philschmid/deep-learning-pytorch-huggingface

Issues

• Fine-tune-llm-in-2024-with-trl.ipynb for LLAMA3.2

  #65 opened 3 months ago by anas-zafar
  0

• ValueError: Must flatten tensors with uniform dtype but got torch.bfloat16 and torch.float32

  #61 opened 4 months ago by daje0601
  0

• OOM error in FSDP QLORA setup

  #60 opened 4 months ago by ss8319
  0

• Regarding the OOM issues with fine-tuning Flan-T5-xl

  #58 opened 6 months ago by Alsodream
  0

• Not able to run training/fsdp-qlora-distributed-llama3.ipynb

  #55 opened 7 months ago by aasthavar
  7

• Quantization question:

  #56 opened 7 months ago by aptum11
  0

• Deprecation warnings.

  #52 opened 8 months ago by hohoCode
  0

• Fine-tune-llm-in-2024-with-trl.ipynb not producing the outputs

  #51 opened 8 months ago by scigeek72
  0

• Out of Memory: Cannot reproduce T5-XXL run on 8xA10G.

  #49 opened 10 months ago by slai-natanijel
  3

• What's the use of "messages" in dpo step?

  #48 opened 10 months ago by katopz
  0

• question about DeepSpeedPeftCallback

  #47 opened 10 months ago by mickeysun0104
  0

• Re. fine-tune-llms-in-2024-with-trl.ipynb

  #45 opened 10 months ago by andysingal
  1

• Target modules all-linear not found in the base model.

  #43 opened a year ago by kassemsabeh
  6

• Does deepspeed partition the model to multi GPUs?

  #15 opened 2 years ago by vikki7777
  4

• Instruction tuning of LLama2 is significantly slower compared to documented 3 hours fine-tuning time on A10G.

  #35 opened a year ago by mlscientist2
  1

• flash attention error on instruction tune llama-2 tutorial on Sagemaker notebook

  #40 opened a year ago by matthewchung74
  2

• CUDA OOM error while saving the model

  #16 opened 2 years ago by aasthavar
  10

• Precision Issue

  #39 opened a year ago by zihaohe123
  4

• Does this work for Llama2 - Fine-tune Falcon 180B with DeepSpeed ZeRO, LoRA & Flash Attention?

  #37 opened a year ago by ibicdev
  11

• Falcon-180B "forward() got an unexpected keyword argument 'position_ids'"

  #38 opened a year ago by aittalam
  0

• Compute metrics while using SFT trainer

  #34 opened a year ago by shubhamagarwal92
  1

• Cannot load tokenizer for llama2

  #33 opened a year ago by smreddy05
  1

• LLama 2 Flash Attention Patch Not Working For 70B

  #32 opened a year ago by mallorbc
  6

• question about the llama instruction code

  #28 opened a year ago by yeontaek
  8

• Is the DataCollator necessary in peft-flan-t5-int8-summarization.ipynb ?

  #29 opened a year ago by brooksbp
  0

• compute_metrics() function

  #3 opened 2 years ago by ybagoury
  1

• How to create a json file for create_flan_t5_cnn_dataset.py

  #25 opened a year ago by andysingal
  1

• Llama patch for FlashAttention support fails with use_cache

  #26 opened a year ago by qmdnls
  2

• gcc/cuda used for training

  #24 opened a year ago by danyaljj
  1

• CPU offload when not using offload deepspeed config file

  #19 opened 2 years ago by siddharthvaria
  3

• Colab notebook fails

  #17 opened 2 years ago by TzurV
  1

• Error when training peft model example

  #18 opened 2 years ago by Tachyon5
  6

• FLAN-T5 XXL using DeepSpeed fits well for training but gives OOM error for inference.

  #12 opened 2 years ago by irshadbhat
  2

• Inference on CNN validation set takes 2+ hours on p4dn.24xlarge machine with 8 A100s, 40GB each

  #13 opened 2 years ago by sverneka
  5

• ValueError

  #14 opened 2 years ago by Martok10
  4

• Sample inference script for FLAN-T5 XXL using DeepSpeed & Hugging Face.

  #11 opened 2 years ago by irshadbhat
  7

• Error when finetuning Flan-T5-XXL on custom dataset

  #10 opened 2 years ago by ngun7
  1

• Error (return code -7) when finetuning FLANT5-xxl on 8* A100

  #7 opened 2 years ago by scofield7419
  3

• OOM when finetuning FLANT5-xxl

  #6 opened 2 years ago by AndrewZhe
  4

• Chat Inference Code

  #4 opened 2 years ago by samarthsarin
  2

• Problem with preprocess_function() in tutorial

  #2 opened 2 years ago by ybagoury
  4