2 रिपॉजिटरी
Generates long video sequences by processing overlapping frame windows to produce high-frame-count outputs with limited VRAM.
Distinct from Video Frame Processing: Distinct from Video Frame Processing: extends video length via overlapping windows rather than decoding or adjusting individual frames.
Explore 2 awesome GitHub repositories matching graphics & multimedia · Context Window Extenders. Refine with filters or upvote what's useful.
I've made everythign less reliant on torch.compile for VRAM efficiency, so things should work better even without it. Also figured workaround for some issues when using compile that made first run use drastically more VRAM, issue I battled with myself a lot.
Implements context-window video extension that processes overlapping frame windows to generate long sequences under VRAM constraints.
Baichuan-7B is an open-source 7 billion parameter bilingual Transformer model designed for text generation and few-shot learning across Chinese and English. It is built on a large Transformer architecture trained on a bilingual corpus, enabling it to produce coherent text in both languages from a single model. The model incorporates several optimization techniques that distinguish it from standard large language models. It uses rotary position embeddings that can extrapolate to longer sequences than seen during training, allowing context extension beyond the original 4096-token training lengt
Extends context beyond training length by using rotary position embeddings that generalize to longer sequences.