Ecosyste.ms: Issues

An open API service for providing issue and pull request metadata for open source projects.

GitHub / Lightning-AI/litgpt issues and pull requests

#1717 - Cannot attend to 9904, block size is only 4096

Issue - State: closed - Opened by starjob42 2 months ago - 1 comment
Labels: question

#1716 - "RuntimeError: All the chunks should have been deleted." on non-Studio machine

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

#1715 - llm.generate issue on CPU machines

Issue - State: closed - Opened by rasbt 2 months ago - 3 comments
Labels: bug

#1714 - `llm.generate` function does not work on Mac (MPS) devices anymore

Issue - State: closed - Opened by rasbt 2 months ago
Labels: bug

#1713 - Llama 3.1 RoPE not implemented -- Finetuning will lead to spelling/grammar mistakes

Issue - State: closed - Opened by calvintwr 2 months ago - 3 comments
Labels: bug

#1712 - Data Loading bug in `pretrain` on resume over multiple epochs

Issue - State: open - Opened by fdalvi 2 months ago
Labels: bug

#1711 - Manual convert_to_litgpt for Phi-3.5-mini-instruct downloaded weights from HF

Issue - State: closed - Opened by wasifferoze 2 months ago - 1 comment
Labels: bug

#1710 - minor Readme update/typos

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

#1709 - Qwen series

Issue - State: open - Opened by Godlikemandyy 2 months ago - 1 comment
Labels: question, model-weights

#1708 - Unable to train using LoRA however training with adapter seems fine

Issue - State: closed - Opened by salokr 2 months ago - 5 comments
Labels: bug

#1707 - Fix device Error in Decode Stream

Pull Request - State: closed - Opened by Motsepe-Jr 3 months ago

#1706 - zsh: illegal hardware instruction

Issue - State: closed - Opened by Suddhasatwa 3 months ago - 6 comments
Labels: bug

#1705 - Way to load quantized model on the fly through Python SDK

Issue - State: closed - Opened by sovit-123 3 months ago - 2 comments
Labels: question

#1704 - `accelerator` argument not working properly with `LitServer`

Issue - State: closed - Opened by sovit-123 3 months ago - 2 comments
Labels: bug

#1703 - Liger Kernel Integration

Issue - State: closed - Opened by ByronHsu 3 months ago - 5 comments
Labels: enhancement

#1702 - Add `batched_generate_fn()`

Pull Request - State: closed - Opened by apaz-cli 3 months ago

#1701 - bump thunder dependency to main

Pull Request - State: closed - Opened by t-vi 3 months ago - 1 comment

#1700 - add support for batched input_pos to model

Pull Request - State: closed - Opened by t-vi 3 months ago

#1699 - [BUG] LLaMA 3.1 RoPE

Issue - State: closed - Opened by zzhhjjj 3 months ago - 3 comments
Labels: bug, question

#1698 - Training not working with default script

Issue - State: closed - Opened by ByteBrigand 3 months ago - 17 comments
Labels: bug

#1697 - Ligpt for windows - triton

Issue - State: closed - Opened by seshganesh 3 months ago - 1 comment
Labels: bug

#1696 - Fix falcon prompt template

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

#1695 - Bumb version to 0.4.11

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

#1694 - Preserve eos in encoding when max_seq_length = -1

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

#1693 - Add `batched_next_token()` and `batched_sample()`

Pull Request - State: closed - Opened by apaz-cli 3 months ago - 8 comments

#1692 - Add `batched_next_token()` and `batched_sample()`

Pull Request - State: closed - Opened by apaz-cli 3 months ago - 1 comment

#1691 - Avoid error when executing benchmark util outside a git folder

Pull Request - State: closed - Opened by rasbt 3 months ago - 1 comment

#1690 - Make number of generated tokens consistent with CLI

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

#1689 - about litgpt command

Issue - State: closed - Opened by aekx 3 months ago - 3 comments
Labels: question

#1688 - Disable attention mask during new token generation

Pull Request - State: closed - Opened by Andrei-Aksionov 3 months ago

#1687 - Add Microsoft Phi 3.5 checkpoint

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

#1686 - Microsoft Phi 3.5 MoE

Issue - State: open - Opened by rasbt 3 months ago
Labels: enhancement, model-weights

#1685 - Spelling fix

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

#1684 - Update check_nvlink_connectivity

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

#1683 - adding a UI for the training and the finetuning

Issue - State: open - Opened by Esmail-ibraheem 3 months ago - 1 comment
Labels: enhancement

#1682 - Llama3 finetuning and generation: Double begin_of_text, no eot_id

Issue - State: open - Opened by sanderland 3 months ago - 9 comments
Labels: bug

#1681 - Added git hash to benchmark utility.

Pull Request - State: closed - Opened by apaz-cli 3 months ago - 1 comment

#1680 - Add PR benchmark util for internal use

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

#1679 - Improved benchmark utils

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

#1678 - Fix KV cache issue in LLM API

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

#1677 - Auto device handling in LLM API

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

#1676 - Add distribute=None to python-api.md

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

#1675 - Combine `generate()` functions

Pull Request - State: closed - Opened by apaz-cli 3 months ago - 10 comments

#1674 - Bumb version to 0.4.10 for next release

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

#1673 - Add Mistral Large 123B

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

#1672 - attention mask is incorrect when generate with softcapping

Issue - State: open - Opened by twaka 3 months ago - 8 comments
Labels: bug

#1671 - Disable KV cache option

Issue - State: open - Opened by rasbt 3 months ago
Labels: enhancement

#1670 - Multi-gpu serving

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

#1669 - Update azure-gpu-test.yml

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

#1668 - Support the refactored API in litgpt serve

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

#1667 - Make LitGPT LLM API compatible with PyTorch Lightning Trainer 1/2

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

#1666 - Swap old Llama model with Phi-3

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

#1665 - Gemma 2B weights seem to have changed

Issue - State: open - Opened by rasbt 3 months ago - 1 comment
Labels: bug

#1664 - Bumb version for 0.4.9 release

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

#1663 - Tensor parallelism generates non-sensical outputs

Issue - State: open - Opened by rasbt 3 months ago - 1 comment
Labels: bug

#1662 - Use FlexAttention

Issue - State: open - Opened by rasbt 3 months ago
Labels: enhancement, performance

#1661 - Support Tensor Parallel in Python API

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

#1660 - Optionally return benchmark info in Python API

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

#1658 - Fix some issues with circular and relative imports

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

#1656 - Add a PyTorch Lightning example

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

#1655 - MPS test gave out gibberish result

Issue - State: closed - Opened by edifice1989 3 months ago - 3 comments
Labels: bug

#1654 - Update LitServe version and tests

Pull Request - State: closed - Opened by rasbt 4 months ago - 1 comment

#1653 - Version bumb for Gemma 2 2B release

Pull Request - State: closed - Opened by rasbt 4 months ago

#1652 - Pin litserve version

Pull Request - State: closed - Opened by rasbt 4 months ago

#1638 - Implement prompt caching to speed up inference

Issue - State: open - Opened by rasbt 4 months ago - 1 comment
Labels: enhancement

#1628 - Support sequential generation in litgpt serve

Issue - State: closed - Opened by rasbt 4 months ago - 1 comment
Labels: enhancement

#1627 - Mistral Large 2 Checkpoints

Issue - State: closed - Opened by rasbt 4 months ago - 1 comment
Labels: enhancement, checkpoints

#1616 - Support downloading and using quantized weights (GGUF)

Issue - State: open - Opened by rasbt 4 months ago - 1 comment
Labels: enhancement

#1610 - Set kv cache size based on resource needs

Issue - State: closed - Opened by rasbt 4 months ago - 1 comment
Labels: enhancement

#1607 - Results of GPT2 (124M) reproduction with LitGPT

Issue - State: closed - Opened by awaelchli 4 months ago - 2 comments
Labels: question

#1606 - Is there any support for visual generation?

Issue - State: open - Opened by dunbar12138 4 months ago - 2 comments
Labels: enhancement, question

#1598 - Apply Sliding Window Attention to Mistral

Issue - State: closed - Opened by rasbt 4 months ago
Labels: enhancement

#1586 - Update Azure workflow to use the latest stable PyTorch version

Pull Request - State: closed - Opened by rasbt 4 months ago - 1 comment

#1578 - Add FlashAttention v3 support

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

#1576 - Restore previous LitData assertions in tests

Pull Request - State: closed - Opened by awaelchli 4 months ago - 4 comments

#1566 - Remove duplicated bos_token for CodeLlama

Pull Request - State: closed - Opened by alealv 4 months ago - 4 comments

#1538 - Do not wrap LoRA layers with FSDP

Pull Request - State: open - Opened by janEbert 5 months ago - 20 comments

#1522 - Tokenizer: prefer HF Tokenizer

Pull Request - State: closed - Opened by rasbt 5 months ago - 1 comment

#1498 - add support for Qwen 2

Issue - State: open - Opened by aniketmaurya 5 months ago - 5 comments
Labels: model-weights

#1483 - Recreate Mistral PR

Pull Request - State: closed - Opened by rasbt 5 months ago - 3 comments

#1443 - validation output during finetuning

Issue - State: closed - Opened by richardzhuang0412 6 months ago - 3 comments

#1431 - Mistral v0.3

Pull Request - State: closed - Opened by rasbt 6 months ago

#1430 - performing continuous pretraining and then finetuning causes error

Issue - State: open - Opened by richardzhuang0412 6 months ago - 4 comments
Labels: bug

#1363 - litgpt download doesn't work

Issue - State: closed - Opened by natanloterio 7 months ago - 10 comments
Labels: bug

#1220 - Can I train a model on 7900XT 4 cards?

Issue - State: open - Opened by win10ogod 8 months ago - 2 comments
Labels: question

#1202 - Finetuning run times out at evaluation step on multiple devices

Issue - State: open - Opened by ecatkins 8 months ago - 13 comments

#1192 - Introduce OptimizerArgs and add support for GaLore

Pull Request - State: closed - Opened by rasbt 8 months ago - 13 comments
Labels: breaking change

#1182 - Config URLs for pinned version

Pull Request - State: closed - Opened by awaelchli 8 months ago

#1156 - Separate out the biases

Pull Request - State: closed - Opened by rasbt 8 months ago - 6 comments

#1107 - Add MPS configs

Issue - State: open - Opened by rasbt 8 months ago - 7 comments

#934 - Adding DoRA (Weight-Decomposed Low-Rank Adaptation) to improve LoRA

Issue - State: open - Opened by rasbt 9 months ago - 20 comments
Labels: enhancement, fine-tuning

#796 - Harcoded incorrect (and repeated) validation example

Issue - State: closed - Opened by DavidGOrtega 12 months ago - 20 comments
Labels: enhancement

#620 - Sample packing for pretraining/fine-tuning

Issue - State: open - Opened by alitirmizi23 about 1 year ago - 16 comments
Labels: enhancement, pre-training

#491 - RuntimeError: Cannot writeback when the parameter shape changes

Issue - State: closed - Opened by ngctnnnn about 1 year ago - 11 comments
Labels: bug, fine-tuning

#100 - Add `incremental_save` to utils, use in `convert_hf_checkpoint`

Pull Request - State: closed - Opened by carmocca over 1 year ago - 2 comments