldl]

@MarsMeng1994 You can enlarge this value, such as:

chatllm.cpp/models/minicpm.cpp

Line 88 in 7c21227

static constexpr size_t SCRATCH_SIZE = 344ull * 1024 * 1024;

See also Issue #6.