Dan Nelson
Dan Nelson
Adding a link to a riffusion demo we have up and running on Replicate: https://replicate.com/hmartiro/riffusion
Hey! Added a `cog.yaml` and `predict.py` file so that we have the code that powers the riffusion demo on Replicate (https://replicate.com/hmartiro/riffusion) colocated with the riffusion source inference code.
Hey @tloen! This is some really great work. Here's a PR that enables containerization of training & inference w/Cog. Should be as simple as `cog run finetune.py / cog predict...
Hey! Really, really cool work you've done here with miniGPT-4. It really represents a substantial leap in capabilities for multi-modal models. I added a demo of minigpt-4 to Replicate. This...
Hey folks! ImageBind is very clever work, really like using images as a way to bind embeddings across modalities to a common space. This PR adds some files (`predict.py` and...
Hey @albarji ! 👋 Really great work with mixture of diffusers! I like the approach you took to composing a scene. This pull request has a link to a demo...
Goal here is to add actions for running tests (rn just a smoke test) when we push a new version of a model, as well as pushing a model to...
Not in mergeable state yet (need to pull out some WIP axolotl stuff) but working
This adds a few features to fine-tuning: * Saves checkpoints every n_epochs/steps * Critically, patches the HF trainer s.t. deepspeed only saves the peft weights, instead of the entire 13/26...