#####   ######  #####    ###    #   #  ###  #   #  ######
##  ##  ##      ##  ##  ## ##   #   #   #   #   #  ##
#####   ####    #####   #   #   #   #   #   #   #  ####
##  #   ##      ##      ## ##    # #    #    # #   ##
##   #  ######  ##       ###      #    ###    #    ######

$ curl repovive.com/roadmaps/llm-fine-tuning/fine-tuning-tools-frameworks/unsloth

░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░█████████████████████████████████████████████████████████████████████████████████

#####   ######  #####    ###    #   #  ###  #   #  ######
##  ##  ##      ##  ##  ## ##   #   #   #   #   #  ##
#####   ####    #####   #   #   #   #   #   #   #  ####
##  #   ##      ##      ## ##    # #    #    # #   ##
##   #  ######  ##       ###      #    ###    #    ######

$ curl repovive.com/roadmaps/llm-fine-tuning/fine-tuning-tools-frameworks/unsloth

Repovive

Unsloth - Fine-Tuning Tools & Frameworks | LLM Fine-Tuning | Repovive

Repovive.

LLM Fine-Tuning10 sections · 313 units1h total

Open in Course

Unsloth

Unsloth optimizes fine-tuning for speed. It rewrites attention and training kernels for $2$ x faster training with $60$ % less memory. Supports Llama, Mistral, Phi, Gemma, and Qwen.

from unsloth import FastLanguageModel
model, tokenizer = FastLanguageModel.from_pretrained(
    model_name="unsloth/llama-3-8b-bnb-4bit",
    max_seq_length=2048
)
model = FastLanguageModel.get_peft_model(model, r=16)

The API mirrors Hugging Face. If you're training on consumer GPUs, Unsloth is almost always worth using.