Merge pull request #50 from Nixtla/feat/finetune_depth

feat: add finetune_depth parameter

R/nixtla_client_cross_validation.R

@@ -11,8 +11,9 @@
 #' @param quantiles Quantiles to forecast. Should be between 0 and 1.
 #' @param n_windows Number of windows to evaluate.
 #' @param step_size Step size between each cross validation window. If NULL, it will equal the forecast horizon (h).
-#' @param finetune_steps Number of steps used to finetune 'TimeGPT' in the new data.
-#' @param finetune_loss Loss function to use for finetuning. Options are: "default", "mae", "mse", "rmse", "mape", and "smape".
+#' @param finetune_steps Number of steps used to fine-tune 'TimeGPT' in the new data.
+#' @param finetune_depth The depth of the fine-tuning. Uses a scale from 1 to 5, where 1 means little fine-tuning and 5 means that the entire model is fine-tuned.
+#' @param finetune_loss Loss function to use for fine-tuning. Options are: "default", "mae", "mse", "rmse", "mape", and "smape".
 #' @param clean_ex_first Clean exogenous signal before making the forecasts using 'TimeGPT'.
 #' @param model Model to use, either "timegpt-1" or "timegpt-1-long-horizon". Use "timegpt-1-long-horizon" if you want to forecast more than one seasonal period given the frequency of the data.
 #'
@@ -27,7 +28,7 @@
 #'   fcst <- nixtlar::nixtla_client_cross_validation(df, h = 8, id_col = "unique_id", n_windows = 5)
 #' }
 #'
-nixtla_client_cross_validation <- function(df, h=8, freq=NULL, id_col="unique_id", time_col="ds", target_col="y", level=NULL, quantiles=NULL, n_windows=1, step_size=NULL, finetune_steps=0, finetune_loss="default", clean_ex_first=TRUE, model="timegpt-1"){
+nixtla_client_cross_validation <- function(df, h=8, freq=NULL, id_col="unique_id", time_col="ds", target_col="y", level=NULL, quantiles=NULL, n_windows=1, step_size=NULL, finetune_steps=0, finetune_depth=1, finetune_loss="default", clean_ex_first=TRUE, model="timegpt-1"){
 
   # Validate input ----
   if(!is.data.frame(df) & !inherits(df, "tbl_df") & !inherits(df, "tsibble")){
@@ -130,6 +131,7 @@ nixtla_client_cross_validation <- function(df, h=8, freq=NULL, id_col="unique_id
     freq = freq,
     clean_ex_first = clean_ex_first,
     finetune_steps = finetune_steps,
+    finetune_depth = finetune_depth,
     finetune_loss = finetune_loss
   )
 

R/nixtla_client_forecast.R

@@ -10,8 +10,9 @@
 #' @param X_df A tsibble or a data frame with future exogenous variables.
 #' @param level The confidence levels (0-100) for the prediction intervals.
 #' @param quantiles Quantiles to forecast. Should be between 0 and 1.
-#' @param finetune_steps Number of steps used to finetune 'TimeGPT' in the new data.
-#' @param finetune_loss Loss function to use for finetuning. Options are: "default", "mae", "mse", "rmse", "mape", and "smape".
+#' @param finetune_steps Number of steps used to fine-tune 'TimeGPT' in the new data.
+#' @param finetune_depth The depth of the fine-tuning. Uses a scale from 1 to 5, where 1 means little fine-tuning and 5 means that the entire model is fine-tuned.
+#' @param finetune_loss Loss function to use for fine-tuning. Options are: "default", "mae", "mse", "rmse", "mape", and "smape".
 #' @param clean_ex_first Clean exogenous signal before making the forecasts using 'TimeGPT'.
 #' @param add_history Return fitted values of the model.
 #' @param model Model to use, either "timegpt-1" or "timegpt-1-long-horizon". Use "timegpt-1-long-horizon" if you want to forecast more than one seasonal period given the frequency of the data.
@@ -27,7 +28,7 @@
 #'   fcst <- nixtlar::nixtla_client_forecast(df, h=8, id_col="unique_id", level=c(80,95))
 #' }
 #'
-nixtla_client_forecast <- function(df, h=8, freq=NULL, id_col="unique_id", time_col="ds", target_col="y", X_df=NULL, level=NULL, quantiles=NULL, finetune_steps=0, finetune_loss="default", clean_ex_first=TRUE, add_history=FALSE, model="timegpt-1"){
+nixtla_client_forecast <- function(df, h=8, freq=NULL, id_col="unique_id", time_col="ds", target_col="y", X_df=NULL, level=NULL, quantiles=NULL, finetune_steps=0, finetune_depth=1, finetune_loss="default", clean_ex_first=TRUE, add_history=FALSE, model="timegpt-1"){
 
   # Validate input ----
   if(!is.data.frame(df) & !inherits(df, "tbl_df") & !inherits(df, "tsibble")){
@@ -116,6 +117,7 @@ nixtla_client_forecast <- function(df, h=8, freq=NULL, id_col="unique_id", time_
     freq = freq,
     clean_ex_first = clean_ex_first,
     finetune_steps = finetune_steps,
+    finetune_depth = finetune_depth,
     finetune_loss = finetune_loss
   )
 
@@ -254,7 +256,20 @@ nixtla_client_forecast <- function(df, h=8, freq=NULL, id_col="unique_id", time_
 
   # Add fitted values if required ----
   if(add_history){
-    fitted <- nixtla_client_historic(df=df, freq=freq, id_col=id_col, time_col=time_col, target_col=target_col, level=level, quantiles=quantiles, finetune_steps=finetune_steps, finetune_loss=finetune_loss, clean_ex_first=clean_ex_first)
+    fitted <- nixtla_client_historic(
+      df=df,
+      freq=freq,
+      id_col=id_col,
+      time_col=time_col,
+      target_col=target_col,
+      level=level,
+      quantiles=quantiles,
+      finetune_steps=finetune_steps,
+      finetune_depth=finetune_depth,
+      finetune_loss=finetune_loss,
+      clean_ex_first=clean_ex_first
+      )
+
     forecast <- dplyr::bind_rows(fitted, forecast)
   }
 

R/nixtla_client_historic.R

@@ -8,8 +8,9 @@
 #' @param target_col Column that contains the target variable.
 #' @param level The confidence levels (0-100) for the prediction intervals.
 #' @param quantiles Quantiles to forecast. Should be between 0 and 1.
-#' @param finetune_steps Number of steps used to finetune 'TimeGPT' in the new data.
-#' @param finetune_loss Loss function to use for finetuning. Options are: "default", "mae", "mse", "rmse", "mape", and "smape".
+#' @param finetune_steps Number of steps used to fine-tune 'TimeGPT' in the new data.
+#' @param finetune_depth The depth of the fine-tuning. Uses a scale from 1 to 5, where 1 means little fine-tuning and 5 means that the entire model is fine-tuned.
+#' @param finetune_loss Loss function to use for fine-tuning. Options are: "default", "mae", "mse", "rmse", "mape", and "smape".
 #' @param clean_ex_first Clean exogenous signal before making the forecasts using 'TimeGPT'.
 #' @param model Model to use, either "timegpt-1" or "timegpt-1-long-horizon". Use "timegpt-1-long-horizon" if you want to forecast more than one seasonal period given the frequency of the data.
 #'
@@ -24,7 +25,7 @@
 #'   fcst <- nixtlar::nixtla_client_historic(df, id_col="unique_id", level=c(80,95))
 #' }
 #'
-nixtla_client_historic <- function(df, freq=NULL, id_col=NULL, time_col="ds", target_col="y", level=NULL, quantiles=NULL, finetune_steps=0, finetune_loss="default", clean_ex_first=TRUE, model="timegpt-1"){
+nixtla_client_historic <- function(df, freq=NULL, id_col=NULL, time_col="ds", target_col="y", level=NULL, quantiles=NULL, finetune_steps=0, finetune_depth=1, finetune_loss="default", clean_ex_first=TRUE, model="timegpt-1"){
 
   # Validate input ----
   if(!is.data.frame(df) & !inherits(df, "tbl_df") & !inherits(df, "tsibble")){

tests/mocks/api.nixtla.io/model_params-3f263f-POST.json

@@ -3,5 +3,5 @@
         "input_size": 120,
         "horizon": 24
     },
-    "request_id": "6UZDCP8S3J"
+    "request_id": "TFKG7KC8EP"
 }

tests/mocks/api.nixtla.io/model_params-9255c1-POST.json

@@ -3,5 +3,5 @@
         "input_size": 840,
         "horizon": 168
     },
-    "request_id": "YDX6MPGAUH"
+    "request_id": "MBSFLHXE2U"
 }