OpenGVLab/LLaMA-Adapter issues and pull requests

#153 - created a model on colab but cannot load for inference

Issue - State: open - Opened by sedici16 4 months ago

#130 - I have problem with downloading 7B_chinese in imagebind_LLM.

Issue - State: closed - Opened by JINAILAB 12 months ago - 1 comment

#101 - AttributeError: module 'clip' has no attribute 'load'

Issue - State: closed - Opened by parasmech about 1 year ago - 2 comments

#100 - error when pretrin the llama-adapterv2-multimodal

Issue - State: open - Opened by adda1221 about 1 year ago - 2 comments

#99 - the format of dataset for fintuning on llama_adapter_v2_multimodal

Issue - State: open - Opened by adda1221 about 1 year ago - 4 comments

#98 - Code result not the same as in the gradio GUI.

Issue - State: open - Opened by Practicing7 about 1 year ago - 3 comments

#97 - Support for llama-2 70B

Issue - State: open - Opened by qizzzh about 1 year ago - 7 comments

#96 - Reproduce problems with the model llama-adapter-multimodal-v2

Issue - State: open - Opened by dongzhiwu about 1 year ago - 7 comments

#95 - alpaca_finetuning_v1 does not support llama2 checkpoint

Issue - State: open - Opened by GentleZhu about 1 year ago - 2 comments

#94 - Training data for audio model?

Issue - State: open - Opened by jpgard about 1 year ago - 1 comment

#93 - Pretrain model

Issue - State: open - Opened by yuntaodu over 1 year ago - 1 comment

#92 - Error when inference

Issue - State: closed - Opened by sci-m-wang over 1 year ago - 3 comments

#91 - Error when running example of imagebind_LLM

Issue - State: closed - Opened by basteran over 1 year ago - 7 comments

#90 - Learning rate and batch size

Issue - State: open - Opened by dhyani15 over 1 year ago - 3 comments

#89 - llamav2 base Chinese multimodal

Issue - State: open - Opened by lucasjinreal over 1 year ago

#88 - training time

Issue - State: open - Opened by cissoidx over 1 year ago - 2 comments

#87 - How many training epochs to replicate scienceQA result using llama v1

Issue - State: open - Opened by dhyani15 over 1 year ago - 12 comments

#86 - Question: How to convert weights to required format (.bin --> .pth)

Issue - State: closed - Opened by r3shma over 1 year ago - 1 comment

#84 - Could not replicate the same result.

Issue - State: open - Opened by poonehmousavi over 1 year ago - 1 comment

#83 - How long does it take to train v2-multimodal model on Image-Text-V1(a concatenation of LAION400M, COYO, MMC4, SUB, Conceptual Captions, and COCO) via 8 A100(80GB)?

Issue - State: open - Opened by ifshinelx over 1 year ago

#82 - Does it support multi image input？

Issue - State: open - Opened by hangzeli08 over 1 year ago - 1 comment

#81 - Fine Tune on AMD GPU

Issue - State: open - Opened by LeLaboDuGame over 1 year ago

#80 - LLaMA-Adapter-v2-multimodal evaluation code?

Issue - State: open - Opened by heliossun over 1 year ago - 1 comment

#79 - Two different sets of llama module

Issue - State: closed - Opened by dhyani15 over 1 year ago - 1 comment

#78 - Example and label for finetuning

Issue - State: closed - Opened by thao9611 over 1 year ago

#77 - Questions for reproducing llama-adapter v1 (finetuned on COCO Captions)

Issue - State: open - Opened by miso-choi over 1 year ago

#76 - No sympy found

Issue - State: closed - Opened by dhyani15 over 1 year ago - 2 comments

#75 - Save the model weights in a few hundred megabytes size like the BIAS-7B.pth provided by the official.

Pull Request - State: closed - Opened by Enderfga over 1 year ago - 4 comments

#74 - The code is inconsistent with v2 paper

Issue - State: open - Opened by merlinarer over 1 year ago - 4 comments

#73 - how to reproduce coco caption result in paper LLaMA-Adapter: Efficient Fine-tuning of Language Models with Zero-init Attention ?

Issue - State: open - Opened by baiyuting over 1 year ago - 5 comments

#72 - llama-adapterV2 multi modal demo error

Issue - State: open - Opened by dongzhiwu over 1 year ago - 5 comments

#71 - Some questions about imagebind_llm code

Issue - State: closed - Opened by keke-220 over 1 year ago - 7 comments

#70 - fun-tunning of image？

Issue - State: open - Opened by web3creator over 1 year ago - 4 comments

#69 - multi-scale CLIP feature

Issue - State: closed - Opened by qyx1121 over 1 year ago - 1 comment

#68 - Question for v2 finetuning on coco caption

Issue - State: open - Opened by simplewhite9 over 1 year ago - 5 comments

#67 - NaN Loss while training on a custom instruction dataset

Issue - State: closed - Opened by Sleepyhead01 over 1 year ago - 5 comments

#66 - Clarification on Pretrained Weights

Issue - State: closed - Opened by keke-220 over 1 year ago - 7 comments

#65 - Inference speed very slow

Issue - State: closed - Opened by Hsn37 over 1 year ago - 5 comments

#64 - On the 80k data to train chatbot

Issue - State: open - Opened by John-Ge over 1 year ago - 1 comment

#63 - model weights

Issue - State: open - Opened by 1zhangtianqing over 1 year ago - 7 comments

#62 - Code for Multi-modal Reasoning on ScienceQA.

Issue - State: open - Opened by kai-wen-yang over 1 year ago - 4 comments

#61 - A lot of files not found. Will they be released again?

Issue - State: open - Opened by Practicing7 over 1 year ago - 2 comments

#60 - Update misc.py

Pull Request - State: open - Opened by Nikhil-Paleti over 1 year ago

#59 - loss is nan

Issue - State: open - Opened by 1zhangtianqing over 1 year ago - 1 comment

#58 - LLaMA-Adapter V3?

Issue - State: closed - Opened by gusye1234 over 1 year ago - 1 comment

#57 - Question : How to effectively add training to specific domain ?

Issue - State: open - Opened by x4080 over 1 year ago - 2 comments

#56 - Where is the instruction/finetuning data for 3d?

Issue - State: open - Opened by mu-cai over 1 year ago - 2 comments

#55 - Finetuning Of V2?

Issue - State: open - Opened by poonehmousavi over 1 year ago - 3 comments

#54 - Add LangChain integration doc and demo for v1 and v2

Pull Request - State: closed - Opened by SiyuanHuang95 over 1 year ago - 1 comment

#53 - Can this adapter be run with OpenLLaMA?

Issue - State: open - Opened by jawhster over 1 year ago - 2 comments

#52 - Add ImageBind-LLM with 3D point cloud modality

Pull Request - State: closed - Opened by ZiyuGuo99 over 1 year ago

#51 - What dataset do you use for training the (released) multimodal adapter?

Issue - State: closed - Opened by dcahn12 over 1 year ago - 2 comments

#50 - Is it normal to have outputs for example.py that is different from yours?

Issue - State: open - Opened by yxchng over 1 year ago - 1 comment

#49 - Is the GPL License correct?

Issue - State: closed - Opened by RonanKMcGovern over 1 year ago - 4 comments

#48 - Question: How to save checkpoints after every epoch?

Issue - State: closed - Opened by salehshadi over 1 year ago - 2 comments

#47 - Adapter: v1 vs v2 implementation

Issue - State: closed - Opened by Andrei-Aksionov over 1 year ago - 2 comments

#46 - how to merge adapter to original weights

Issue - State: open - Opened by LiuPearl1 over 1 year ago - 8 comments

#45 - llama_adapter v1 script can be trained on v100 32G?

Issue - State: open - Opened by zh25714 over 1 year ago - 1 comment

#44 - Why use 512 as the max sequence length for fine tuning alpaca?

Issue - State: open - Opened by tetratorus over 1 year ago - 2 comments

#43 - Loss does not decrease during v1 finetuning.

Issue - State: closed - Opened by Luzzer over 1 year ago - 2 comments

#42 - A question about adapter code

Issue - State: open - Opened by TheShy-Dream over 1 year ago - 1 comment

#41 - Runtime Error when running fine tuning script.

Issue - State: open - Opened by seelenbrecher over 1 year ago - 1 comment

#40 - look up the implement of multimodal adapter

Issue - State: closed - Opened by TheShy-Dream over 1 year ago - 6 comments

#39 - should we always take tanh on gate

Issue - State: open - Opened by dingran over 1 year ago - 3 comments

#38 - Questions about implementation of llama-adapter-v2's multi-modal ability and training

Issue - State: closed - Opened by PanQiWei over 1 year ago - 10 comments

#37 - KeyError: 'adapter_query.weight' on finetuned adaptor

Issue - State: open - Opened by dittops over 1 year ago - 2 comments

#36 - Inconsitency of learnable scale parameter between code and paper

Issue - State: closed - Opened by theAdamColton over 1 year ago - 2 comments

#35 - Fix hardcoded model embedding size in v1

Pull Request - State: open - Opened by Ar-Kareem over 1 year ago

#34 - Computing output likelihoods?

Issue - State: open - Opened by vishaal27 over 1 year ago - 1 comment

#33 - What is the minimum VRAM of a GPU required for training a model?

Issue - State: open - Opened by xuantoan0406 over 1 year ago - 1 comment

#32 - Finetune using V2

Issue - State: open - Opened by huzerD over 1 year ago - 1 comment

#31 - Update packaging

Pull Request - State: closed - Opened by jxtngx over 1 year ago - 1 comment

#30 - Proper comparison between adapter-tuning, lora-tuning, prompt-tuning, and prefix-tuning?

Issue - State: closed - Opened by jzhang38 over 1 year ago - 2 comments

#29 - potential avenues of size reduction.

Issue - State: open - Opened by Alignment-Lab-AI over 1 year ago - 2 comments

#28 - Quantization support

Issue - State: open - Opened by neuhaus over 1 year ago - 4 comments

#27 - Getting a OOM error when running on a 2xA100-80gb machine

Issue - State: open - Opened by Jameshskelton over 1 year ago - 1 comment

#26 - Can this work on a customer GPU?

Issue - State: open - Opened by davyuan over 1 year ago - 2 comments

#25 - Mac M1 Pro GPU compatibility

Issue - State: open - Opened by kirrukirru over 1 year ago - 1 comment

#24 - Visual Instruction model

Issue - State: closed - Opened by remixer-dec over 1 year ago - 10 comments

#23 - Chat 65b demo

Pull Request - State: closed - Opened by linziyi96 over 1 year ago

#22 - Errors throwns when finetuning

Issue - State: open - Opened by Petrichoeur over 1 year ago - 4 comments

#21 - which GPU could I used if I want to do alpaca_finetuning_v1 on a single GPU?

Issue - State: open - Opened by pengwei-iie over 1 year ago - 2 comments

#20 - Issues with fine-tuning on the ScienceQA dataset.

Issue - State: open - Opened by Gary3410 over 1 year ago - 5 comments

#19 - How to fine-tune with the trained alpaca adaptor as the starting point?

Issue - State: closed - Opened by alpayariyak over 1 year ago - 2 comments

#18 - Multi-image inputs to the model

Issue - State: open - Opened by vishaal27 over 1 year ago - 1 comment

#17 - what about gpuduring training? I use 16G*4, batch size to 1, and then OOM

Issue - State: open - Opened by pengwei-iie over 1 year ago - 2 comments

#16 - multi-gpu

Issue - State: open - Opened by pengwei-iie over 1 year ago - 1 comment

#15 - Alpaca finetuning issues

Issue - State: open - Opened by Gary3410 over 1 year ago - 3 comments

#14 - Error when running example.py

Issue - State: open - Opened by reddiamond1234 over 1 year ago - 3 comments

#13 - Training error occurred

Issue - State: closed - Opened by laosuan over 1 year ago - 3 comments

#12 - Question: CUDA synchronize when training

Issue - State: open - Opened by Jack-ZC8 over 1 year ago

#11 - Question about pytorch and cudatool version

Issue - State: closed - Opened by wendyhan2020 over 1 year ago - 2 comments

#10 - Results on more multimodal datasets

Issue - State: open - Opened by sachit-menon over 1 year ago - 1 comment

#9 - Question about the initialization of Adapter.

Issue - State: open - Opened by TitleZ99 over 1 year ago - 7 comments

#8 - Error while inference

Issue - State: open - Opened by satani99 over 1 year ago - 4 comments

#7 - When Training code will be Released? I have prepared the japanese Version after training on my dataset, I would like to see its efficiency.

Issue - State: open - Opened by ankur92009 over 1 year ago - 4 comments

#6 - Code for reproducing evaluation results on ScienceQA

Issue - State: open - Opened by TJKlein over 1 year ago - 9 comments

#5 - HuggingFace Models

Issue - State: open - Opened by slavakurilyak over 1 year ago - 4 comments

#4 - Question: Can the number of "adaptable" layers be significantly reduced? Can training be optimized for "extremely consumer" grade GPUs like the 3060 Ti (8GB VRAM)?

Issue - State: closed - Opened by Andrey36652 over 1 year ago - 3 comments

#3 - some questions about LoRA

Issue - State: open - Opened by suc16 over 1 year ago - 3 comments

GitHub / OpenGVLab/LLaMA-Adapter issues and pull requests