huggingface/peft

Issues

• 训练时使用的QLoRA 4rank，进行cuda模型合并导出时出现，KeyError: 'base_model.model.model.model.layers.14.mlp.down_proj'

  #2213 opened 9 days ago by xiaoheiyue
  14

• modules_to_save Incorrect Overlap in Multiple LoRA Adapters

  #2206 opened 14 days ago by saeid93
  3

• When I use peft to finetune llama2, the gpu memory keeps growing

  #2141 opened 5 days ago by xuanzhangyang
  7

• Bug: BOFT forward/merging with CUDA

  #2219 opened 5 days ago by BenjaminBossan
  1

• KeyError: Parameter containing

  #2205 opened 14 days ago by Amerehei
  6

• How to specify the coefficients of loading lora during inference?

  #2216 opened 5 days ago by laolongboy
  1

• KeyError: 'messages'

  #2204 opened 5 days ago by rickeyhhh
  9

• Prompt Tuning Crash with Llama-3.2 in torch.embedding

  #2161 opened a month ago by hrsmanian
  5

• update `layers_pattern` logic to to accept layer index

  #2165 opened a month ago by fcakyon
  7

• LoraConfig conflict when using `layers_to_transform` in `LlamaModel`

  #2155 opened a month ago by Evan02580
  4

• support for heterogeneous types for `modules_to_save`

  #2136 opened 8 days ago by saeid93
  5

• about run_unsloth_peft.sh

  #2152 opened a month ago by opentld
  5

• the lack of adapter_model.bin and adapter_config.json after fine-tuning

  #2211 opened 10 days ago by TracyGuo2001
  14

• Documentation for LoRAConfig.

  #2212 opened 10 days ago by brynhayder
  1

• TypeError: LoraConfig.__init__() got an unexpected keyword argument 'exclude_modules'

  #2208 opened 14 days ago by imrankh46
  25

• Unable to merge lora into base model properly?

  #2209 opened 11 days ago by hgftrdw45ud67is8o89
  2

• `PeftModelForCausalLM.generate` ignores prompt tuning parameters unless `use_cache=False`

  #2123 opened 12 days ago by mattlgarber
  2

• RuntimeError: element 0 of tensors.. OpenCLIP model

  #2200 opened 17 days ago by EngEmmanuel
  4

• Add Assertions for `task_type` in `LoraConfig`

  #2203 opened 15 days ago by d-kleine
  1

• could not finetune gemma 2 9b with lora and fsdp

  #2111 opened 16 days ago by imadoualid
  14

• PEFT doesn't inject virtual tokens into generate forward pass

  #2134 opened 16 days ago by Kami-chanw
  7

• Memory Inefficiency for LoRA & DoRA during fine-tuning.

  #2196 opened 20 days ago by gslama12
  3

• [BUG] Issue with using `rank_pattern` and `alpha_pattern` together in `LoraConfig`

  #2194 opened 22 days ago by sirluk
  2

• Strange GPU MEM Occupation on GPU0 when using torchrun

  #2193 opened 21 days ago by ma787639046
  9

• ”peft_prefix_tuning_seq2seq.ipynb“RuntimeError Due to Tensor Dimension Mismatch

  #2192 opened 22 days ago by 1hb6s7t
  2

• merge_and_unload docs do not clarify behaviour for quantized base models

  #2105 opened 2 months ago by RonanKMcGovern
  7

• Ineffective Fine-Tuning Bug: Using `get_peft_model()` Before Loading LoRA Produces Outputs Identical to the Base Model

  #2115 opened 24 days ago by Hoper-J
  4

• How to change 'modules_to_save' setting when reloading a lora finetuned model

  #2188 opened 24 days ago by dengchengxifrank
  1

• PiSSA Updates Base Model Weights Silently

  #2184 opened 24 days ago by njbrake
  2

• Integration of merge-kit into PEFT

  #2179 opened a month ago by ParagEkbote
  11

• Evaluation of peft models using lm-eval-harness

  #2182 opened a month ago by JINO-ROHIT
  4

• Tensor Expansion Size Mismatch During Forward Pass

  #2154 opened 25 days ago by VecherVhatuX
  2

• Xlora cannot reload model from last checkpoint by using trainer.train(resume_from_checkpoint="checkpp")

  #2185 opened 25 days ago by SongHanKen
  0

• Key mismatch when trying to load a LORA adapter into an XLORA model

  #2132 opened 2 months ago by p4arth
  4

• Lora PISSA init: not support gpt2

  #2103 opened a month ago by suyang160
  4

• How can I do to export mode format as gguf

  #2181 opened a month ago by xu756
  0

• LoRA + DeBERTa: loading model gives erratic, non-deterministic results

  #2171 opened a month ago by jchook
  12

• Invalid `none` check for loftq_config attribute in `LoraConfig`

  #2170 opened a month ago by sirluk
  5

• fsdp_auto_wrap_policy is not working when FSDP_TRANSFORMER_CLS_TO_WRAP and the model's _no_split_modules are None.

  #2166 opened a month ago by eljandoubi
  0

• Request to Include Named Entity Recognition and Relation Extraction Model Finetuning Examples and Guidance Request

  #2119 opened a month ago by HarikrishnanK9
  6

• Is it possible to support the new Bone method?

  #2138 opened a month ago by JL-er
  5

• Optimize DoRA computation when there is no dropout

  #2107 opened a month ago by BenjaminBossan
  2

• LoKrModel doesn't support LLM Model

  #2147 opened a month ago by WhuanY
  3

• RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu!

  #2129 opened a month ago by brunopistone
  8

• SFTTrainer: element 0 of tensors does not require grad and does not have a grad_fn

  #2125 opened a month ago by brunopistone
  2

• p-tuning or prefix tuning or prompt tuning code given does not work during inference

  #2131 opened a month ago by chandar-l
  5

• ValueError: Please specify `target_modules` in `peft_config`, issue exist with gemma

  #2128 opened 2 months ago by yananchen1989
  4

• Peft should update the way of passing past_key_values of prefix-tuning, prompt tuning etc

  #2121 opened 2 months ago by Kami-chanw
  1

• PEFT Config checking update request

  #2112 opened 2 months ago by lemingshen
  4

• Update `huggingface_hub` requirement version

  #2116 opened 2 months ago by fmartiescofet
  1