h2oai/h2o-llmstudio

Issues

• [BUG] Default model is added twice to the LLM Backbone list

  #738 opened 3 days ago by pascal-pfeiffer
  0

• [BUG] Deepseek tokenizer error on main (worked before)

  #743 opened 4 days ago by tmostak
  4

• [FEATURE] Add (experimental) FP8 support

  #748 opened 5 days ago by psinger
  0

• Compare Zero-Epoch Prediction with Fine-Tuned Prediction as well as Validation Score Comparison

  #695 opened a month ago by meganjkurka
  1

• [DOCS] Duplicate Questions in the FAQ's

  #734 opened 6 days ago by meganjkurka
  0

• [FEATURE] Freezing layers

  #727 opened 9 days ago by psinger
  0

• [BUG] Memory allocation left resident in GPU(s) after model upload to HuggingFace

  #736 opened 9 days ago by tmostak
  0

• [FEATURE] Connection with LLM DataStudio

  #735 opened 10 days ago by meganjkurka
  1

• [FEATURE] Select multiple training dataframes

  #733 opened 11 days ago by psinger
  0

• [FEATURE] Implement SimPO

  #732 opened 12 days ago by psinger
  0

• [UX] Screen hangs when you click Download Model

  #728 opened 13 days ago by meganjkurka
  0

• Can't access meta-llama/Llama-2-7b using llmstudio cli tool

  #725 opened 14 days ago by sunilswain
  1

• [FEATURE] Add ability to specify dataset problem type during data import

  #690 opened 16 days ago by MartinBarus
  1

• [BUG] Code rendering in the validation prediction insights replaces characters

  #722 opened 17 days ago by pascal-pfeiffer
  0

• [BUG] HuggingFace export does not preserve bfloat16 weights but converts to float16 silently when using CPU for upload

  #702 opened 20 days ago by tmostak
  9

• [CODE IMPROVEMENT] Custom HF model for classification

  #713 opened 20 days ago by psinger
  0

• [FEATURE] add support for multigpu, splitting model across gpus without using deepspeed/fsdp

  #710 opened 22 days ago by Quetzalcohuatl
  3

• [FEATURE] Add DoRA: Weight-Decomposed Low-Rank Adaptation

  #645 opened 22 days ago by pascal-pfeiffer
  1

• Data Format section has a broken link

  #701 opened a month ago by cemremengu
  4

• [FEATURE] Option to plot train/eval plots with epoch instead of step on x-axis

  #700 opened a month ago by tmostak
  0

• [FEATURE] Support for minimum learning rate

  #671 opened 2 months ago by tmostak
  1

• [FEATURE] Use local LLM deployment as Judge

  #694 opened a month ago by pascal-pfeiffer
  1

• [FEATURE] Option for not saving checkpoint

  #691 opened a month ago by psinger
  1

• ValueError: invalid literal for int() with base 10: ‘Failed to initialize NVML: Unknown Error’

  #668 opened 2 months ago by jldroid19
  8

• [FEATURE] Option to add BOS token

  #636 opened 3 months ago by psinger
  4

• [BUG] Chat window generation parameters not updated

  #679 opened a month ago by psinger
  0

• [FEATURE] Random validation sample for chat interface

  #683 opened 2 months ago by psinger
  0

• [CODE IMPROVEMENT] Use build-in H2O Wave Chatbot features

  #630 opened 2 months ago by pascal-pfeiffer
  0

• [BUG] UI freezes when using "Stop streaming" button with a text in the input box

  #674 opened 2 months ago by pascal-pfeiffer
  0

• [FEATURE] Fine-tune CohereForCausalLM Models

  #677 opened 2 months ago by pascal-pfeiffer
  0

• [BUG] Exporting / downloading model larger, than VRAM available (trained with DeepSpeed) fails

  #670 opened 2 months ago by AZ777xx
  8

• [CODE IMPROVEMENT] Default for max_time

  #653 opened 2 months ago by psinger
  0

• [FEATURE] Mixed Precision Dtype

  #672 opened 2 months ago by psinger
  0

• [BUG] Mixed precision not working with bfloat16

  #628 opened 2 months ago by maxjeblick
  2

• [CODE IMPROVEMENT] Avoid gradient sync during gradient accumulation

  #664 opened 2 months ago by rohitgr7
  0

• [CODE IMPROVEMENT] Sort data files alphabetically

  #665 opened 2 months ago by psinger
  0

• [BUG] Pipenv missing as a requirement for the `make llmstudio` command

  #669 opened 2 months ago by pascal-pfeiffer
  1

• [CHORE] Upgrade dependencies to fix vulnerabilities

  #656 opened 2 months ago by pascal-pfeiffer
  0

• [FEATURE] Add additional digit of precision for specifying learning rate and other parameters

  #661 opened 2 months ago by tmostak
  1

• [FEATURE] Add danube2 to default model list

  #658 opened 2 months ago by pascal-pfeiffer
  0

• [BUG] Scheduler should consider gradient accumulation while assigning `epoch_steps`?

  #663 opened 2 months ago by rohitgr7
  0

• [CODE IMPROVEMENT] Flash_attn installation may be wrong if the wheel is cached

  #651 opened 2 months ago by pascal-pfeiffer
  0

• [FEATURE] Use Free Form Chat Template to Define Input Styling

  #654 opened 2 months ago by pascal-pfeiffer
  0

• [BUG] Tokenizer config has add_bos_token=true while LLM Studio is training with add_special_tokens=False

  #644 opened 3 months ago by pascal-pfeiffer
  1

• [BUG] Downloading a fine-tuned model fails with error

  #639 opened 3 months ago by vgorovoy
  2

• [DOCS] Adding the description syntax

  #621 opened 3 months ago by shaunyogeshwaran
  0

• [BUG] Error when pushing model to HuggingFace

  #635 opened 3 months ago by jeffwang0516
  2

• [CODE IMPROVEMENT] Resolve High Vulnerabilities

  #633 opened 3 months ago by codyharris-h2o-ai
  1

• [FEATURE] Training with QLoRA + FSDP

  #631 opened 3 months ago by pascal-pfeiffer
  0

• [BUG] Recent regression causing an error when loading tokenizer for Deepseek models

  #623 opened 3 months ago by tmostak
  2