Falcon 7B Code Alpaca
Jina's Finetuner team adapted Falcon 7B to the Code Alpaca dataset. This is a trained derivative of the Technology Innovation Institute model, not a claim that Jina built Falcon's original backbone.
Research release
How the model works
The release documents LoRA rank 8 on query-key-value modules, a learning rate of 0.0003 and a batch size of 128. This 7B version ran for 6 epochs with a micro-batch size of 8. Its training can be reproduced with Jina's Jerboa repository.Falcon 7B Code Alpaca model details
Use and access
Jina publishes full 16-bit merged weights and a separate LoRA adapter repository. They represent the same fine-tune and are grouped in this entry. The card labels the release CC BY-NC 4.0. It is a historical research checkpoint; no current managed API for this specific fine-tune was verified.Falcon 7B Code Alpaca model details
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Research release |
| Launch date | |
| Access | Downloadable |
| Model identifiers | falcon-7b-code-alpacajinaai/falcon-7b-code-alpacafalcon-7b-code-alpaca-lora |
| Parameters | 7 billionFalcon 7B Code Alpaca model details |
| Weight license | CC BY-NC 4.0Falcon 7B Code Alpaca model details |
| Base model | TII Falcon 7BFalcon 7B Code Alpaca model details |
Related models
- StarCoder 1B Textbook
Jina AI · Jina Code Fine-tunes
A small code-completion model trained on synthetic programming exercises.
Launch date Research releaseDownloadable - Falcon 40B Code Alpaca
Jina AI · Jina Code Fine-tunes
A Jina-trained Falcon 40B adaptation using Code Alpaca examples.
Launch date Research releaseDownloadable