CLIP

Classifies images into arbitrary categories without training by encoding images and text descriptions into a shared embedding space, then computing cosine similarity between image and text embeddings. The dual-encoder architecture (separate image and text encoders) projects both modalities into the same vector space where semantically related concepts cluster together, enabling direct comparison without fine-tuning on target classes.

Computes semantic similarity between images and text by encoding both into a 512-dimensional (or larger, depending on model variant) shared embedding space using separate image and text encoders, then calculating cosine similarity between the resulting vectors. The contrastive training objective aligns related image-text pairs close together in this space while pushing unrelated pairs apart, enabling ranking and matching tasks.

Tokenizes text strings using a custom byte-pair encoding (BPE) tokenizer with a 49,152-token vocabulary trained on the pre-training corpus. The tokenizer is accessed via clip.tokenize(text) and converts text to token IDs, automatically padding or truncating to a fixed context length of 77 tokens. The tokenizer handles special tokens (start-of-sequence, end-of-sequence, padding) and produces integer token tensors suitable for the text encoder.

Extracts images into fixed-size embedding vectors (512 to 768 dimensions depending on model variant) by passing images through the image encoder (either a modified ResNet or Vision Transformer backbone) and projecting the output into the shared embedding space. These embeddings can be stored, indexed, and used for downstream tasks like clustering, retrieval, or as input to other models.

Converts text strings into fixed-size embedding vectors (512 to 768 dimensions) by first tokenizing text using a byte-pair encoding (BPE) tokenizer with a 49,152-token vocabulary, then passing tokenized sequences through a Transformer encoder with causal attention masking, and finally projecting the output into the shared embedding space. The tokenizer handles arbitrary text up to 77 tokens (context length) and pads or truncates as needed.

Provides 9 pre-trained model variants with different architectural choices (ResNet-50/101/50x4/50x16/50x64 or Vision Transformer B/32, B/16, L/14, L/14@336px) and parameter counts (50M to 400M), allowing users to select based on accuracy-speed-memory trade-offs. Models are loaded via clip.load(model_name) which downloads from OpenAI's Azure endpoint, caches locally, and returns the model plus preprocessing transform. Each variant has different input image sizes (224×224 to 448×448) and embedding dimensions.

Processes multiple images or text samples in batches through the model with automatic GPU/CPU device placement and optional JIT compilation for faster inference. The clip.load() function accepts a 'device' parameter (e.g., 'cuda', 'cpu') and a 'jit' boolean flag that compiles the model to TorchScript for optimized execution. Batch processing is significantly faster than single-sample inference due to GPU parallelization and reduced overhead.

Provides two distinct image encoder architectures: Vision Transformers (ViT-B/32, ViT-B/16, ViT-L/14, ViT-L/14@336px) that divide images into patches and process them with self-attention, and modified ResNets (RN50, RN101, RN50x4, RN50x16, RN50x64) that use convolutional layers with additional attention mechanisms. Both architectures are trained end-to-end with the text encoder using contrastive loss, and the choice affects accuracy, speed, and memory trade-offs.

Implements a contrastive pre-training objective where image-text pairs from the training corpus are pulled together in embedding space while negative pairs (unrelated images and text) are pushed apart. The loss function computes similarity between all image-text pairs in a batch, creating a symmetric contrastive objective that aligns both modalities. This training approach enables the learned embeddings to capture semantic relationships without explicit labels for downstream tasks.

Applies model-specific image preprocessing including resizing to the correct input dimensions (224×224 for most variants, 336×336 for ViT-L/14@336px, 448×448 for RN50x64), center cropping, conversion to tensors, and normalization using ImageNet statistics (mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]). The clip.load() function returns a preprocessing transform (torchvision.transforms.Compose) that encapsulates these operations, ensuring consistency with training-time preprocessing.

Provides clip.available_models() function that returns a list of all available pre-trained model names, and clip.load() automatically downloads models from OpenAI's Azure endpoint on first use, caches them locally in ~/.cache/clip/, and loads from cache on subsequent calls. This enables users to discover available models, automatically manage model downloads, and avoid re-downloading large model files.