Crow 8B: Cleaned Training Data for Language Model Fine-Tuning

615,000 tokens of cleaned text data used for training the Crow 8B language model. The dataset was created by Crownelius and last updated on Hugging Face in March 2026. It consists of prompt-completion pairs with an average of 6.65 tokens per row.

Use Cases

• Fine-tuning language models based on the provided prompt-completion structure.
• Benchmarking model performance on text generation tasks using the described token counts.
• Studying data cleaning techniques for AI training datasets as suggested by the title.
• Analyzing cost efficiency for model training using the provided token and cost metrics.

Strengths

• Contains 615,000 tokens of cleaned text data.
• Provides specific cost metrics, estimating a total cost of $3.08 USD for generation.
• Has a consistent structure with an average of 1.00 turns and 6.65 tokens per row.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count, file formats, and license information are unknown.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

huggingface

Collection Method

Likely generated or cleaned for training the Crow 8B model.

Freshness

Last updated 2026-03-15 07:03:32; freshness should be verified.