• multimodal — A model that works with more than text — reading images, audio, or video, and sometimes generating them too.

If you're wrangling multitask, multimodal training pipelines, OFASys reduces the boilerplate of setting up heterogeneous tasks and modalities into a declarative interface, tackling the batchification and training-stability headaches that usually derail generalist-model experiments.

• articleChinese CLIP: Contrastive Vision-Language Pretraining in Chinese
• articleQwen1.5-MoE: Matching 7B Model Performance with 1/3 Activated Parameters
• articleGeneralizing an LLM from 8k to 1M Context using Qwen-Agent
• articleQwen2-Audio: Chat with Your Voice!

• articleMix Data Or Merge Models Optimizing For Diverse Multi Task Learning 2024 10 15Cohere
• articleMultimodality and Large Multimodal Models (LMMs)Chip Huyen
• clipEnd-to-end omnimodal removes the text bottleneck but is still half duplexJulia Turc

Checking sign-in…

Loading comments…