Entdecken

30 technologies shown

• Ollama tool Runs open-weight models locally behind a simple command and HTTP API.
• Claude model Anthropic's hosted model family, positioned around long-context work and safety.
• Colossal Clean Crawled Corpus dataset Heavily filtered Common Crawl used to train T5.
• Elasticsearch database Distributed search engine with vector support alongside full-text search.
• Gradient Descent technique Updating weights in the direction that reduces the loss.
• LMArena benchmark Public arena where anonymous model pairs are ranked by human votes.
• MS MARCO dataset Real search queries with passages and answers, the standard retrieval benchmark.
• OpenAI Text Embedding 3 model Hosted embedding models with configurable output dimensions.
• PostgreSQL database General-purpose relational database, and a vector store through an extension.
• Self-Attention concept Weighting every position by every other position in the same sequence.
• TensorFlow framework Google's deep learning framework, oriented toward deployment at scale.
• Whisper model Speech recognition model trained on large-scale weakly supervised audio.
• Chatbot application A conversational interface that answers a user turn by turn.
• Deep Learning concept Machine learning using many-layered neural networks to learn representations from raw data.
• Fine-tuning technique Continue training a pre-trained model on task-specific data to adapt its behaviour.
• Hybrid Search technique Combining keyword and vector retrieval, then merging the two rankings.
• pgvector library Vector similarity search as a PostgreSQL extension, keeping embeddings beside your data.
• Bi-Encoder concept Model that embeds queries and passages separately so they can be compared cheaply.
• CLIP model Model that places images and text in one embedding space.
• DeepSeek model Open-weight model family notable for efficient training at frontier scale.
• FlashAttention technique Attention computed in tiles to avoid materialising the full matrix.
• H200 hardware Refresh of the H100 with substantially more high-bandwidth memory.
• HNSW algorithm Graph index that finds approximate nearest neighbours in logarithmic time.
• Inference concept Running a trained model to produce output, as opposed to training it.
• Instruction Tuning technique Fine-tuning a model on examples of instructions and their responses.
• LAION-5B dataset Five billion image-text pairs collected from the public web.
• MTEB benchmark Benchmark suite ranking embedding models across many tasks.
• Multi-Head Attention concept Running several attention operations in parallel over different projections.
• Multimodal AI application Models that take more than one kind of input, such as text and images together.
• NVLink hardware High-bandwidth interconnect between GPUs, letting them share memory directly.