diff options
Diffstat (limited to 'common/train.cpp')
| -rw-r--r-- | common/train.cpp | 10 |
1 files changed, 5 insertions, 5 deletions
diff --git a/common/train.cpp b/common/train.cpp index e4c3d5df..0dbfd24d 100644 --- a/common/train.cpp +++ b/common/train.cpp @@ -31,7 +31,7 @@ struct train_state * init_train_state() { state->opt = new struct ggml_opt_context; state->opt->ctx = NULL; - state->opt->params = ggml_opt_default_params(GGML_OPT_ADAM); + state->opt->params = ggml_opt_default_params(GGML_OPT_TYPE_ADAM); state->opt->params.graph_size = LLAMA_TRAIN_MAX_NODES; state->opt->loss_after = 0.0f; @@ -556,7 +556,7 @@ void load_opt_context_gguf(struct gguf_context * fctx, struct ggml_context * f_g std::string opt_type; GGUF_GET_KEY(fctx, opt_type, gguf_get_val_str, GGUF_TYPE_STRING, true, LLM_KV_OPTIMIZER_TYPE); if (opt_type == LLM_KV_OPTIMIZER_TYPE_ADAM) { - opt->params.type = GGML_OPT_ADAM; + opt->params.type = GGML_OPT_TYPE_ADAM; GGUF_GET_KEY(fctx, opt->adam.fx_best, gguf_get_val_f32, GGUF_TYPE_FLOAT32, true, LLM_KV_OPTIMIZER_ADAM_BEST_LOSS); GGUF_GET_KEY(fctx, opt->adam.fx_prev, gguf_get_val_f32, GGUF_TYPE_FLOAT32, true, LLM_KV_OPTIMIZER_ADAM_PREVIOUS_LOSS); @@ -568,7 +568,7 @@ void load_opt_context_gguf(struct gguf_context * fctx, struct ggml_context * f_g copy_tensor_by_name(opt->adam.v, f_ggml_ctx, LLM_TENSOR_OPTIMIZER_ADAM_SECOND_MOMENTS); copy_tensor_by_name(opt->adam.pf, f_ggml_ctx, LLM_TENSOR_OPTIMIZER_ADAM_PAST_LOSS_VALUES); } else if (opt_type == LLM_KV_OPTIMIZER_TYPE_LBFGS) { - opt->params.type = GGML_OPT_LBFGS; + opt->params.type = GGML_OPT_TYPE_LBFGS; GGUF_GET_KEY(fctx, opt->params.lbfgs.m, gguf_get_val_u32, GGUF_TYPE_UINT32, true, LLM_KV_OPTIMIZER_LBFGS_APPROX_HESSIAN_COUNT); GGUF_GET_KEY(fctx, opt->lbfgs.fx_best, gguf_get_val_f32, GGUF_TYPE_FLOAT32, true, LLM_KV_OPTIMIZER_LBFGS_BEST_LOSS); @@ -603,7 +603,7 @@ void save_opt_context_gguf(struct gguf_context * fctx, struct ggml_opt_context * gguf_set_val_bool(fctx, LLM_KV_OPTIMIZER_JUST_INITIALIZED, opt->just_initialized); switch (opt->params.type) { - case GGML_OPT_ADAM: + case GGML_OPT_TYPE_ADAM: { gguf_set_val_str(fctx, LLM_KV_OPTIMIZER_TYPE, LLM_KV_OPTIMIZER_TYPE_ADAM); gguf_set_val_f32(fctx, LLM_KV_OPTIMIZER_ADAM_BEST_LOSS, opt->adam.fx_best); @@ -622,7 +622,7 @@ void save_opt_context_gguf(struct gguf_context * fctx, struct ggml_opt_context * gguf_add_tensor(fctx, opt->adam.pf); } } break; - case GGML_OPT_LBFGS: + case GGML_OPT_TYPE_LBFGS: { gguf_set_val_str(fctx, LLM_KV_OPTIMIZER_TYPE, LLM_KV_OPTIMIZER_TYPE_LBFGS); gguf_set_val_u32(fctx, LLM_KV_OPTIMIZER_LBFGS_APPROX_HESSIAN_COUNT, opt->params.lbfgs.m); |