artste

Containerized end-to-end workflow

1

### Summary This Pull Request introduces a containerized environment to seamlessly run the `lm-hackers` notebook from start to finish. It includes two shell scripts: `build.sh` for building the `lmhackers` Docker...

OPENAI: upgrade to openai api v1.0.0

1

This PR focuses on migrate the openai api code to version 1.0.0 . For example, old api uses this to create chat completion: `c = ChatCompletion.create(...` Will become: `c =...

This PR fixes two issues in the `sql.yml` configuration file for axolotl: 1. **Error**: `ValueError: eval_table_size and eval_sample_packing are not supported together with sample_packing. Please set 'eval_sample_packing' to false.` -...

This PR is meant to add `environment.yml` with the specifications of complete conda environment needed to run the notebook. NOTE: this will not contains axolotl/peft support.

Fix echo and logprobs incompatibility in Completion

1

This PR aims to resolve the following error ([probably due to API change](https://community.openai.com/t/why-will-gpt-3-5-turbo-instruct-no-longer-support-echo-true-and-logprobs-1/404932)): `InvalidRequestError: Setting 'echo' and 'logprobs' at the same time is not supported for this model.` The original...

Set axolotl qlora-out as subfolder of notebook directory.

1

This PR sets the `qlora-out` path as a subfolder of the folder containing `lm-hackers` notebook. It's important to note that this works fine only if you run the fine tuning...

Run notebook end-to-end with GTX 3090/4090

1

The main purpose of this PR is to enable users with RTX 3090/4090 cards to run the `lm-hackers.ipynb` notebook end-to-end. To achieve this, we need to free both the host...

Downloading GGML model from Huggingface hub

1

Downloading `llama-2-7b-chat.Q4_K_M.gguf` from Huggingface when needed instead of relying on the pre-downloaded copy.