Fine-Tuning Qwen3 on Your Data using Single GPU | Sentiment Analysis for Cryptocurrency Tweets

Subscribe to MLExpert Pro for the upcoming live AI bootcamp (9-11 May): https://www.mlexpert.io/ How much better can LLM become after fine-tuning on custom data? In this video, you'll learn how to fine-tune Qwen3 (0.6B) model on your own dataset. You'll learn how to prepare your data, evaluate the model before training, use LoRA for training and evaluate the performance of the fine-tuned model. Original dataset: https://huggingface.co/datasets/StephanAkkerman/financial-tweets-crypto Qwen 3’s Thinking Blog: https://muellerzr.github.io/til/end_thinking.html AI Bootcamp: https://www.mlexpert.io/ LinkedIn: https://www.linkedin.com/in/venelin-valkov/ Follow me on X: https://twitter.com/venelin_valkov Discord: https://discord.gg/UaNPxVD6tv Subscribe: http://bit.ly/venelin-subscribe GitHub repository: https://github.com/curiousily/AI-Bootcamp 👍 Don't Forget to Like, Comment, and Subscribe for More Tutorials! 00:00 - Welcome 00:58 - Live bootcamp sessions on MLExpert.io 01:48 - Dataset 02:57 - Notebook setup 04:18 - Loading the data and preprocessing 08:21 - Creating HuggingFace datasets (including prompt) 13:21 - Tokenizer 15:36 - Counting tokens 16:55 - Model loading and quantization 19:53 - LoRA configuration 20:30 - Baseline (untrained) model evaluation 25:43 - Training arguments and training 33:35 - Training logs review in Tensorboard 35:47 - Saving and merging the trained model 37:35 - Evaluating the trained model 41:40 - Training on completions only 44:13 - Qwen3 thinking budget 45:58 - Conclusion Join this channel to get access to the perks and support my work: https://www.youtube.com/channel/UCoW_WzQNJVAjxo4osNAxd_g/join #qwen3 #lora #artificialintelligence #llm #finetuning