Skip to content

Commit da54f9f

Browse files
authored
presets : add qwen3-30B-a3b FIM (ggml-org#15616)
1 parent 4737327 commit da54f9f

File tree

1 file changed

+16
-0
lines changed

1 file changed

+16
-0
lines changed

common/arg.cpp

Lines changed: 16 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -3538,6 +3538,22 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
35383538
}
35393539
).set_examples({LLAMA_EXAMPLE_SERVER}));
35403540

3541+
add_opt(common_arg(
3542+
{"--fim-qwen-30b-default"},
3543+
string_format("use default Qwen 3 Coder 30B A3B Instruct (note: can download weights from the internet)"),
3544+
[](common_params & params) {
3545+
params.model.hf_repo = "ggml-org/Qwen3-Coder-30B-A3B-Instruct-Q8_0-GGUF";
3546+
params.model.hf_file = "qwen3-coder-30b-a3b-instruct-q8_0.gguf";
3547+
params.port = 8012;
3548+
params.n_gpu_layers = 99;
3549+
params.flash_attn = true;
3550+
params.n_ubatch = 1024;
3551+
params.n_batch = 1024;
3552+
params.n_ctx = 0;
3553+
params.n_cache_reuse = 256;
3554+
}
3555+
).set_examples({LLAMA_EXAMPLE_SERVER}));
3556+
35413557
add_opt(common_arg(
35423558
{ "--diffusion-steps" }, "N",
35433559
string_format("number of diffusion steps (default: %d)", params.diffusion.steps),

0 commit comments

Comments
 (0)