Ecosyste.ms: Issues

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

GitHub / pytorch/torchtune issues and pull requests

#931 - Llama-3 Inference and Uploading to Huggingface

Issue - State: closed - Opened by fabriceyhc 10 months ago - 19 comments

#930 - lm harness distributed evaluation?

Issue - State: open - Opened by monk1337 10 months ago - 3 comments
Labels: enhancement

#929 - Add recipe test for llama3

Pull Request - State: closed - Opened by SLR722 10 months ago - 2 comments
Labels: CLA Signed

#928 - [WIP] Free generation for evals

Pull Request - State: closed - Opened by joecummings 10 months ago - 1 comment
Labels: CLA Signed

#925 - add llama-70B memory and perf numbers to the README table

Issue - State: closed - Opened by soumith 10 months ago - 3 comments

#922 - How I can find all the checkpoints and merge it manually? (Lora)

Issue - State: closed - Opened by monk1337 10 months ago - 4 comments
Labels: question

#921 - Request support for Eleuther generative task

Issue - State: closed - Opened by ScottHoang 10 months ago - 3 comments

#916 - Unify example dataset in configs

Issue - State: closed - Opened by joecummings 10 months ago - 4 comments

#915 - no error when lacking hugging face permissions

Issue - State: closed - Opened by dangbert 10 months ago - 4 comments

#914 - Document model tokenizers

Issue - State: closed - Opened by joecummings 10 months ago
Labels: documentation

#912 - Finish additions needed for Phi-3 Mini 4K

Pull Request - State: open - Opened by joecummings 10 months ago - 1 comment
Labels: CLA Signed

#911 - Loading mistral reward model checkpoints

Pull Request - State: closed - Opened by SalmanMohammadi 10 months ago - 1 comment
Labels: CLA Signed

#910 - New integration - CometLogger

Pull Request - State: closed - Opened by Lothiraldan 10 months ago - 9 comments
Labels: CLA Signed

#909 - enable QLoRA + FSDP2

Pull Request - State: closed - Opened by weifengpy 10 months ago - 3 comments
Labels: CLA Signed

#908 - Remove reference to 2.2.2 as latest stable in README

Pull Request - State: closed - Opened by rohan-varma 10 months ago - 1 comment
Labels: CLA Signed

#907 - Remove usage of LRU cache from peft_utils

Pull Request - State: closed - Opened by ebsmothers 10 months ago - 1 comment
Labels: CLA Signed

#906 - document integration with bitsandbytes?

Issue - State: open - Opened by Titus-von-Koeller 10 months ago - 3 comments
Labels: documentation

#905 - Test directory structure for models doesn't match corresponding implementation files

Issue - State: closed - Opened by ebsmothers 10 months ago - 1 comment
Labels: best practice

#904 - `get_adapter_params` does not free GPU memory

Issue - State: closed - Opened by Optimox 10 months ago - 2 comments

#903 - total_training_steps -> global_step

Pull Request - State: closed - Opened by tcapelle 10 months ago - 1 comment
Labels: CLA Signed

#902 - Remove non-existant objects from `__all__`

Pull Request - State: closed - Opened by vmoens 10 months ago - 1 comment
Labels: CLA Signed

#901 - [Help needed] Impact of padding on causal attention ?

Issue - State: open - Opened by Optimox 10 months ago - 7 comments
Labels: discussion

#900 - Remove recipe_state from eleuther config

Pull Request - State: closed - Opened by rohan-varma 10 months ago - 1 comment
Labels: CLA Signed

#899 - Simplify eleuther eval config

Issue - State: closed - Opened by rohan-varma 10 months ago - 1 comment

#898 - Remove unnecessary system role's index check for llama3

Pull Request - State: open - Opened by musab-mk 10 months ago - 6 comments
Labels: CLA Signed

#897 - Compute grad norm

Pull Request - State: open - Opened by tcapelle 10 months ago - 10 comments
Labels: CLA Signed

#896 - Add documentation information

Pull Request - State: closed - Opened by joecummings 10 months ago - 1 comment
Labels: CLA Signed

#895 - [Feature addition] Clearml logger integration

Pull Request - State: open - Opened by Prakyathkantharaju 10 months ago - 9 comments
Labels: CLA Signed

#894 - Feature Request : ORPO

Issue - State: closed - Opened by nivibilla 10 months ago - 4 comments
Labels: documentation

#893 - [FR] (Q)DoRA

Issue - State: closed - Opened by DreamGenX 10 months ago - 8 comments
Labels: enhancement, rfc

#892 - [FR] Sample Packing with correct attention mask

Issue - State: closed - Opened by DreamGenX 10 months ago - 3 comments

#891 - Runtime Error: BF16 unsupported on supported hardware

Issue - State: closed - Opened by slobodaapl 10 months ago - 8 comments

#890 - Support `conversation_style` of `openai` format (OpenAI API style)

Pull Request - State: closed - Opened by xingyaoww 10 months ago - 4 comments
Labels: CLA Signed

#889 - Feat: Add support of multiple datasets in config

Pull Request - State: open - Opened by EvilFreelancer 10 months ago - 14 comments
Labels: CLA Signed

#888 - Mistral testing

Pull Request - State: closed - Opened by SalmanMohammadi 10 months ago - 6 comments
Labels: CLA Signed

#887 - Compile workflows seem broken on 2.3

Issue - State: closed - Opened by rohan-varma 10 months ago - 1 comment

#886 - Remove unused imports in models/__init__.py

Pull Request - State: closed - Opened by ebsmothers 10 months ago - 1 comment
Labels: CLA Signed

#885 - Enable profiler only on rank 0

Issue - State: closed - Opened by rohan-varma 10 months ago - 1 comment
Labels: enhancement, help wanted

#884 - Add support for 8da4w quantization

Pull Request - State: closed - Opened by andrewor14 10 months ago - 3 comments
Labels: CLA Signed

#883 - Validation and early stopping during training

Issue - State: open - Opened by kinggongzilla 10 months ago - 7 comments
Labels: enhancement, high-priority, community help wanted

#882 - Decouple ModelType from checkpointer

Issue - State: closed - Opened by rohan-varma 10 months ago

#881 - refacto: expose output layer for gemma models

Pull Request - State: closed - Opened by Optimox 10 months ago - 9 comments
Labels: CLA Signed

#880 - Support for MS Phi-3, please.

Issue - State: closed - Opened by razvanab 10 months ago - 1 comment
Labels: enhancement

#879 - Got error when download llama3 via the 'tune download'

Issue - State: closed - Opened by mazzzystar 10 months ago - 1 comment

#877 - Understanding contents of the final checkpoint file

Issue - State: closed - Opened by man-shar 10 months ago - 2 comments

#876 - Add Phi3 Mini 4K Instruct Model to torchtune

Pull Request - State: closed - Opened by kartikayk 10 months ago - 1 comment
Labels: CLA Signed

#875 - Sample packing for map datasets with correct RoPE encoding and no cross-contamination

Pull Request - State: closed - Opened by RdoubleA 10 months ago - 18 comments
Labels: CLA Signed

#874 - Feature/raft fine tuning

Pull Request - State: open - Opened by efenocchi 10 months ago - 4 comments
Labels: CLA Signed

#873 - Correctly pass TORCHTUNE_VERSION_DOCS during the build

Pull Request - State: closed - Opened by svekars 10 months ago - 2 comments
Labels: CLA Signed

#872 - Update Llama capitalization in docs

Pull Request - State: closed - Opened by joecummings 10 months ago - 2 comments
Labels: CLA Signed

#871 - Support stopping on more than just eos during generation

Pull Request - State: closed - Opened by ebsmothers 10 months ago - 1 comment
Labels: CLA Signed

#870 - Can't seem to get latest version via pip

Issue - State: closed - Opened by man-shar 10 months ago - 2 comments

#869 - Understanding QLora memory consumption for inference

Issue - State: closed - Opened by Optimox 10 months ago - 6 comments

#868 - Support for unstructured text corpus datasets for CPT

Pull Request - State: closed - Opened by RdoubleA 10 months ago - 5 comments
Labels: CLA Signed

#867 - Add missing Gemma recipes to registry, update default log_every_n_steps

Pull Request - State: closed - Opened by ebsmothers 10 months ago - 1 comment
Labels: CLA Signed

#866 - Revise the implementation of set_activation_checkpointing

Pull Request - State: closed - Opened by rohan-varma 10 months ago - 2 comments
Labels: CLA Signed

#865 - FSDP Llama3 wrapping improvements for full finetune

Pull Request - State: closed - Opened by rohan-varma 10 months ago - 9 comments
Labels: CLA Signed

#864 - File issues with 70B Lora setup

Issue - State: closed - Opened by BedirT 10 months ago - 4 comments

#863 - Is there a plan for supporting full fine-tuning 70B model?

Issue - State: closed - Opened by dmammfl 10 months ago - 6 comments
Labels: question

#862 - Update README to include 70B download command

Pull Request - State: closed - Opened by rohan-varma 10 months ago - 2 comments
Labels: CLA Signed

#861 - Validate all paths before doing any expensive work

Issue - State: open - Opened by rohan-varma 10 months ago
Labels: good first issue, community help wanted, better engineering

#860 - LLama 3 : can't load the hugging face state dict

Issue - State: closed - Opened by Optimox 10 months ago - 2 comments

#859 - token

Pull Request - State: closed - Opened by stefanrattay1 10 months ago - 3 comments

#858 - Can not download llama3 via the 'tune download'

Issue - State: closed - Opened by MaxwelsDonc 10 months ago - 1 comment

#857 - Documentation: Clarify all llama3 recipes

Pull Request - State: closed - Opened by musabgultekin 10 months ago - 5 comments
Labels: CLA Signed

#856 - Context Length Increse Results in OOM

Issue - State: closed - Opened by BedirT 10 months ago - 3 comments
Labels: documentation

#855 - enable LoRA + FSDP2

Pull Request - State: closed - Opened by weifengpy 10 months ago - 4 comments
Labels: CLA Signed

#854 - Cant load llama3 8B into memory

Issue - State: closed - Opened by abpani 10 months ago - 4 comments

#853 - Add eval results to QLoRA tutorial

Issue - State: closed - Opened by rohan-varma 10 months ago

#852 - update readme.md - add command to download llama3

Pull Request - State: closed - Opened by lessw2020 10 months ago - 1 comment
Labels: CLA Signed

#851 - Default to llama3-8b-instruct

Pull Request - State: closed - Opened by ebsmothers 10 months ago - 6 comments
Labels: CLA Signed

#850 - `tune run` crashes w/ ` `NotImplementedError`

Issue - State: closed - Opened by optimass 10 months ago - 4 comments

#849 - `tune download` doesn't download the weights

Issue - State: closed - Opened by optimass 10 months ago - 3 comments

#848 - Testing for Mistral models

Issue - State: open - Opened by SalmanMohammadi 10 months ago

#847 - adding model builders for code-llama2 7b, 13b, and 70b

Pull Request - State: closed - Opened by SalmanMohammadi 10 months ago - 13 comments
Labels: CLA Signed

#846 - Can not import torchtune

Issue - State: closed - Opened by abpani 10 months ago - 4 comments

#845 - Can i fine tune "dolphin-2.2.1-mistral-7b.Q2_K.gguf" with torchtune ? using cpu ?

Issue - State: closed - Opened by walidbet18 10 months ago - 14 comments
Labels: documentation, question

#844 - Multi-GPU QLoRA?

Issue - State: closed - Opened by cuichenx 10 months ago - 7 comments

#843 - [FSDP1] reduce GPU memory usage from 78G instead of 23G

Pull Request - State: closed - Opened by weifengpy 10 months ago - 3 comments
Labels: CLA Signed

#842 - Update PR template

Pull Request - State: closed - Opened by ebsmothers 10 months ago - 1 comment
Labels: CLA Signed

#841 - Better dataset and data documentation

Pull Request - State: closed - Opened by joecummings 10 months ago - 2 comments
Labels: CLA Signed

#840 - Transformer classifier

Pull Request - State: closed - Opened by SalmanMohammadi 10 months ago - 12 comments
Labels: CLA Signed

#839 - vRAM usage display for 70B models

Issue - State: closed - Opened by BedirT 10 months ago - 4 comments
Labels: documentation

#833 - Update nightly build instructions

Pull Request - State: closed - Opened by joecummings 10 months ago - 1 comment
Labels: CLA Signed

#832 - How to save a trained model so it can be loaded with HF `from_pretrained()`?

Issue - State: closed - Opened by calmitchell617 10 months ago - 30 comments
Labels: enhancement

#831 - Metric logger improvements

Pull Request - State: closed - Opened by ebsmothers 10 months ago - 4 comments
Labels: CLA Signed

#830 - VRAM Usage / Training Time in comparison to Huggingface

Issue - State: closed - Opened by bdytx5 10 months ago - 5 comments

#829 - Make Linter requirements for PR more prominent

Issue - State: closed - Opened by kartikayk 10 months ago - 1 comment

#828 - Running LoRA + QLoRA on Colab Notebooks

Issue - State: open - Opened by kartikayk 10 months ago - 1 comment
Labels: enhancement

#827 - feat: added packed continued pretraining dataset functionality

Pull Request - State: open - Opened by calmitchell617 10 months ago - 5 comments
Labels: CLA Signed

#826 - Finetuning LLama 2 Code Instruct without using Hugging Face

Issue - State: closed - Opened by sgupta1007 10 months ago - 7 comments

#825 - e2e torchtune tutorial example

Issue - State: open - Opened by seyeint 10 months ago - 7 comments
Labels: enhancement

#824 - Llama3 ChatFormat?

Issue - State: closed - Opened by Broyojo 10 months ago - 14 comments

#823 - Chat data + prompt template tutorial

Pull Request - State: closed - Opened by RdoubleA 10 months ago - 1 comment
Labels: CLA Signed

#822 - Delete unused APIs

Pull Request - State: closed - Opened by rohan-varma 10 months ago - 5 comments
Labels: CLA Signed

#812 - [RFC] Proximal Policy Optimisation

Issue - State: closed - Opened by SalmanMohammadi 10 months ago - 9 comments
Labels: rfc

#811 - [Feature Request] Support multimodal LLM, e.g., llava

Issue - State: closed - Opened by StarCycle 10 months ago - 2 comments
Labels: enhancement

#809 - Seeking guidance on continuing pretraining

Issue - State: closed - Opened by calmitchell617 10 months ago - 5 comments