4 Pull requests merged by 3 people

• Contextual keywords generation for RAG using Llama-3.1

  #889 merged Mar 26, 2025

• Bump tj-actions/changed-files from 41.0.0 to 45.0.8 in /.github/workflows

  #900 merged Mar 17, 2025

• format markdown faq

  #895 merged Mar 9, 2025

• update test for guard prefix

  #892 merged Mar 6, 2025

7 Pull requests opened by 5 people

• fix sharding strategy str to enum conversion

  #891 opened Mar 3, 2025

• Add SalesBot Weaviate Notebook

  #897 opened Mar 10, 2025

• Notebook showing how to fine tune llama guard with torchtune

  #898 opened Mar 11, 2025

• update HELM link

  #902 opened Mar 21, 2025

• Clean / update LangGraph Tutorial

  #905 opened Mar 26, 2025

• Google Calendar Assistant with with Llama 3.2 3B Tool Calling

  #906 opened Mar 27, 2025

• Update `hello_llama_cloud.ipynb` source

  #907 opened Mar 28, 2025

8 Issues closed by 5 people

• 🏦 mascot

  #903 closed Mar 22, 2025

• Llama-3.3-70B-Instruct Infernce speed too slow

  #901 closed Mar 21, 2025

• converting checkpoint from dist to hf fails

  #899 closed Mar 14, 2025

• Fine-tuning for small datasets

  #893 closed Mar 9, 2025

• No chat template for processor

  #845 closed Mar 5, 2025

• safetensors_rust.SafetensorError: Error while deserializing header: HeaderTooLarge

  #884 closed Mar 4, 2025

• Finetuning llama 3.1 7B models using FSDP+4-bit-quantization+PEFT not reducing GPU consumption

  #872 closed Mar 3, 2025

• continue pre-training Example

  #820 closed Mar 3, 2025

2 Issues opened by 2 people

• MultiGPU finetuning: AttributeError: 'Params4bit' object has no attribute 'absmax'

  #904 opened Mar 24, 2025

• Local LangGraph RAG agent with Llama 3 error

  #894 opened Mar 6, 2025

13 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• Batch Inference with Llama 3.2 Generate Function: Only the First Result is Correct

  #816 commented on Mar 4, 2025 • 0 new comments

• Is there a flag for using random weights to benchmark pre-training speed?

  #813 commented on Mar 4, 2025 • 0 new comments

• Minimum GPU VRAM requirement for vision model inference

  #815 commented on Mar 4, 2025 • 0 new comments

• custom function name for "get_custom_dataset" also updates the name of the function called to retreive a custom data loader.

  #828 commented on Mar 4, 2025 • 0 new comments

• Full fine-tuned model is not working, which LoRA fine-tuned model works fine.

  #836 commented on Mar 4, 2025 • 0 new comments

• FSDP not work well with PEFT and some error in mixed precision when use peft in fine-tune

  #837 commented on Mar 4, 2025 • 0 new comments

• Reproduction issue of task GSM8K with Llama3.2-1B-Instruct

  #810 commented on Mar 6, 2025 • 0 new comments

• Add flag to turn on activation checkpointing on single GPU

  #835 commented on Mar 6, 2025 • 0 new comments

• Missing `input_ids` Error When Going through Llama 3.2 Vision Models Fine-Tuning Recipe

  #812 commented on Mar 6, 2025 • 0 new comments

• FSDP sharding_strategy parameter fails with KeyError when passed as string

  #809 commented on Mar 6, 2025 • 0 new comments

• Adding a Vision RAG Notebook to Llama Recipes

  #781 commented on Mar 7, 2025 • 0 new comments

• Update wandb.py to accept setting run name from command line argument (e.g., --wandb_config.name "run_name") for fine tuning

  #772 commented on Mar 4, 2025 • 0 new comments

• update parsing of dataset_config.file to prevent custom-function-name from clobbering data-collator name.

  #829 commented on Mar 4, 2025 • 0 new comments