AGI-Edgerunners/LLM-Adapters

Issues

• Reproduce the commense results on Boolq

  #64 opened 8 months ago by Zhenyu001225
  23

• How to see the evalution results?

  #72 opened 2 months ago by gf457832386
  0

• Questions about the accuracy of eight commonsense reasoning datasets vs the Llama paper

  #70 opened 3 months ago by Yonghao-Tan
  1

• Baseline evaluation

  #71 opened 3 months ago by Yonghao-Tan
  0

• How to evaluate a model fine-tuned with prefix?

  #32 opened a year ago by heart-and-soul
  6

• Question on the source of commonsense_15k

  #69 opened 6 months ago by clarenceluo78
  3

• Question about datasets variants

  #66 opened 7 months ago by ZeguanXiao
  1

• Question about the reproducation of the results in the math_10k

  #58 opened 9 months ago by zeyuliu1037
  13

• Can not find BottleneckConfig

  #68 opened 7 months ago by 1148514800
  2

• Training loss goes to 0 and eval loss goes to nan

  #67 opened 7 months ago by ZeguanXiao
  5

• Errors when I run generation

  #36 opened a year ago by ChaoGaoUCR
  5

• about loss

  #65 opened 7 months ago by haoyuwangwhy
  1

• Problems I came across when I try to reprocude the results

  #37 opened a year ago by ChaoGaoUCR
  3

• Full-Parameter Fine-Tuning on commonsense

  #62 opened 7 months ago by lucasliunju
  16

• Gibberish output

  #63 opened 8 months ago by Aradhye2002
  6

• Possible Bug In Handling Batch Size During Common Sense Evaluation

  #61 opened 8 months ago by mchorton
  1

• AttributeError: 'tuple' object has no attribute 'update'

  #60 opened 8 months ago by YananLi18
  5

• Is there any way to evaluate models without any adapters?

  #59 opened 9 months ago by smkim0220
  0

• FT with bottleneck : cannot perform fine-tuning on purely quantized models

  #57 opened 9 months ago by Lao-yy
  2

• p-tuning in finetune.py?

  #56 opened 10 months ago by smkim0220
  0

• Can't fine tune/train when the model is loaded in 8bit

  #55 opened 10 months ago by Wonigox
  7

• 请问为如何两次加载不同的微调后生成的lora权重？

  #54 opened 10 months ago by jinlong7790
  1

• How to Reproduce BLOOMz-7b and GPT-j-6 results?

  #50 opened a year ago by Ocean-627
  1

• weird evaluation results: 0% accuracy

  #48 opened a year ago by wum67
  1

• ValueError: The version of PEFT you are using is not compatible, please use a version that is greater than 0.5.0

  #47 opened a year ago by nbasyl
  6

• Couldn't get the same accuracy with eight commonsense reasoning datasets.

  #38 opened a year ago by ello0211
  7

• How does chatglm support p-tuning in code?

  #53 opened a year ago by lyt719
  0

• Question regarding the source of math_10k.json

  #43 opened a year ago by HuangOwen
  10

• Upload evaluation outputs and adapters

  #46 opened a year ago by mkeoliya
  3

• Guidance Request for Reproducing OpenbookQA Dataset Results

  #49 opened a year ago by FairyFali
  1

• Details on provided peft

  #45 opened a year ago by aksh555
  2

• how to download the dataset

  #44 opened a year ago by ello0211
  1

• Questions about evaluate time

  #40 opened a year ago by Yuan0320
  4

• Training reproduce

  #41 opened a year ago by ChaoGaoUCR
  2

• Peft version problem

  #39 opened a year ago by marlin-codes
  2

• Couldn't get the same accuracy as the table (7B model LoRA)

  #26 opened a year ago by ywen666
  4

• finetune accuracy is much higher than what is in the README table

  #30 opened a year ago by CrazyElements
  4

• AdapterH, AdapterP code

  #35 opened a year ago by ChaoGaoUCR
  2

• Eval without Tuning/Using OPT-1.3B

  #34 opened a year ago by ChaoGaoUCR
  2

• [Bug] Lora finetuning memory keeps rising until it is Out Of Memory

  #33 opened a year ago by angelOnly
  1

• Could you give me an example to fintuning the Chatglm with adapter bottleneck please?

  #29 opened a year ago by zhaojunGUO
  0

• how to tune chatglm6b with dialogue dataset?

  #28 opened a year ago by zhaojunGUO
  0

• How to use llama-13B or bigger models?

  #27 opened a year ago by feiyuehchen
  1

• any code to merge the adapter weight with the original base model?

  #25 opened a year ago by sohuren
  1

• How to overwrite the Adapter

  #24 opened a year ago by YuChen17Heaven
  3

• Can not reproduce GSM8K zero-shot result

  #16 opened 2 years ago by simplelifetime
  4

• Questions about inconsistent results between the paper and the the README table.

  #21 opened 2 years ago by lpyhdzx
  4

• Question about the train and eval dataset for reporting the `Finetuned Result` table

  #20 opened 2 years ago by ToheartZhang
  2

• RuntimeError: "addmm_impl_cpu_" not implemented for 'Half'

  #19 opened 2 years ago by ljbigbang
  1

• About the fp16 parameter setting

  #17 opened 2 years ago by noob-ctrl
  1