---

name: clip_contrastive_loss description: Compute CLIP-style normalized similarity logits and symmetric contrastive loss for image-text batches.

CLIP Contrastive Loss

Use this skill when implementing or checking the CLIP training mechanism on paired image/text embeddings. It is appropriate for full, reduced, or synthetic recovery if the result is clearly labeled.

Inputs

• Equal-length image and text embedding batches.
• A positive logit scale, corresponding to inverse temperature.

Outputs

• Scaled cosine-similarity logits.
• Image-to-text and text-to-image cross entropy.
• Average symmetric loss and top-1 retrieval accuracy.

Workflow

1. Normalize each image and text feature vector.
2. Compute all pairwise dot products and multiply by the logit scale.
3. Use diagonal indices as positives for both image-to-text and text-to-image directions.
4. Average the two cross-entropy losses and report retrieval accuracy.

Validation

Run python tests/test_contrastive_loss.py or the generated skill tree validator.

Limitations

This skill operates on supplied embeddings. It does not implement the visual or text transformer encoders.