mrm8488's picture
Update README.md
52bd0aa
metadata
license: other
base_model: microsoft/phi-2
tags:
  - generated_from_trainer
model-index:
  - name: phi2-ft-no_robots
    results: []
library_name: peft

phi2-ft-no_robots

This model is a fine-tuned version of microsoft/phi-2 on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 2.0917

logo

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

The following bitsandbytes quantization config was used during training:

  • quant_method: bitsandbytes
  • load_in_8bit: True
  • load_in_4bit: False
  • llm_int8_threshold: 6.0
  • llm_int8_skip_modules: None
  • llm_int8_enable_fp32_cpu_offload: False
  • llm_int8_has_fp16_weight: False
  • bnb_4bit_quant_type: fp4
  • bnb_4bit_use_double_quant: False
  • bnb_4bit_compute_dtype: float32

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 2.5e-05
  • train_batch_size: 2
  • eval_batch_size: 8
  • seed: 66
  • gradient_accumulation_steps: 32
  • total_train_batch_size: 64
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 1

Training results

Training Loss Epoch Step Validation Loss
2.2187 0.15 20 2.1602
2.1689 0.29 40 2.1049
2.1977 0.44 60 2.0958
2.1587 0.59 80 2.0910
2.0382 0.74 100 2.0920
2.1622 0.88 120 2.0917

Framework versions

  • PEFT 0.4.0
  • Transformers 4.37.0.dev0
  • Pytorch 2.1.0+cu121
  • Datasets 2.16.0
  • Tokenizers 0.15.0