Ecosyste.ms: Issues

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

GitHub / TransformerLensOrg/TransformerLens issues and pull requests

#755 - Upstream update

Pull Request - State: closed - Opened by bryce13950 4 months ago

#754 - [Proposal] Ensure TransformerLens does not load from hugging face when config is passed in

Issue - State: open - Opened by hamind 4 months ago - 2 comments
Labels: complexity-moderate

#753 - Can't load gpt2 or gpt2-medium locally

Issue - State: closed - Opened by hamind 4 months ago

#752 - Demo colab compatibility

Pull Request - State: closed - Opened by bryce13950 4 months ago

#751 - Add support for `Mistral-Nemo-Base-2407` model

Pull Request - State: closed - Opened by ryanhoangt 4 months ago - 8 comments

#749 - add transformer diagram

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

#747 - [Bug Report] hook_normalized is inconsistent between RMSNorm and LayerNorm

Issue - State: open - Opened by neelnanda-io 4 months ago - 1 comment
Labels: bug, complexity-moderate, breaking-change

#746 - [Proposal] Add example of collecting activations from a single layer.

Issue - State: closed - Opened by adamkarvonen 4 months ago - 6 comments
Labels: demo, complexity-moderate

#745 - release 2.7.1

Pull Request - State: closed - Opened by bryce13950 4 months ago

#744 - Upstream update

Pull Request - State: closed - Opened by bryce13950 4 months ago

#742 - Updated broken Slack link

Pull Request - State: closed - Opened by neelnanda-io 4 months ago - 3 comments

#741 - Llama3.1

Pull Request - State: closed - Opened by mylesgoose 4 months ago - 22 comments

#740 - added llama 3.1 models and base for working on mllama

Pull Request - State: closed - Opened by mylesgoose 4 months ago - 5 comments

#739 - Avoid warning in `utils.download_file_from_hf`

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

#738 - [Proposal] Usage with openai/transformer-debugger

Issue - State: closed - Opened by fzyzcjy 4 months ago - 4 comments

#737 - [Bug Report] Q cannot be reshaped correctly when model is loaded in 4bit

Issue - State: open - Opened by po13on 4 months ago - 4 comments
Labels: bug, needs-investigation

#736 - Upstream update

Pull Request - State: closed - Opened by bryce13950 5 months ago

#735 - Release 2.7

Pull Request - State: closed - Opened by bryce13950 5 months ago

#734 - Model llama 3.2

Pull Request - State: closed - Opened by bryce13950 5 months ago

#733 - `utils.test_prompt` compares multiple prompts

Pull Request - State: closed - Opened by bryce13950 5 months ago

#732 - Type hooked encoder

Pull Request - State: closed - Opened by bryce13950 5 months ago

#731 - Model llama 3.2

Pull Request - State: closed - Opened by bryce13950 5 months ago

#730 - Fine tune model and using this framework

Issue - State: closed - Opened by nitay16 5 months ago - 3 comments
Labels: question, needs-information

#729 - [Proposal] Guide to adding new models

Issue - State: open - Opened by deven367 5 months ago - 9 comments
Labels: documentation, complexity-moderate

#728 - `utils.test_prompt` compares multiple prompts

Pull Request - State: closed - Opened by callummcdougall 5 months ago - 12 comments

#727 - fixed typo

Pull Request - State: closed - Opened by bryce13950 5 months ago

#726 - [Proposal] Warn people when trying to load t5 into HookedTransformer

Issue - State: closed - Opened by bryce13950 5 months ago - 4 comments
Labels: good first issue, complexity-simple

#725 - Fix the bug that tokenize_and_concatenate function not working for small dataset

Pull Request - State: closed - Opened by xy-z-code 5 months ago - 1 comment

#724 - Load state dict with assign to avoid OOMs

Pull Request - State: closed - Opened by cyber-chris 5 months ago - 6 comments

#723 - 2.6

Pull Request - State: closed - Opened by bryce13950 5 months ago

#722 - Redo of #713

Pull Request - State: closed - Opened by bryce13950 5 months ago

#721 - Release 2.5

Pull Request - State: closed - Opened by bryce13950 5 months ago

#720 - [Bug Report] Review current matmul function usages

Issue - State: open - Opened by bryce13950 5 months ago
Labels: bug, complexity-high

#719 - [Proposal] Add frequency-based RoPE support for Llama 3.1 models

Issue - State: closed - Opened by frances720 5 months ago - 3 comments

#718 - Add `allenai/OLMoE-1B-7B-0924`.

Pull Request - State: closed - Opened by joelburget 5 months ago - 7 comments

#717 - Allow loading only first n layers.

Pull Request - State: closed - Opened by joelburget 5 months ago

#716 - HookedTransformerConfig docs string: `weight_init_mode` => `init_mode`

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

#715 - Fix typo in bug issue template

Pull Request - State: closed - Opened by JasonGross 5 months ago

#714 - [Bug Report] Torch FutureWarning when calling `utils.download_file_from_hf` with `torch==2.4.1`

Issue - State: closed - Opened by albertsgarde 5 months ago - 6 comments
Labels: good first issue, complexity-simple

#713 - Ungrouping GQA

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

#712 - v2.4.1

Pull Request - State: closed - Opened by bryce13950 5 months ago

#711 - support for the Amber model, including checkpoints

Pull Request - State: closed - Opened by tadakeigo 5 months ago

#710 - [Proposal] Add MVP Support For 1-2 Models Per-Modality

Issue - State: open - Opened by 4gatepylon 5 months ago - 4 comments
Labels: complexity-high, discussion

#709 - [Bug Report] Gemma 2 unsupported?

Issue - State: closed - Opened by jasonlim131 6 months ago

#708 - [Bug Report] Gemma-2-2b not found

Issue - State: closed - Opened by jasonlim131 6 months ago - 1 comment

#706 - revised loading to recycle state dict

Pull Request - State: closed - Opened by bryce13950 6 months ago

#705 - Updated state loading to copy by reference

Pull Request - State: closed - Opened by bryce13950 6 months ago

#704 - [Proposal] Add support for TracrBench

Issue - State: open - Opened by HannesThurnherr 6 months ago - 3 comments
Labels: new-architecture, complexity-high

#703 - Upstream commit update

Pull Request - State: closed - Opened by bryce13950 6 months ago

#702 - Release 2.4.0

Pull Request - State: closed - Opened by bryce13950 6 months ago

#701 - Release v2.3.1

Pull Request - State: closed - Opened by bryce13950 6 months ago

#700 - Recent release

Pull Request - State: closed - Opened by bryce13950 6 months ago

#699 - Improve attention masking

Pull Request - State: closed - Opened by UFO-101 6 months ago - 1 comment

#698 - Vram not used rtx 4090's

Issue - State: closed - Opened by mylesgoose 6 months ago - 1 comment

#697 - How to get the Activation cache while the LLM is generating new tokens?

Issue - State: open - Opened by Meehaohao 6 months ago - 2 comments
Labels: complexity-moderate

#696 - About the cached layernorm scale factors

Issue - State: open - Opened by Meehaohao 6 months ago - 2 comments

#695 - Changed fold_value_biases method to be able to handle multi-gpu.

Pull Request - State: closed - Opened by Heigke 6 months ago - 4 comments

#694 - Update Gemma2 attention scale

Pull Request - State: closed - Opened by mntss 6 months ago - 4 comments

#693 - [Bug Report] Gemma-2-2b-it output logit doesn't match with huggingface

Issue - State: closed - Opened by yeutong 6 months ago - 7 comments
Labels: complexity-high, implementation-inaccuracy

#692 - add a demo for Patchscopes and Generation with Patching

Pull Request - State: closed - Opened by HenryCai11 6 months ago - 2 comments

#691 - [Proposal] Add Lllama 3.1 support

Issue - State: closed - Opened by ssuukk 6 months ago - 20 comments
Labels: new-architecture, complexity-moderate

#690 - Python 3.8 removal

Pull Request - State: closed - Opened by bryce13950 6 months ago

#689 - Added gemma-2 2b (#687)

Pull Request - State: closed - Opened by bryce13950 6 months ago

#688 - 2.3.0

Pull Request - State: closed - Opened by bryce13950 6 months ago

#687 - Added gemma-2 2b

Pull Request - State: closed - Opened by curt-tigges 6 months ago - 1 comment

#685 - [Bug Report] Different results from HuggingFace when using the GPT2 small example

Issue - State: closed - Opened by nreHieW 7 months ago - 1 comment
Labels: complexity-high, needs-investigation, implementation-inaccuracy

#684 - [Proposal] Expand quantization model support

Issue - State: open - Opened by miguel-kjh 7 months ago - 1 comment
Labels: complexity-high

#683 - [Bug Report] Qwen model implementation is too inaccurate

Issue - State: closed - Opened by bryce13950 7 months ago - 4 comments
Labels: complexity-high, needs-investigation, implementation-inaccuracy

#682 - updated dependencies

Pull Request - State: open - Opened by bryce13950 7 months ago

#681 - Test arena cleanup

Pull Request - State: closed - Opened by bryce13950 7 months ago

#680 - [Proposal] Demo and Tutorial on Patchscopes and "Patching + Generation"

Issue - State: closed - Opened by HenryCai11 7 months ago - 5 comments
Labels: demo, complexity-moderate

#679 - NamesFilter can be a string

Pull Request - State: closed - Opened by jettjaniak 7 months ago - 1 comment

#678 - Add Mixtral to `test_match_huggingface` test

Pull Request - State: closed - Opened by joelburget 7 months ago - 1 comment

#677 - Fix typo in `embed.py` docs

Pull Request - State: closed - Opened by ArthurConmy 7 months ago

#675 - Release 2.2.2

Pull Request - State: closed - Opened by bryce13950 7 months ago

#674 - added arena content as a notebook

Pull Request - State: closed - Opened by bryce13950 7 months ago

#673 - fix: fixing broken backward hooks change

Pull Request - State: closed - Opened by chanind 7 months ago - 1 comment

#672 - [Bug Report] Backward hooks are broken as of v2.0.0

Issue - State: closed - Opened by chanind 7 months ago - 1 comment

#671 - [Proposal] Allow tied embeddings

Issue - State: open - Opened by neelnanda-io 7 months ago - 1 comment
Labels: enhancement, complexity-moderate

#670 - ValueError: microsoft/Phi-3-mini-128k-instruct not found.

Issue - State: open - Opened by joykirat18 7 months ago - 1 comment
Labels: complexity-moderate, model-request

#669 - does run_with_cache method support data parallel , how can I do it ?

Issue - State: closed - Opened by Yang-bug-star 7 months ago - 1 comment

#668 - Release 2.2.1

Pull Request - State: closed - Opened by bryce13950 7 months ago

#667 - [Bug Report] Einops shape error when `use_attn_result = True`

Issue - State: closed - Opened by dtch1997 7 months ago - 1 comment

#666 - Fix attention result projection

Pull Request - State: closed - Opened by callummcdougall 7 months ago - 2 comments

#665 - [Proposal] Allow recent versions of beartype

Issue - State: open - Opened by jettjaniak 7 months ago - 6 comments
Labels: complexity-simple, tooling

#664 - [Question] Offline Error HookedTransformer.from_pretrained

Issue - State: closed - Opened by pbernabeup 7 months ago - 3 comments

#663 - Adding RMSNorm to apply_ln_to_stack

Pull Request - State: closed - Opened by gaabrielfranco 7 months ago - 1 comment

#662 - Add support for Qwen2 models

Pull Request - State: closed - Opened by g-w1 7 months ago - 3 comments

#661 - [Bug Report] Pythia output inconsistent across batch sizes when use_split_qkv_input=True

Issue - State: open - Opened by oliveradk 7 months ago
Labels: bug, complexity-high, implementation-inaccuracy

#660 - removed einsum causing error when use_atten_result is enabled

Pull Request - State: closed - Opened by oliveradk 7 months ago - 2 comments

#659 - [Bug Report] Attn Result hook not working

Issue - State: closed - Opened by oliveradk 7 months ago - 2 comments

#658 - docs: update Main_Demo.ipynb

Pull Request - State: closed - Opened by eltociear 7 months ago - 1 comment

#657 - [Bug Report] RMSNormPre in Transformer_lens is maybe different from Llama source code?

Issue - State: open - Opened by wangyifei0047 7 months ago - 1 comment
Labels: complexity-moderate, needs-investigation

#656 - Release 2.2

Pull Request - State: closed - Opened by bryce13950 7 months ago