Hugging Face's logo

blackhole33
/
experiments

PEFT
TensorBoard
Safetensors
trl
sft
Generated from Trainer
Model card Files Metrics Community
experiments / README.md
blackhole33's picture
blackhole33
Update README.md
8d8f1a9 verified
preview code
|
raw
history blame contribute delete
2.08 kB
metadata
base_model: meta-llama/Meta-Llama-3-8B-Instruct
library_name: peft
license: llama3
tags:
  - trl
  - sft
  - generated_from_trainer
model-index:
  - name: experiments
    results: []

experiments

This model is a fine-tuned version of meta-llama/Meta-Llama-3-8B-Instruct on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 1.4332

Model description 


MODEL_NAME = "/content/blackhole33/llama-5000-sample-peft"
quantization_config = BitsAndBytesConfig(
    load_in_4bit=True, bnb_4bit_quant_type="nf4", bnb_4bit_compute_dtype=torch.bfloat16
)

tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME, use_fast=True)

model = AutoModelForCausalLM.from_pretrained(
    MODEL_NAME, quantization_config=quantization_config, device_map="auto"
)

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 0.0001
  • train_batch_size: 2
  • eval_batch_size: 2
  • seed: 42
  • gradient_accumulation_steps: 4
  • total_train_batch_size: 8
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: constant
  • lr_scheduler_warmup_ratio: 0.1
  • num_epochs: 1
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss
1.3475 0.2 100 1.5142
1.4979 0.4 200 1.4703
1.4307 0.6 300 1.4510
1.3795 0.8 400 1.4434
1.3847 1.0 500 1.4332

Framework versions

  • PEFT 0.12.0
  • Transformers 4.44.1
  • Pytorch 2.3.1+cu121
  • Datasets 2.21.0
  • Tokenizers 0.19.1