1 Repo
Awesome GitHub RepositoriesDynamic Depth Batching
Techniques for executing varying numbers of recurrent iterations within a single batch to optimize compute based on input complexity.
Distinct from Batch Sequence Training: Distinct from standard batch processing by allowing variable iteration depths (early exit) per sequence in a batch.
Explore 1 awesome GitHub repository matching artificial intelligence & ml · Dynamic Depth Batching. Refine with filters or upvote what's useful.
Awesome Dynamic Depth Batching GitHub Repositories
Finde die besten Repos mit KI.Wir suchen mit KI nach den am besten passenden Repositories.
- kyegomez/openmythos
kyegomez/OpenMythos
14,176Auf GitHub ansehenOpenMythos is a framework for implementing recurrent large language model architectures. It utilizes recurrent transformer blocks to enable compute-adaptive reasoning and variable processing depth through multiple iterative passes over the same weights. The system features a mixture of experts framework that routes tokens between shared and specialized layers to optimize parameter usage. It also includes parameter-efficient fine-tuning tools using low-rank adaptation modules to modify model behavior with minimal weight updates. The framework covers distributed training pipelines using data p
Implements depth-wise batching to allow early exit for simple inputs, increasing inference throughput.
Pythonaianthropicattention
