Ecosyste.ms: Issues

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

GitHub / artidoro/qlora issues and pull requests

#292 - Error when loading model

Issue - State: open - Opened by m000lie 7 months ago - 3 comments

#145 - Loading Lora Adapter weights into 4bit model to continue fine tuning

Issue - State: closed - Opened by simsim314 over 1 year ago - 5 comments

#100 - Multiplication error

Issue - State: open - Opened by two-tiger over 1 year ago - 6 comments

#100 - Multiplication error

Issue - State: open - Opened by two-tiger over 1 year ago - 6 comments

#98 - Errors happen during loading llama 65B for tuning.

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

#98 - Errors happen during loading llama 65B for tuning.

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

#97 - Is there anyone who is running this successfully?

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

#97 - Is there anyone who is running this successfully?

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

#96 - Multi-GPU Training

Issue - State: open - Opened by wcy1122 over 1 year ago - 17 comments

#96 - Multi-GPU Training

Issue - State: open - Opened by wcy1122 over 1 year ago - 17 comments

#94 - Discussion on formula (4), i.e. quantile quantization

Issue - State: closed - Opened by Sapium59 over 1 year ago - 7 comments

#94 - Discussion on formula (4), i.e. quantile quantization

Issue - State: closed - Opened by Sapium59 over 1 year ago - 7 comments

#92 - How to load LLaMA65B without enough GPU memory?

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

#91 - How to use the fine-turned model?

Issue - State: closed - Opened by tulunlxj2017 over 1 year ago - 1 comment

#91 - How to use the fine-turned model?

Issue - State: closed - Opened by tulunlxj2017 over 1 year ago - 1 comment

#89 - Question : what loss value is good value ?

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

#87 - add optional epoch stop

Pull Request - State: closed - Opened by dred0n over 1 year ago - 2 comments

#87 - add optional epoch stop

Pull Request - State: closed - Opened by dred0n over 1 year ago - 2 comments

#86 - add support local training data

Pull Request - State: closed - Opened by dred0n over 1 year ago - 1 comment

#86 - add support local training data

Pull Request - State: closed - Opened by dred0n over 1 year ago - 1 comment

#85 - ValueError:paged_adamw_32bit is not a valid OptimizerNames

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

#85 - ValueError:paged_adamw_32bit is not a valid OptimizerNames

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

#84 - Loss spike during training phase

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

#84 - Loss spike during training phase

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

#83 - how to load safetensors model checkpoint?

Issue - State: open - Opened by shawei3000 over 1 year ago

#82 - RuntimeError: CUDA error: an illegal memory access was encountered

Issue - State: open - Opened by flaviadeutsch over 1 year ago - 6 comments

#81 - How do I tuning llama 65B with 6000 ADA?

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

#80 - demo error

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

#79 - Bug Fix: Resume training from checkpoints

Pull Request - State: closed - Opened by KKcorps over 1 year ago - 6 comments

#78 - training/eval loss doesn't decrease when using paged_adamw_8bit

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

#77 - The time needed for finetuning a model is inconsistant with the reported result

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

#75 - [Feature request] Add custom dataset compatibility

Issue - State: open - Opened by zohfur over 1 year ago - 6 comments

#73 - Error while trying to run training in Windows

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

#71 - V100 can not supprt load_in_4bit and fp16?

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

#70 - memory leak?

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

#69 - Model finished training, but adapter_model.bin is empty?

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

#68 - Multi GPU inference example

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

#67 - Trained model output seems illegible

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

#66 - Update eval_gpt_review.py

Pull Request - State: open - Opened by mzamini92 over 1 year ago

#65 - Core dump. Not sure if caused by earlier problem with pad_token

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

#64 - Only cpu ram getting used...

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

#63 - Suppress pad_token warning message

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

#62 - Syntax/Logic error? pad_token is used before it is defined.

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

#61 - .

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

#60 - Update EVAL_README.md

Pull Request - State: closed - Opened by mohamed20050508 over 1 year ago - 3 comments

#59 - Open_Llama compatibility

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

#58 - Set per_device_eval_batch_size in finetune.sh

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

#56 - Which specific checkpoints are supported?

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

#55 - Finetuned T5 checkpoints

Issue - State: open - Opened by i-am-neo over 1 year ago - 1 comment

#54 - try finetune guanaco-33b-merged with default params and some problems

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

#52 - EleutherAI/gpt-j-6b not supported

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

#51 - Fix link to inference notebook

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

#50 - Can't resume from checkpoint

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

#49 - OverflowError: out of range integral type conversion attempted

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

#46 - Can wav2vec2 be finetuned?

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

#44 - Bug Fix: 443 Bytes `adapter_model.bin` files

Pull Request - State: closed - Opened by KKcorps over 1 year ago - 5 comments

#43 - RuntimeError: self and mat2 must have the same dtype

Issue - State: closed - Opened by baibaiw5 over 1 year ago - 4 comments

#41 - lora weights are not saved correctly

Issue - State: open - Opened by taishan1994 over 1 year ago - 10 comments

#40 - TypeError: __init__() got an unexpected keyword argument 'load_in_4bit'

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

#38 - Adapter model is just 400 bytes when using finetune.py

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

#37 - Can the quantized model run on multiple GPUs?

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

#36 - live demo always error, when type in chinese.

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

#35 - How can I run this repo across multiple machines?

Issue - State: open - Opened by lmc8133 over 1 year ago

#34 - pip install -q -U bitsandbytes

Issue - State: open - Opened by 2662007798 over 1 year ago

#33 - The VRAM usage is more than 48GB.

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

#32 - 4bit inference is slow

Issue - State: open - Opened by yangjianxin1 over 1 year ago - 18 comments

#31 - undefined symbol: cquantize_blockwise_fp16_fp4

Issue - State: open - Opened by DamonGuzman over 1 year ago - 11 comments

#30 - RecursionError: maximum recursion depth exceeded

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

#29 - Cannot merge LORA layers when the model is loaded in 8-bit mode

Issue - State: open - Opened by yangjianxin1 over 1 year ago - 26 comments

#28 - LORA Merge fails in 4-bit mode

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

#27 - Cannot resume from checkpoint because it is not detected as valid

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

#26 - Problems with bitsandbytes

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

#25 - How do you process oasst1 to get 9209 examples

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

#24 - V100 not support int4 and bf16?

Issue - State: open - Opened by shibing624 over 1 year ago - 2 comments

#23 - Getting error while trying to replicate

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

#22 - Add sophia optimizer

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

#21 - Fine-tuning with unlabelled data? (Causal language modelling)

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

#20 - Check for LlamaTokenizerFast rather than infer type from path name.

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

#19 - guanaco-13b Model fails on Google Colab Free tier T4 GPU

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

#17 - finetune.py 65b on A6000 48GB crashes with OOM

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

#16 - guanaco-7B-demo-colab.ipynb breaks with 4bit

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