CLIP · ViT-L/14 · Foundation Model
Text → Image Retrieval
Prompt-aware retrieval: tokenize → text encoder → semantic pre-filter → FAISS cosine similarity → diversity re-rank → XAI.
⌘K
GPU78%
Loss0.184
Throughput312/s
Execution pipeline
- 1Tokenize prompt
- 2Text encoder (Transformer)
- 3Candidate selection (keyword pre-filter)
- 4Query FAISS IVF-PQ index
- 5Cosine similarity + diversity re-rank
- 6Explainability re-ranker
- 7Metric evaluation & report
Top-K
Enter a prompt and run retrieval — each run produces a new ranking.