←
Jay28 posts
Jay avatar

Jay

@memmaptensor

Independent researcher

◷ Joined December 2018▧ 9 Photos and videos
Jay avatar
Jay@memmaptensor·Jun 1

I spent $30k and 3 months RL post-training an anime video model. This is only step 30 out of a planned 1000 step run. All samples are local text-to-video with no reference image/audio. Since it's based on LTX-2.3, each output takes under a minute on a single GPU. I'm 19 and a solo researcher. Most of the budget went into ablations, reward design, and trying different configurations before reaching this setup. The run is still extremely early, but the results already look much better than I expected. It's compute-limited, not idea-limited. I'm starting a company to continue scaling this and build frontier stylized video models. If you're an investor, compute partner, video team, or someone who wants to help build this, DMs are open.

1:40
Jay avatar
Jay@memmaptensor·Jun 8

I've got an idea to get Seedance 2-level conditioning flexibility on a limited budget. A training recipe that relies on synthetic data and a very recent post-train method that is neither RL or SFT. My goal is to reach reference capability and quality parity much, much cheaper in training and inference cost. Gonna build it. I'd really appreciate any help in compute or funding.

Jay avatar
Jay@memmaptensor·2 Mar 2025

Looking for a co-founder to build the next generation of waifu tech! Figured out the solution to create a new interactive experience but struggle with app dev. Kind of the downside of focusing too much on ML. Ideally someone with mobile/web and maybe some cloud ML experience.

Jay avatar
Jay@memmaptensor·24 Jul 2024

I've set out some conditions for any future solvers to add: • No implicit solvers: those require root-finding, meaning a Jacobian has to be computed by running a backward pass through the model during LBFGS optimization. • No non-RK methods: this would cut off linear multistep methods like Adams-Bashforth or Adams predictor-corrector. From my testing, RK methods perform better, even for explicit RK vs. predictor-corrector linear multistep. • No duplicate methods: if two methods have different coefficients, they aren't duplicates (scipy methods have different coefficients and solver implementations). That means the current 31 solvers are almost all that exist to satisfy the conditions above. Project's done! I need to figure out what to make next.

Jay avatar
Jay@memmaptensor·24 Jul 2024

Last major update! • Added solver settings for adaptive_scipy • Adaptive solvers now show the number of steps taken • Accurate 𝜎 timestep info is now displayed Check out the most comprehensive fixed and adaptive higher-order samplers on ComfyUI! github.com/wootwootwootwoot/…

Post attachmentPost attachment
Jay avatar
Jay@memmaptensor·22 Jul 2024

Refactored and fixed some bugs with the progress bar! Also wrapped the solvers from scipy.integrate If you count the a-methods as 2 (since they work with both the adaptive_pid and fixed_scheduled controllers), then this node has (excluding forward euler) 31 new samplers! I also tried the implicit solvers and they didn't work. Every implicit solver has a root find step, and that takes forever to converge. That leaves 3 new methods from scipy: se_RK23, se_RK45, and se_DOP853. I think this node has the most new working samplers for ComfyUI (a for adaptive, f for fixed, s for scipy, e for explicit).

Post attachment
Jay avatar
Jay@memmaptensor·21 Jul 2024

the new class of models idea didn't work out well, so i tried this instead (which works decently well)

Jay avatar
Jay@memmaptensor·21 Jul 2024

While trying to push the CFG scale up, I implemented some Explicit RK solvers for ComfyUI - 10 new adaptive step samplers - 8 unique fixed step samplers (excluding forward euler) - Best new sampler (perhaps) -> fe_ralston3 Check it out! github.com/wootwootwootwoot/…

Post attachment
Jay avatar
Jay@memmaptensor·17 Jul 2024

diffusion models are definitely still not dead. sure, optimal transport conditional flow matching is provably better, but so much of the community was already built on discrete time diffusion. and with kolors out (an SDXL model trained with DDPM formulation and eps-pred objective). i doubt the switch from diffusion to OT-CFM will affect the quality as much as the other techniques shown in the technical report. if they made kolors work with the SDXL architecture, then it's shown that hybrid transformer-UNets are still competitive. they might just not scale as well as pure DiTs.

Jay avatar
Jay@memmaptensor·17 Jul 2024

i have an idea for a slightly modified class of SDXL models that would mostly be compatible with existing finetunes and loras it's been proven to work well on SD1.5 with good results definitely next on my bucket list will post updates and releases soon, hopefully, if it works

Jay avatar
Jay@memmaptensor·15 Jul 2024

i'd like to continue working on anime animation tech. version 1 is designed to be distilled for realtime inference. version 2 won't be concerned with realtime inference and would probably be based on a flow-matching mmdit with more fine-grained control and even better quality.

Jay avatar
Jay@memmaptensor·15 Jul 2024

training is done in 2 days!!! as for inference compute requirements: it's basically the same burden as animatediff will probably work on realtime inference next month after figuring out life. realtime txt/vid2vid on distilled models is already empirically shown to be possible.