zephyr-quiklang-3b / README.md
Walmart-the-bag's picture
Update README.md
bfe8891
|
raw
history blame
948 Bytes
metadata
license: other
pipeline_tag: text-generation
tags:
  - causal_lm
datasets:
  - teknium/openhermes
  - unalignment/toxic-dpo-v0.1
base_model: stabilityai/stablelm-zephyr-3b
inference: false
metrics:
  - bleu
  - rouge

Model Description

This is a finetune of StableLM-Zephyr-3B with 2 datasets, toxic-dpo and openhermes with 10000 samples. This was finetuned at 1024 context, for 4k version, go here: https://huggingface.co/Walmart-the-bag/zephyr-quiklang-3b-4K.

Training Parameters

  • 1xA6000-48GB
  • batch_size: 6
  • learning_rate: 5e-5

Datasets:

  • unalignment/toxic-dpo-v0.1
  • teknium/openhermes

Metrics/Basic Eval:

"predict_bleu-4": 31.594154999999997,
"predict_rouge-1": 44.092935,
"predict_rouge-2": 22.276081000000005,
"predict_rouge-l": 34.506909,
"predict_runtime": 121.7549,
"predict_samples_per_second": 0.821,
"predict_steps_per_second": 0.107