LLM optimization

The embedding layer’s parameter count is determined by a simple formula:

Embedding parameters = vocabulary_size × embedding_dimension

For Gemma 3 270M:

• Vocabulary size: ~256,000 tokens (Gemma uses a very large multilingual vocabulary)
• Embedding dimension: 640

LMCache

The first open-source Knowledge Delivery Network (KDN) that accelerates LLM applications up to 8x faster, at 8x lower cost.

• https://lmcache.ai/

• https://gaurigupta19.github.io/llms/distributed%20ml/optimization/2025/10/02/efficient-ml.html

• https://blog.cloudflare.com/how-cloudflare-runs-more-ai-models-on-fewer-gpus/

• https://jax-ml.github.io/scaling-book/

• https://blog.cloudflare.com/cloudflares-most-efficient-ai-inference-engine/oinf