Pinned
Manu Romero
45.8K posts
CSO/Co-founder @maisaAI_. Head Contrib/ Ambassador🤗 @huggingface. Research 🌸@bigsciencew/@BigCodeProject | ex @narrativaAI
- I fine-tuned @BigscienceW 🌸 BLOOM (7B) on Standford's #Alpaca 🦙 dataset using @huggingface PEFT (LoRA), and the results are awesome! Any name suggestion for the model?
- 📒Colab for: Fine-tuning a deep-learning model (bert-base-cased) with @huggingface 🤗Datasets library and @PyTorch (included in the quick start guide) on glue / mrpc
- Data augmentation in #NLP is a very insteresting research field. I created this colab colab.research.google.com/drive/1oOHBPZ7… where @tobias_sterbak shows Data augmentation with (@huggingface) transformer models for #NER
- Replying to @IzquierdaUnidano os dais cuenta de que estáis haciendo el ridículo felicitando a un dictador en pleno S. XXI??
- It seems 🤖Reinforcement learning👾 is back! Very good resources for getting started on it from @huggingface and @ThomasSimonini
- Many of you asked me for a Colab Notebook 📒 to train a model to translate from SQL 💽 to Natural Language 🗣️ (and vice versa). Tomorrow I will release it using Google's T5 and @huggingface ecosystem. #nlp #nlg #seq2seq #tuto
- 🧵When using #LoRA it is important to apply it to ALL `Linear` layers of the model to get similar results to "full fine-tuning." If you are using @huggingface/peft library, you will see some LoRA configs like the following
- I fine-tuned #Falcon 🦅7B (huggingface.co/tiiuae/falcon-…) on a Spanish instruction dataset using #QLoRA (4bit), and the results are promising! Thanks to @younesbelkada for the support!
- I fine-tuned @BigscienceW 🌸 BLOOM (7B) on Standford's #Alpaca 🦙 dataset using @huggingface PEFT (LoRA), and the results are awesome! Any name suggestion for the model?







