diff --git a/README.md b/README.md
index 7faf222..c4b86f4 100644
--- a/README.md
+++ b/README.md
@@ -52,10 +52,9 @@ Use `run_pretraining.py` to pre-train an ELECTRA model. It has the following arg
 If training is halted, re-running the `run_pretraining.py` with the same arguments will continue the training where it left off.
 
 You can continue pre-training from the released ELECTRA checkpoints by
-1. Setting the model-name to point to a downloaded model (e.g., `--model-name electra_small` if you downloaded weights to `$DATA_DIR/electra_small`).
-2. Setting `num_train_steps` by (for example) adding `"num_train_steps": 4010000` to the `--hparams`. This will continue training the small model for 10000 more steps (it has already been trained for 4e6 steps).
-3. Increase the learning rate to account for the linear learning rate decay. For example, to start with a learning rate of 2e-4 you should set the `learning_rate` hparam to 2e-4 * (4e6 + 10000) / 10000.
-4. For ELECTRA-Small, you also need to specifiy `"generator_hidden_size": 1.0` in the `hparams` because we did not use a small generator for that model.
+1. Setting the model-name to point to a downloaded model (e.g., `--model-name electra_small`).
+2. Specify an initial checkpoint (usually from a released ELECTRA checkpoint, e.g., `--init_checkpoint pretrained_models_dir/electra_small`)
+3. For ELECTRA-Small, you also need to specifiy `"generator_hidden_size": 1.0` in the `hparams` because we did not use a small generator for that model.
 
 ##  Quickstart: Pre-train a small ELECTRA model.
 These instructions pre-train a small ELECTRA model (12 layers, 256 hidden size). Unfortunately, the data we used in the paper is not publicly available, so we will use the [OpenWebTextCorpus](https://skylion007.github.io/OpenWebTextCorpus/) released by Aaron Gokaslan and Vanya Cohen instead. The fully-trained model (~4 days on a v100 GPU) should perform roughly in between [GPT](https://s3-us-west-2.amazonaws.com/openai-assets/research-covers/language-unsupervised/language_understanding_paper.pdf) and BERT-Base in terms of GLUE performance. By default the model is trained on length-128 sequences, so it is not suitable for running on question answering. See the "expected results" section below for more details on model performance.
diff --git a/configure_pretraining.py b/configure_pretraining.py
index f576563..fb402e7 100644
--- a/configure_pretraining.py
+++ b/configure_pretraining.py
@@ -44,6 +44,7 @@ def __init__(self, model_name, data_dir, **kwargs):
     self.num_warmup_steps = 10000
 
     # training settings
+    self.init_checkpoint = None
     self.iterations_per_loop = 200
     self.save_checkpoints_steps = 1000
     self.num_train_steps = 1000000
diff --git a/run_pretraining.py b/run_pretraining.py
index 599c9f9..039d884 100644
--- a/run_pretraining.py
+++ b/run_pretraining.py
@@ -272,6 +272,35 @@ def model_fn(features, labels, mode, params):
     model = PretrainingModel(config, features,
                              mode == tf.estimator.ModeKeys.TRAIN)
     utils.log("Model is built!")
+
+    # Load pre-trained weights from checkpoint
+    tvars = tf.trainable_variables()
+
+    init_checkpoint = tf.train.latest_checkpoint(config.init_checkpoint)
+    utils.log("Using checkpoint", init_checkpoint)
+    tvars = tf.trainable_variables()
+
+    initialized_variable_names = {}
+    scaffold_fn = None
+    if init_checkpoint:
+      assignment_map, initialized_variable_names = modeling.get_assignment_map_from_checkpoint(
+          tvars, init_checkpoint)
+      if config.use_tpu:
+        def tpu_scaffold():
+          tf.train.init_from_checkpoint(init_checkpoint, assignment_map)
+          return tf.train.Scaffold()
+        scaffold_fn = tpu_scaffold
+      else:
+        tf.train.init_from_checkpoint(init_checkpoint, assignment_map)
+
+    utils.log("**** Trainable Variables ****")
+    for var in tvars:
+      init_string = ""
+      if var.name in initialized_variable_names:
+        init_string = ", *INIT_FROM_CKPT*"
+      utils.log("  name = %s, shape = %s%s", var.name, var.shape,
+                      init_string)
+
     if mode == tf.estimator.ModeKeys.TRAIN:
       train_op = optimization.create_optimizer(
           model.total_loss, config.learning_rate, config.num_train_steps,
@@ -284,6 +313,7 @@ def model_fn(features, labels, mode, params):
           mode=mode,
           loss=model.total_loss,
           train_op=train_op,
+          scaffold_fn=scaffold_fn,
           training_hooks=[training_utils.ETAHook(
               {} if config.use_tpu else dict(loss=model.total_loss),
               config.num_train_steps, config.iterations_per_loop,
@@ -293,6 +323,7 @@ def model_fn(features, labels, mode, params):
       output_spec = tf.estimator.tpu.TPUEstimatorSpec(
           mode=mode,
           loss=model.total_loss,
+          scaffold_fn=scaffold_fn,
           eval_metrics=model.eval_metrics,
           evaluation_hooks=[training_utils.ETAHook(
               {} if config.use_tpu else dict(loss=model.total_loss),