Lightning-AI/litgpt issues and pull requests

#1817 - 0.5.3 post release setup

Pull Request - State: open - Opened by rasbt 18 days ago

#1816 - Update pyproject.toml

Pull Request - State: closed - Opened by rasbt 18 days ago

#1815 - Bumb version for 0.5.4 bugfix release

Pull Request - State: closed - Opened by rasbt 18 days ago

#1814 - Remove hardcoded 32-precision conversion

Pull Request - State: closed - Opened by rasbt 18 days ago

#1813 - fully fine-tuning specific set of layer

Issue - State: open - Opened by sorobedio 18 days ago - 1 comment
Labels: question

#1812 - Preicison none erronously always sets to 32-true

Issue - State: open - Opened by AurelioAguirre 18 days ago - 1 comment
Labels: bug

#1811 - Litgpt download

Issue - State: open - Opened by Wu-tn 18 days ago - 1 comment
Labels: question

#1810 - Move distributed all_reduce import into a function

Pull Request - State: closed - Opened by IvanYashchuk 19 days ago

#1809 - Support for mini-omni and mini-omni2 pre training, fine tuning on custom dataset.

Issue - State: open - Opened by satheeshkola-532 20 days ago
Labels: question

#1808 - Question about loss calculated in finetuning

Issue - State: closed - Opened by richardzhuang0412 22 days ago - 3 comments
Labels: question

#1807 - typo in convert_to_litgpt command

Pull Request - State: closed - Opened by wasifferoze 22 days ago

#1806 - Add support for quantized Llama 3.2 checkpoints

Issue - State: open - Opened by rasbt 23 days ago
Labels: model-weights

#1805 - Pretrain tok sec

Pull Request - State: closed - Opened by rasbt 23 days ago

#1804 - Choose evaluation example from test set

Pull Request - State: closed - Opened by rasbt 23 days ago

#1803 - Add Nvidia Llama 3.1 70B Nemotron weights

Pull Request - State: closed - Opened by rasbt 23 days ago

#1802 - Fixed the issue that precision is always "32-true".

Pull Request - State: closed - Opened by jianpingw 24 days ago

#1801 - Add token counts to compute performance

Pull Request - State: closed - Opened by rasbt 24 days ago

#1800 - Add Phi 3 128k model

Pull Request - State: closed - Opened by deveworld 24 days ago

#1799 - Post-release setup for 0.5.3.dev1

Pull Request - State: closed - Opened by rasbt 25 days ago

#1798 - 0.5.2 release

Pull Request - State: closed - Opened by rasbt 25 days ago

#1797 - finetune with lora tutorial error

Issue - State: closed - Opened by micrem73 25 days ago - 3 comments
Labels: question

#1796 - Pin to PyTorch 2.4.1

Pull Request - State: closed - Opened by rasbt 26 days ago

#1795 - Housekeeping: remove unused code and fix tests

Pull Request - State: closed - Opened by rasbt 26 days ago

#1794 - Fix step iteration bug in finetuning scripts

Pull Request - State: closed - Opened by rasbt 26 days ago

#1793 - Training on n epochs results in 1 more step in n+1 epoch

Issue - State: closed - Opened by Oseltamivir 29 days ago - 4 comments
Labels: bug

#1792 - litgpt cannot be imported with PyTorch build without distributed support

Issue - State: closed - Opened by IvanYashchuk about 1 month ago - 3 comments
Labels: bug, question

#1791 - litgpt download ..model_name... --convert_checkpoint false is performing conversion

Issue - State: open - Opened by wiiiktor about 1 month ago - 2 comments
Labels: enhancement

#1790 - Fix prompt style usage in evaluation example

Pull Request - State: closed - Opened by rasbt about 1 month ago

#1789 - Post 0.5.1 release setup

Pull Request - State: closed - Opened by rasbt about 1 month ago

#1788 - 8-bit precision training

Issue - State: open - Opened by rasbt about 1 month ago - 2 comments
Labels: enhancement

#1787 - Bump version for 0.5.1 bugfix release

Pull Request - State: closed - Opened by rasbt about 1 month ago

#1786 - Tidy up RoPE

Pull Request - State: closed - Opened by rasbt about 1 month ago

#1785 - Use more realistic RoPE tests

Pull Request - State: closed - Opened by rasbt about 1 month ago

#1784 - Add missing test dependency: flaky

Pull Request - State: closed - Opened by rasbt about 1 month ago

#1783 - `litgpt download list` not working

Issue - State: closed - Opened by a2jinhee about 1 month ago - 1 comment
Labels: bug

#1782 - Fixing Llama 3.1 and 3.2 Maximum Context Length

Pull Request - State: closed - Opened by rasbt about 1 month ago

#1781 - Be more specific about missign RoPE parameters

Pull Request - State: closed - Opened by rasbt about 1 month ago

#1780 - Fix CI

Pull Request - State: closed - Opened by rasbt about 1 month ago

#1779 - Use correct Llama 3.1 and 3.2 context lengths

Pull Request - State: closed - Opened by rasbt about 1 month ago

#1778 - prompts.py and "evaluating on a custom test set" section in the model evaluation tutorial

Issue - State: closed - Opened by zty12713 about 1 month ago - 2 comments
Labels: question

#1777 - Config files are missing rope_adjustments

Issue - State: closed - Opened by mtasic85 about 1 month ago - 2 comments
Labels: enhancement

#1776 - Code evaluation using bigcode-evaluation-harness framework

Issue - State: open - Opened by mtasic85 about 1 month ago - 1 comment
Labels: enhancement

#1775 - AMD (MI250X) support

Pull Request - State: closed - Opened by TensorTemplar about 1 month ago - 10 comments

#1774 - v0.5.0 post release setup

Pull Request - State: closed - Opened by rasbt about 1 month ago

#1773 - Bumb version for 0.5.0 release

Pull Request - State: closed - Opened by rasbt about 1 month ago

#1772 - Fix Llama 3.2 tokenizer

Pull Request - State: closed - Opened by rasbt about 1 month ago

#1771 - Investigate tokenizer discrepancy

Issue - State: closed - Opened by rasbt about 1 month ago
Labels: bug

#1770 - Add bnb.nn.StableEmbedding for quantized training

Pull Request - State: closed - Opened by rasbt about 1 month ago

#1769 - bnb optimizers could use bnb.nn.StableEmbedding instead of torch.nn.Embedding

Issue - State: closed - Opened by mtasic85 about 1 month ago - 2 comments
Labels: enhancement

#1768 - Add Salamandra models - Multilingual LLaMA-based models including minority languages

Issue - State: open - Opened by miguel-kjh about 1 month ago - 1 comment
Labels: enhancement

#1767 - Use latest bitsandbytes version

Issue - State: open - Opened by rasbt about 1 month ago
Labels: enhancement

#1766 - [fix][1760] Added fix for the missing `context` key issue in dolly!

Pull Request - State: closed - Opened by pytholic about 2 months ago - 10 comments

#1765 - Bump lightning version to latest stable release (2.4.0)

Pull Request - State: closed - Opened by rasbt about 2 months ago

#1764 - Errors when try to save checkpoints during the full fine-tuning

Issue - State: open - Opened by lemon-awa about 2 months ago
Labels: question

#1763 - Molmo support

Issue - State: open - Opened by win4r about 2 months ago
Labels: enhancement

#1762 - Abnormal Output from Gemma Pretrained Model After Conversion to Hugging Face Format

Issue - State: open - Opened by SKNahin about 2 months ago - 2 comments
Labels: bug

#1761 - Changing default activation function

Issue - State: closed - Opened by mtasic85 about 2 months ago - 6 comments
Labels: question

#1760 - Issue with Dolly Dataloader: `context` key not found!

Issue - State: open - Opened by pytholic about 2 months ago - 5 comments
Labels: bug

#1759 - TypeError: TextInputSequence must be str

Issue - State: open - Opened by hemanth about 2 months ago - 1 comment
Labels: bug

#1758 - invalid decimal literal?

Issue - State: closed - Opened by hemanth about 2 months ago - 2 comments
Labels: question

#1757 - How to set seed when generation outputs are not the same everytime?

Issue - State: closed - Opened by mrpeerat about 2 months ago - 3 comments
Labels: question

#1756 - Add missing explanation on how to use a finetuned model

Pull Request - State: closed - Opened by rasbt about 2 months ago

#1755 - Post 0.4.13 release set up

Pull Request - State: closed - Opened by rasbt about 2 months ago

#1754 - Temporarily take out thunder dependency for deployment

Pull Request - State: closed - Opened by rasbt about 2 months ago

#1753 - Bump version to 0.4.13 for new release

Pull Request - State: closed - Opened by rasbt about 2 months ago

#1752 - Repair json files

Pull Request - State: closed - Opened by rasbt about 2 months ago

#1751 - Increase rtol and atol in Gemma 2 for macOS

Pull Request - State: closed - Opened by rasbt about 2 months ago

#1750 - Llama 3.2 weights

Pull Request - State: closed - Opened by rasbt about 2 months ago

#1749 - Make json parsing more robust

Pull Request - State: closed - Opened by rasbt about 2 months ago

#1748 - Fix jsonarparse version

Pull Request - State: closed - Opened by rasbt about 2 months ago - 1 comment

#1747 - Test CI

Pull Request - State: closed - Opened by rasbt about 2 months ago - 1 comment

#1746 - Update RoPE tests

Pull Request - State: closed - Opened by rasbt about 2 months ago

#1745 - Improve rope

Pull Request - State: closed - Opened by rasbt about 2 months ago

#1744 - Support for optimizers which don't have "fused" parameter such as grokadamw and 8bit bnb

Pull Request - State: closed - Opened by mtasic85 about 2 months ago - 4 comments

#1743 - grokadamw and 8bit bnb optimizers

Issue - State: closed - Opened by mtasic85 about 2 months ago - 5 comments
Labels: enhancement

#1742 - Extend context length for sliding window tests

Pull Request - State: closed - Opened by rasbt about 2 months ago

#1741 - Add sliding window attention to Mistral and Phi 3

Pull Request - State: closed - Opened by rasbt about 2 months ago

#1740 - Update Thunder README.md

Pull Request - State: closed - Opened by rasbt about 2 months ago

#1739 - Support 128k token version of Phi 3

Issue - State: closed - Opened by rasbt about 2 months ago
Labels: enhancement, model-weights

#1738 - Add a more informative message in case text exceeds context size

Pull Request - State: closed - Opened by rasbt about 2 months ago

#1737 - Improve filepath handling in unit tests

Pull Request - State: closed - Opened by rasbt about 2 months ago

#1736 - Typo fix and formatting improvements in API Trainer docs

Pull Request - State: closed - Opened by rasbt about 2 months ago

#1735 - Add pretrain conversion

Pull Request - State: closed - Opened by rasbt about 2 months ago

#1734 - Fixing the tokenizer for slimpajama data preparation

Pull Request - State: closed - Opened by tomaslaz about 2 months ago

#1733 - cuda error when serve with `workers_per_device` > 1 and using concurrency request

Issue - State: closed - Opened by puppyapple about 2 months ago - 6 comments
Labels: enhancement

#1732 - downloading model

Issue - State: closed - Opened by alabazazzade about 2 months ago - 3 comments
Labels: question

#1731 - Pretraining with other strategies than FSDP

Issue - State: closed - Opened by HariSeldon11988 about 2 months ago - 1 comment
Labels: question

#1730 - Download checkpoints with `litgpt serve`

Issue - State: closed - Opened by aniketmaurya about 2 months ago - 4 comments
Labels: bug

#1729 - use initial_checkpoint_dir for continue-pretraining but can't load model correctly

Issue - State: closed - Opened by wodelt about 2 months ago - 6 comments
Labels: documentation, enhancement, question

#1728 - Add Chainlit Studio

Pull Request - State: closed - Opened by rasbt 2 months ago

#1727 - Question about tie_embeddings

Issue - State: open - Opened by twaka 2 months ago
Labels: question

#1726 - Simplify MPS support

Pull Request - State: closed - Opened by rasbt 2 months ago - 2 comments

#1725 - bump macos to m1

Pull Request - State: closed - Opened by t-vi 2 months ago - 9 comments

#1724 - Enable MPS support for LitGPT

Pull Request - State: closed - Opened by rasbt 2 months ago - 3 comments

#1723 - Is Support for the DeepSeek v2.5 model on the roadmap?

Issue - State: open - Opened by ColumbusAI 2 months ago - 1 comment
Labels: checkpoints

#1722 - Make 0.4.13.dev1 version

Pull Request - State: closed - Opened by rasbt 2 months ago

#1721 - Bump for v0.4.12

Pull Request - State: closed - Opened by rasbt 2 months ago

#1720 - Change default precision on macOS

Pull Request - State: closed - Opened by rasbt 2 months ago

#1719 - Add uvloop as a dependency

Pull Request - State: closed - Opened by rasbt 2 months ago

#1718 - Temporarily remove MPS device support

Pull Request - State: closed - Opened by rasbt 2 months ago

GitHub / Lightning-AI/litgpt issues and pull requests