# Continuous training of google-bert/bert-base-uncased

**URL:** </static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fcontinuous-training-of-google-bert-bert-base-uncased%2F95679%3E%5C
**Category:** Beginners\
**Created:** [July 5, 2024, 9:40am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fcontinuous-training-of-google-bert-bert-base-uncased%2F95679 "2024-07-05T09:40:58Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![Jairnetojp](https://avatars.discourse-cdn.com/v4/letter/j/71c47a/32.png) [@Jairnetojp](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2FJairnetojp)%5C
**Post date:** [July 5, 2024, 9:40am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fcontinuous-training-of-google-bert-bert-base-uncased%2F95679%2F1 "2024-07-05T09:40:58Z")

</div>

Hi everyone!

I’m fine-tuning the `google-bert/bert-base-multilingual-cased` model for content moderation classification, using `AutoModelForSequenceClassification` with the problem set as `multi_label_classification`.

In the initial training, I used 500k messages for training and 100k for validation. This model will go into production, and all blocked messages will be moderated by a team, allowing us to collect corrected data daily.

I have some questions about best practices for continuous training:

1. **Continuous training with new data:**

2. **Training from scratch with new and old data:**

3. **Minimum data quantity:**

I appreciate any help or guidance in advance!

Thank you!

---

<div class="post-metadata">

**Author:** ![vagavr](https://avatars.discourse-cdn.com/v4/letter/v/7feea3/32.png) [@vagavr](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fvagavr)%5C
**Post date:** [January 13, 2025, 10:01am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fcontinuous-training-of-google-bert-bert-base-uncased%2F95679%2F2 "2025-01-13T10:01:30Z")

</div>

I have the same problem, so would appreciate any updates on this topic