Dataset · argilla
Argilla curated Dolly 15k instruction dataset
argilla/databricks-dolly-15k-curated-en
Proofread version of the English Databricks Dolly 15k instruction set with fixed contexts and answers.
What's inside
The set contains 10 to 100 thousand English records, each including task category, instruction, context, and answer. The Argilla team manually cleaned the original dataset, correcting instruction formulations, inputs, and responses. The main goal of the editing is to make texts accurate, task-category compliant, and as concise as possible without losing completeness.
What you can do with it
- Quality SFT fine-tuning or LoRA training for English instruction following
- Evaluating model response quality in tasks with a given context
- Languages
- English
- Size
- 10K–100K rows
- Downloads this month
- 33.3K
- Likes
- 45
Models trained on it
- mradermacher/TinyLLama-v0-GGUF 128
- SystemAdmin123/tiny-random-LlamaForCausalLMText generation 86
- SystemAdmin123/pythia-160mText generation 55
- SystemAdmin123/llama-68mText generation 35
- SystemAdmin123/opt-125mText generation 33
- SystemAdmin123/tiny-llama-fast-tokenizerText generation 31
- SystemAdmin123/tiny-dummy-qwen2Text generation 26
- SystemAdmin123/SmolLM-360MText generation 25
How to load it
from datasets import load_dataset
ds = load_dataset("argilla/databricks-dolly-15k-curated-en")