train-sentence-transformers

You can train sentence-transformers models for retrieval and similarity by automating the fine-tuning process with this Skill. It handles loss selection, hard-negative mining, and evaluation to produce optimized text embeddings.

Fine-tuning a sentence-transformers model for classification is best handled by selecting the appropriate model architecture and loss function. This Skill provides instructions for choosing the right configuration to optimize your classification embeddings.

Yes, you can publish trained sentence-transformers models to the Hugging Face Hub. This Skill supports direct Hub publishing using the huggingface_hub library to share your fine-tuned embeddings.

Yes, you need the datasets library to load training data for sentence-transformers models. This Skill requires datasets and huggingface_hub dependencies to automate paraphrase mining and deduplication tasks.

Knowledge distillation for sentence-transformers transfers knowledge from a larger model to a smaller one during training. This Skill supports distillation alongside hard-negative mining and evaluators to improve embedding accuracy.

Yes, this supports multimodal applications and sparse embeddings for reranking. It provides instructions for selecting appropriate model architectures to handle diverse data inputs and optimize reranking pipelines.