Improvements from v1: > 2x high-quality training data vs DeepSeek-VL1
> Dynamic image tiling for flexible resolutions + efficient DeepSeek-MoE for LM
> 3-stage training + new multi-modal parallel strategies for efficiency
OPEN SOURCE
-
DeepSeek-VL2: Enhanced Multimodal Model with Dynamic Tiling
By
–
-
DeepSeek-VL2 Achieves SoTA Vision Performance with Fewer Parameters
By
–
The whale strikes again! DeepSeekVL 2 > DeepSeek-VL2-Tiny, DeepSeek-VL2-Small, and DeepSeek-VL2, with 1.0B, 2.8B, and 4.5B activated parameters
> SoTA perf with similar or fewer activated parameters compared to Qwen 2 VL
> Excels at visual question answering, optical -
Open AI ML Advances Span Multiple Modalities and Research Labs
By
–
It's been an ABSOLUTELY smashing last couple of weeks – across modalities, sizes and research labs! There's no stopping the open AI/ ML train!
-
Signal Denies Vulnerability Rumor Reiterates Open Source Code
By
–
PSA: we've addressed this rumor repeatedly. Tldr there is no evidence of a vulnerability in Signal, and our code is open source.
— Meredith Whittaker (@mer__edith) 13 décembre 2024
No report of this issue was ever made to security@signal.org, and no CVE was filed. There is no truth to the assertion that a phone number enables… https://t.co/YoFMpFMUitPSA: we've addressed this rumor repeatedly. Tldr there is no evidence of a vulnerability in Signal, and our code is open source. No report of this issue was ever made to security@signal.org, and no CVE was filed. There is no truth to the assertion that a phone number enables
-
Signal Addresses Security Vulnerability Rumor with Open Source Code
By
–
PSA: we've addressed this rumor repeatedly. Tldr there is no evidence of a vulnerability in Signal, and our code is open source. No report of this issue was ever made to security@signal.org, and no CVE was filed. There is no truth to the assertion that a phone number enables
-
Model Release Support: Hugging Face Weights Distribution
By
–
Congrats on the release! – let me know if you need any help with putting the weights on Hugging Face!
-
Meta FAIR Releases Nine Open Source AI Research Artifacts
By
–
Wrapping up the year and coinciding with #NeurIPS2024, today at Meta FAIR we’re releasing a collection of nine new open source AI research artifacts across our work in developing agents, robustness & safety and new architectures. More in the video from @jpineau1
. All of this -
Cultural institutions must share content freely for AI training
By
–
Every institution, library, foundation, cultural group, and government around the world that possesses cultural content should make it available for training ***free and open*** AI foundation models.
Links to full posts -
Keith Witek Discusses Open Source at Future of Computing Conference
By
–
COO Keith Witek on all things #opensource at the Future of Computing Conference at @TU_Muenchen this week.
-
SambaNova’s DataComp-LM Advances LLM Pretraining Data Curation
By
–
SambaNova researcher @etash_guha is presenting DataComp-LM at @NeurIPSConf
! This paper finds a pretraining data curation pipeline for LLMs that surpasses existing open source models. Please reach out to him to discuss the DCLM paper! Paper here