diff options
Diffstat (limited to 'common')
| -rw-r--r-- | common/common.cpp | 13 | ||||
| -rw-r--r-- | common/common.h | 7 |
2 files changed, 18 insertions, 2 deletions
diff --git a/common/common.cpp b/common/common.cpp index 1c0b7c40..dbe7e922 100644 --- a/common/common.cpp +++ b/common/common.cpp @@ -335,6 +335,16 @@ bool gpt_params_parse_ex(int argc, char ** argv, gpt_params & params) { break; } params.yarn_beta_slow = std::stof(argv[i]); + } else if (arg == "--pooling") { + if (++i >= argc) { + invalid_param = true; + break; + } + std::string value(argv[i]); + /**/ if (value == "none") { params.pooling_type = LLAMA_POOLING_TYPE_NONE; } + else if (value == "mean") { params.pooling_type = LLAMA_POOLING_TYPE_MEAN; } + else if (value == "cls") { params.pooling_type = LLAMA_POOLING_TYPE_CLS; } + else { invalid_param = true; break; } } else if (arg == "--defrag-thold" || arg == "-dt") { if (++i >= argc) { invalid_param = true; @@ -1014,6 +1024,8 @@ void gpt_print_usage(int /*argc*/, char ** argv, const gpt_params & params) { printf(" --yarn-attn-factor N YaRN: scale sqrt(t) or attention magnitude (default: 1.0)\n"); printf(" --yarn-beta-slow N YaRN: high correction dim or alpha (default: %.1f)\n", params.yarn_beta_slow); printf(" --yarn-beta-fast N YaRN: low correction dim or beta (default: %.1f)\n", params.yarn_beta_fast); + printf(" --pooling {none,mean,cls}\n"); + printf(" pooling type for embeddings, use model default if unspecified\n"); printf(" -dt N, --defrag-thold N\n"); printf(" KV cache defragmentation threshold (default: %.1f, < 0 - disabled)\n", params.defrag_thold); printf(" --ignore-eos ignore end of stream token and continue generating (implies --logit-bias 2-inf)\n"); @@ -1296,6 +1308,7 @@ struct llama_context_params llama_context_params_from_gpt_params(const gpt_param cparams.yarn_beta_fast = params.yarn_beta_fast; cparams.yarn_beta_slow = params.yarn_beta_slow; cparams.yarn_orig_ctx = params.yarn_orig_ctx; + cparams.pooling_type = params.pooling_type; cparams.defrag_thold = params.defrag_thold; cparams.offload_kqv = !params.no_kv_offload; diff --git a/common/common.h b/common/common.h index ab62bdb8..d3682b7a 100644 --- a/common/common.h +++ b/common/common.h @@ -76,8 +76,11 @@ struct gpt_params { float yarn_beta_slow = 1.0f; // YaRN high correction dim int32_t yarn_orig_ctx = 0; // YaRN original context length float defrag_thold = -1.0f; // KV cache defragmentation threshold - int32_t rope_scaling_type = LLAMA_ROPE_SCALING_TYPE_UNSPECIFIED; - ggml_numa_strategy numa = GGML_NUMA_STRATEGY_DISABLED; + + ggml_numa_strategy numa = GGML_NUMA_STRATEGY_DISABLED; + + llama_rope_scaling_type rope_scaling_type = LLAMA_ROPE_SCALING_TYPE_UNSPECIFIED; + llama_pooling_type pooling_type = LLAMA_POOLING_TYPE_UNSPECIFIED; // pooling type for embeddings // // sampling parameters struct llama_sampling_params sparams; |