https://github.com/truefoundry/llm-finetune

LLM Finetuning with Axolotl with decent defaults + Optional TrueFoundry Experiment Tracking Extension
https://github.com/truefoundry/llm-finetune

llm-finetuning

Last synced: 28 days ago
JSON representation

LLM Finetuning with Axolotl with decent defaults + Optional TrueFoundry Experiment Tracking Extension

• Host: GitHub
• URL: https://github.com/truefoundry/llm-finetune
• Owner: truefoundry
• Created: 2023-12-07T07:30:29.000Z (about 1 year ago)
• Default Branch: main
• Last Pushed: 2024-12-05T20:37:57.000Z (28 days ago)
• Last Synced: 2024-12-05T21:28:54.831Z (28 days ago)
• Topics: llm-finetuning
• Language: Python
• Homepage:
• Size: 1.32 MB
• Stars: 4
• Watchers: 3
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
> [!important]

> Please prefer using commits from [release tags](https://github.com/truefoundry/llm-finetune/releases). `main` branch is work in progress and may have partially working commits.

## LLM Finetuning with Truefoundry

Test QLoRA w/ Deepspeed Stage 2

```

./sample_run.sh

```

---

TODO:

- [ ] Setup C/I Tests

- [ ] Track and publish VRAM and Speed benchmarks for popular models and GPUs

---

Generally we always try to optimize for memory footprint because that allows higher batch size and more gpu utilization

Speedup is second priority but we take what we can easily get