diff --git a/CHANGELOG.md b/CHANGELOG.md
index 1cb2e35..7420687 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -1,5 +1,6 @@
 ## 0.1.1 (unreleased)
 
+- Added disk-level training
 - Added more metrics
 
 ## 0.1.0 (2021-07-26)
diff --git a/README.md b/README.md
index 7df5124..38d77cb 100644
--- a/README.md
+++ b/README.md
@@ -160,6 +160,24 @@ Calculate AUC (for one-class MF)
 model.auc(&data, transpose);
 ```
 
+## Disk-Level Training
+
+Train directly from files
+
+```ruby
+model.fit_disk("train.txt")
+model.fit_eval_disk("train.txt", "validate.txt")
+model.cv_disk("train.txt")
+```
+
+Data should be in the format `row_index column_index value`:
+
+```txt
+0 0 5.0
+0 2 3.5
+1 1 4.0
+```
+
 ## Reference
 
 Specify the initial capacity for a matrix
diff --git a/src/bindings.rs b/src/bindings.rs
index a5df017..438a6ec 100644
--- a/src/bindings.rs
+++ b/src/bindings.rs
@@ -55,8 +55,11 @@ extern "C" {
     pub fn mf_save_model(model: *const MfModel, path: *const c_char) -> c_int;
     pub fn mf_load_model(path: *const c_char) -> *mut MfModel;
     pub fn mf_train(prob: *const MfProblem, param: MfParameter) -> *mut MfModel;
+    pub fn mf_train_on_disk(tr_path: *const c_char, param: MfParameter) -> *mut MfModel;
     pub fn mf_train_with_validation(tr: *const MfProblem, va: *const MfProblem, param: MfParameter) -> *mut MfModel;
+    pub fn mf_train_with_validation_on_disk(tr_path: *const c_char, va_path: *const c_char, param: MfParameter) -> *mut MfModel;
     pub fn mf_cross_validation(prob: *const MfProblem, nr_folds: c_int, param: MfParameter) -> c_double;
+    pub fn mf_cross_validation_on_disk(prob: *const c_char, nr_folds: c_int, param: MfParameter) -> c_double;
     pub fn mf_predict(model: *const MfModel, u: c_int, v: c_int) -> c_float;
     pub fn calc_rmse(prob: *const MfProblem, model: *const MfModel) -> c_double;
     pub fn calc_mae(prob: *const MfProblem, model: *const MfModel) -> c_double;
diff --git a/src/model.rs b/src/model.rs
index 495e7d4..3fe9190 100644
--- a/src/model.rs
+++ b/src/model.rs
@@ -35,17 +35,33 @@ impl Model {
         self.model = unsafe { mf_train(&prob, self.param()) };
     }
 
+    pub fn fit_disk(&mut self, path: &str) {
+        let cpath = CString::new(path).expect("CString::new failed");
+        self.model = unsafe { mf_train_on_disk(cpath.as_ptr(), self.param()) };
+    }
+
     pub fn fit_eval(&mut self, train_set: &Matrix, eval_set: &Matrix) {
         let tr = train_set.to_problem();
         let va = eval_set.to_problem();
         self.model = unsafe { mf_train_with_validation(&tr, &va, self.param()) };
     }
 
+    pub fn fit_eval_disk(&mut self, train_path: &str, eval_path: &str) {
+        let trpath = CString::new(train_path).expect("CString::new failed");
+        let vapath = CString::new(eval_path).expect("CString::new failed");
+        self.model = unsafe { mf_train_with_validation_on_disk(trpath.as_ptr(), vapath.as_ptr(), self.param()) };
+    }
+
     pub fn cv(&mut self, data: &Matrix, folds: i32) {
         let prob = data.to_problem();
         unsafe { mf_cross_validation(&prob, folds, self.param()); }
     }
 
+    pub fn cv_disk(&mut self, path: &str, folds: i32) {
+        let cpath = CString::new(path).expect("CString::new failed");
+        unsafe { mf_cross_validation_on_disk(cpath.as_ptr(), folds, self.param()); }
+    }
+
     pub fn predict(&self, row_index: i32, column_index: i32) -> f32 {
         assert!(self.is_fit());
         unsafe { mf_predict(self.model, row_index, column_index) }
@@ -196,6 +212,9 @@ impl Model {
 mod tests {
     use crate::{Matrix, Model};
 
+    const TRAIN_PATH: &str = "vendor/libmf/demo/real_matrix.tr.txt";
+    const EVAL_PATH: &str = "vendor/libmf/demo/real_matrix.te.txt";
+
     fn generate_data() -> Matrix {
         let mut data = Matrix::new();
         data.push(0, 0, 1.0);
@@ -217,6 +236,13 @@ mod tests {
         model.bias();
     }
 
+    #[test]
+    fn test_fit_disk() {
+        let mut model = Model::new();
+        model.quiet = true;
+        model.fit_disk(TRAIN_PATH);
+    }
+
     #[test]
     fn test_fit_eval() {
         let data = generate_data();
@@ -225,6 +251,13 @@ mod tests {
         model.fit_eval(&data, &data);
     }
 
+    #[test]
+    fn test_fit_eval_disk() {
+        let mut model = Model::new();
+        model.quiet = true;
+        model.fit_eval_disk(TRAIN_PATH, EVAL_PATH);
+    }
+
     #[test]
     fn test_cv() {
         let data = generate_data();
@@ -233,6 +266,13 @@ mod tests {
         model.cv(&data, 5);
     }
 
+    #[test]
+    fn test_cv_disk() {
+        let mut model = Model::new();
+        model.quiet = true;
+        model.cv_disk(TRAIN_PATH, 5);
+    }
+
     #[test]
     fn test_save_load() {
         let data = generate_data();