MCPcopy Create free account
hub / github.com/Tiiny-AI/PowerInfer / llama_context_default_params

Function llama_context_default_params

smallthinker/src/llama-context.cpp:2184–2218  ·  view source on GitHub ↗

Source from the content-addressed store, hash-verified

2182//
2183
2184llama_context_params llama_context_default_params() {
2185 llama_context_params result = {
2186 /*.n_ctx =*/ 512,
2187 /*.n_batch =*/ 2048,
2188 /*.n_ubatch =*/ 512,
2189 /*.n_seq_max =*/ 1,
2190 /*.n_threads =*/ GGML_DEFAULT_N_THREADS, // TODO: better default
2191 /*.n_threads_batch =*/ GGML_DEFAULT_N_THREADS,
2192 /*.rope_scaling_type =*/ LLAMA_ROPE_SCALING_TYPE_UNSPECIFIED,
2193 /*.pooling_type =*/ LLAMA_POOLING_TYPE_UNSPECIFIED,
2194 /*.attention_type =*/ LLAMA_ATTENTION_TYPE_UNSPECIFIED,
2195 /*.rope_freq_base =*/ 0.0f,
2196 /*.rope_freq_scale =*/ 0.0f,
2197 /*.yarn_ext_factor =*/ -1.0f,
2198 /*.yarn_attn_factor =*/ 1.0f,
2199 /*.yarn_beta_fast =*/ 32.0f,
2200 /*.yarn_beta_slow =*/ 1.0f,
2201 /*.yarn_orig_ctx =*/ 0,
2202 /*.defrag_thold =*/ -1.0f,
2203 /*.cb_eval =*/ nullptr,
2204 /*.cb_eval_user_data =*/ nullptr,
2205 /*.type_k =*/ GGML_TYPE_F16,
2206 /*.type_v =*/ GGML_TYPE_F16,
2207 /*.abort_callback =*/ nullptr,
2208 /*.abort_callback_data =*/ nullptr,
2209 /*.embeddings =*/ false,
2210 /*.offload_kqv =*/ true,
2211 /*.flash_attn =*/ false,
2212 /*.no_perf =*/ true,
2213 /*.op_offload =*/ true,
2214 /*.swa_full =*/ true,
2215 };
2216
2217 return result;
2218}
2219
2220llama_context * llama_init_from_model(
2221 llama_model * model,

Callers 13

to_llama_cparamsMethod · 0.50
mainFunction · 0.50
initMethod · 0.50
mainFunction · 0.50
mainFunction · 0.50
mainFunction · 0.50
mainFunction · 0.50
mainFunction · 0.50
mainFunction · 0.50
mainFunction · 0.50
mainFunction · 0.50

Calls

no outgoing calls

Tested by 6

mainFunction · 0.40
mainFunction · 0.40
mainFunction · 0.40
mainFunction · 0.40
mainFunction · 0.40
mainFunction · 0.40