# How to continue training a pre-trained model?

**URL:** <https://discuss.ai.google.dev/t/how-to-continue-training-a-pre-trained-model/27833>\
**Category:** General Discussion\
**Tags:** keras, models, help\_request\
**Created:** [August 31, 2021, 10:59pm UTC](https://discuss.ai.google.dev/t/how-to-continue-training-a-pre-trained-model/27833 "2021-08-31T22:59:05Z")\
**Posts on this page:** 4\
**Page:** 1

<div class="post-metadata">

**Author:** ![Ashley](https://avatars.discourse-cdn.com/v4/letter/a/45deac/32.png) [@Ashley](https://discuss.ai.google.dev/u/Ashley)\
**Post date:** [August 31, 2021, 10:59pm UTC](https://discuss.ai.google.dev/t/how-to-continue-training-a-pre-trained-model/27833/1 "2021-08-31T22:59:05Z")

</div>

I’m currently using VGG16 pretrained model on my set of images  
model = VGG16()  
model = Model(inputs = model.inputs, outputs = model.layers[-2].output)  
and then for each image I get its features `model.predict(img, use_multiprocessing=True)`  
but I also would like to take some subset of my images to make a little further training of VGG16 model on that subset (so the final model would be trained on VGG16’s images + on my images)

how can I implement that?

---

<div class="post-metadata">

**Author:** ![Bhack](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/bhack/32/6388_2.png) [@Bhack](https://discuss.ai.google.dev/u/Bhack)\
**Post date:** [August 31, 2021, 11:06pm UTC](https://discuss.ai.google.dev/t/how-to-continue-training-a-pre-trained-model/27833/2 "2021-08-31T23:06:42Z")

</div>

It is very similar to the fine-tuning described in:

> **[Keras documentation: Keras Applications](https://keras.io/api/applications/)**
>
> Keras documentation

---

<div class="post-metadata">

**Author:** ![Bhack](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/bhack/32/6388_2.png) [@Bhack](https://discuss.ai.google.dev/u/Bhack)\
**Post date:** [August 31, 2021, 11:19pm UTC](https://discuss.ai.google.dev/t/how-to-continue-training-a-pre-trained-model/27833/3 "2021-08-31T23:19:46Z")

</div>

For a more extended tutorial take a look also to:

> **[Keras documentation: Transfer learning & fine-tuning](https://keras.io/guides/transfer_learning/)**
>
> Keras documentation

---

<div class="post-metadata">

**Author:** ![lgusm](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/lgusm/32/6230_2.png) [@lgusm](https://discuss.ai.google.dev/u/lgusm)\
**Post date:** [September 2, 2021, 11:38am UTC](https://discuss.ai.google.dev/t/how-to-continue-training-a-pre-trained-model/27833/5 "2021-09-02T11:38:00Z")

</div>

I’d also add this great tutorial: [Retraining an Image Classifier &nbsp;|&nbsp; TensorFlow Hub](https://www.tensorflow.org/hub/tutorials/tf2_image_retraining)

It’s not specific for VGG but might give some insights