SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2409.04429 · 2024

VILA-U: a Unified Foundation Model Integrating Visual Understanding and Generation

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 7 functions out of this paper's own repositories and ran 5 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.

RepositoryRoleRan
mit-han-lab/vila-u canonical 5 of 7
FunctionStatusWhere it lives
build_llm_and_tokenizer Ran mit-han-lab/vila-u/vila_u/model/language_model/builder.py
code served (permissive licence) · get_code("7edecde48cb876d8")
context_length_extension Ran mit-han-lab/vila-u/vila_u/model/language_model/builder.py
code served (permissive licence) · get_code("2e470faeeb1e0979")
get_frame_from_vcap Ran mit-han-lab/vila-u/vila_u/mm_utils.py
code served (permissive licence) · get_code("94815e9616152485")
get_model_config Ran mit-han-lab/vila-u/vila_u/model/utils.py
code served (permissive licence) · get_code("11b7b07963f16286")
opencv_extract_frames Ran mit-han-lab/vila-u/vila_u/mm_utils.py
code served (permissive licence) · get_code("c1efca00f5418683")
build_vision_tower Not yet run mit-han-lab/vila-u/vila_u/model/multimodal_encoder/builder.py
code served (permissive licence) · get_code("6ef3dc934bf9933a")
load_image_from_base64 Not yet run mit-han-lab/vila-u/vila_u/mm_utils.py
code served (permissive licence) · get_code("0dde2e782959c0bd")

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

VILA-U is a Unified foundation model that integrates Video, Image, Language understanding and generation. Traditional visual language models (VLMs) use separate modules for understanding and generating visual content, which can lead to misalignment and increased complexity. In contrast, VILA-U employs a single autoregressive next-token prediction framework for both tasks, eliminating the need for additional components like diffusion models. This approach not only simplifies the model but also achieves near state-of-the-art performance in visual language understanding and generation. The success of VILA-U is attributed to two main factors: the unified vision tower that aligns discrete visual tokens with textual inputs during pretraining, which enhances visual perception, and autoregressive image generation can achieve similar quality as diffusion models with high-quality dataset. This allows VILA-U to perform comparably to more complex models using a fully token-based autoregressive framework.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2409.04429")
get_code_for_paper("2409.04429")
have("2409.04429")

Connect an agent — have() is free.