Updates the LR used for subsequent backward passes. Can be called between epochs to implement LR scheduling.
ggml_opt_set_lr(lr_ud, adamw_lr = NA, sgd_lr = NA)NULL invisibly
LR userdata pointer (from `ggml_opt_init_for_fit()$lr_ud`)
New AdamW learning rate (NA to keep current)
New SGD learning rate (NA to keep current)
Other optimization:
ggml_fit_opt(),
ggml_opt_alloc(),
ggml_opt_context_optimizer_type(),
ggml_opt_dataset_data(),
ggml_opt_dataset_free(),
ggml_opt_dataset_get_batch(),
ggml_opt_dataset_init(),
ggml_opt_dataset_labels(),
ggml_opt_dataset_ndata(),
ggml_opt_dataset_shuffle(),
ggml_opt_default_params(),
ggml_opt_epoch(),
ggml_opt_eval(),
ggml_opt_fit(),
ggml_opt_free(),
ggml_opt_get_lr(),
ggml_opt_grad_acc(),
ggml_opt_init(),
ggml_opt_init_for_fit(),
ggml_opt_inputs(),
ggml_opt_labels(),
ggml_opt_loss(),
ggml_opt_loss_type_cross_entropy(),
ggml_opt_loss_type_mean(),
ggml_opt_loss_type_mse(),
ggml_opt_loss_type_sum(),
ggml_opt_ncorrect(),
ggml_opt_optimizer_name(),
ggml_opt_optimizer_type_adamw(),
ggml_opt_optimizer_type_sgd(),
ggml_opt_outputs(),
ggml_opt_pred(),
ggml_opt_prepare_alloc(),
ggml_opt_reset(),
ggml_opt_result_accuracy(),
ggml_opt_result_free(),
ggml_opt_result_init(),
ggml_opt_result_loss(),
ggml_opt_result_ndata(),
ggml_opt_result_pred(),
ggml_opt_result_reset(),
ggml_opt_static_graphs()