Modelpublic

mpasila/Llama-3.1-Discord-Short-8B

sourceHugging Facellama3.1updated 2y agoView on Hugging Face

1likes20downloads

Model Card

Trained on Discord chatlogs from this dataset.

Uses Llama 3.1 formatting.

LoRA: mpasila/Llama-3.1-Discord-Short-LoRA-8B

Trained with regular LoRA (not quantized/QLoRA) and LoRA rank was 128 and Alpha set to 32. Trained for 1 epoch using A40 for about 5,5 hours.

python

    args = UnslothTrainingArguments(
        per_device_train_batch_size = 1,
        gradient_accumulation_steps = 8,

        warmup_ratio = 0.1,
        num_train_epochs = 1,

        learning_rate = 5e-5,
        embedding_learning_rate = 5e-6,

        fp16 = not is_bfloat16_supported(),
        bf16 = is_bfloat16_supported(),
        logging_steps = 1,
        optim = "adamw_8bit",
        weight_decay = 0.00,
        lr_scheduler_type = "cosine",
        seed = 3407,
        output_dir = "outputs",
    ),

Uploaded model

—Developed by: mpasila
—License: Llama 3.1 Community License Agreement
—Finetuned from model : unsloth/Meta-Llama-3.1-8B

This llama model was trained 2x faster with Unsloth and Huggingface's TRL library.