Presents open-source 7B models for million-token video and language understanding via Blockwise RingAttention, setting new benchmarks in retrieval and long video tasks.
We trained our models using TPUv4-1024, which is approximately equivalent to 450 A100s, with a batch size of 8M using FSDP (Facebook,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2024 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
World Model on Million-Length Video And Language With Blockwise RingAttention
Presents open-source 7B models for million-token video and language understanding via Blockwise RingAttention, setting new benchmarks in retrieval and long video tasks.