h2o-llmstudio
h2o-llmstudio copied to clipboard
h2oai
H2O LLM Studio - a framework and no-code GUI for fine-tuning LLMs. Documentation: https://docs.h2o.ai/h2o-llmstudio/
### 🐛 Bug I tried to push a model to HF, but got the following error with instructions to post the bug here :=) ``` q.app script_sources: ['/_f/68bd9c4d-864b-44b4-b1c2-3f3f9b12805c/tmpx7ptm808.min.js'] initialized: True...
### 🔧 Proposed code refactoring Currently, model weights are stored in LLm Studio format which is a small wrapper around `AutoModelForCausalLM`. Instead, store model weights in `AutoModelForCausalLM` format, as well...
Currently, we only push the model weights to Huggingface. We could improve this process by adding some of these additional artifacts: - [x] Tokenizer - [ ] LLM Studio CFG...
Would be great to have some FAQs and templates/notebooks for common questions. - [ ] How to generate outputs outside of LLM Studio with trained weights pushed to HF -...
### 🚀 Feature Support tree-like conversation data - i.e. chain of thoughts such as the OASST data provdes. ### Motivation Currently, we only support prompt/output data structures. While one can...
This PR converts pickled configurations into yaml configuration. It introduces a breaking change to the current UI (meaning that old experiments won't work). As the repo is in its initial...
### 🔧 Proposed code refactoring Add the separator tokens as special tokens. Potentially then also add a separate setting to use the separator tokens as stop tokens. We should at...
### 🚀 Feature Add a "download model weights" buttons ### Motivation UI improvement over ssh connection into a remote machine to grab the weights.
There are several potential things to improve experience of the chat window: - Block the chat window if other training runs are active - Make the actual model loading procedure...
While we describe steps to get and load OASST demo data, one useful improvement could be to directly load the data into the GUI by default.