Meihua Dang @meihuadang Twitter profile

Last Seen Profiles

@StarSouth46

@JoCowenArch

@yasereba_inoni

@reinimau

@McginnOlivia

@k_witzgall

@S4kVMe_hololive

@liulianjingrua

@highoffluhjaae

@nothingonevents

@symmetrymag

@MissBeaHavin

@liuminribao

@evligiz60125833

@PappasTyro26055

@BscProjectOrg

@samanthamaiden

@acellera

@sarpbegum

@ronisia_mds

@SHAYEXR

@sport24live

@Sayedmaymunah

@GmoSheridan

@sissybutylee

@scp_wboc

@shuji24hr

@Marie_Lukassen

@sekaibunko

@vanneloph

@ri_cc_o_

@strghtoutamag

@lnsathletics

@AndaLookkaew_CN

@dianhuji

@rednosebulldozn

Meihua Dang

@meihuadang

5 months

Excited to share that the checkpoints are online at 🤗 HuggingFace. Feel free to check it out! SDXL-DPO: SD1.5-DPO:

Rafael Rafailov

@rm_rafailov

5 months

Excited to announce DPO has gone multi-modal! New paper out on RLHF for text-to-image diffusion models! We obtain large-scale state of the art results with 70% win rates against Stable Diffusion XL on human evals! Deep dive below 🧵

11

86

486

26

116

575

Meihua Dang

@meihuadang

1 year

Our paper "Sparse Probabilistic Circuits via Pruning and Growing" was selected as Oral at #NeurIPS2022 . Check here , or at poster session 3 #904 . Can we scale up probabilistic circuits simply by increasing model sizes? 1/2

Sparse Probabilistic Circuits via Pruning and Growing

Probabilistic circuits (PCs) are a tractable representation of probability distributions allowing for exact and efficient computation of likelihoods and marginals. There has been significant...

arxiv.org

2

8

55

Meihua Dang

@meihuadang

1 year

We propose a model pruning technique leveraging the probabilistic semantics of parameters; our approach can prune 97% of the model parameters without hurting performance, and based on which we are able to effectively learn large yet sparse probabilistic circuits. 3/3

1

0

4

Meihua Dang

@meihuadang

1 year

We show that PC training performance plateaus as model size increases; most capacity in existing large PC structures is wasted: fully-connected parameter layers are only sparsely used. 2/3

1

0

3

Meihua Dang

@meihuadang

5 months

@SCarbain You are right. We train the model from the paired dataset end-to-end. One benefit of this method is that we don’t require a reward function e.g. PickScore.

1

0

Meihua Dang

@meihuadang

5 months

@_arbitrarity yes

1

0

2

Meihua Dang

@meihuadang

5 months

@HorridWeasell Just realized I had a typo, can you try with `guidance_scale=5`, the default one. Also updated the example code.

1

0

2

Meihua Dang

@meihuadang

1 year

Thanks to the amazing collaborators @liu_anji and @guyvdb !

0

2

Meihua Dang

@meihuadang

5 months

@rishabh063 it's SDXL trained with DPO, paper here .

Diffusion Model Alignment Using Direct Preference Optimization

Large language models (LLMs) are fine-tuned using human comparison data with Reinforcement Learning from Human Feedback (RLHF) methods to make them better aligned with users' preferences. In...

arxiv.org

1

0

3

Meihua Dang

@meihuadang

4 years

@tetraduzione @martin_trapp @guyvdb @2020Pgm As a follow up, PSDD also has strong determinism properties, which is used for logic constraints, but not needed for probabilistic circuits.

0

1

Meihua Dang

@meihuadang

5 months

@SCarbain Will come soon!

3

0

1

Meihua Dang

@meihuadang

1 year

@LucaAmb Thank you! 😃

0

1