ToolBench Toolllama G123 Dfs: Processed Training Data for Tool-Using LLMs

Yhyu13 uploaded this dataset to Hugging Face on 2024-01 03. It contains processed JSON files for the ToolBench project, intended for training and evaluating language models on tool-use tasks. The data is derived from the original ToolBench repository and has been tailored for use with frameworks like LLaMA-Factory.

Use Cases

• Fine-tune language models for tool-use based on the processed instruction and API call sequences.
• Benchmark model performance on tool invocation tasks using the provided evaluation split.
• Adapt training data scale for resource-efficient experimentation, as mentioned in the description.

Strengths

• Data is pre-processed by the original author, ready for integration into the ToolBench framework.
• Includes separate files for training and evaluation, as specified in the command-line arguments.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

OpenBMB/ToolBench GitHub repository and associated Google Drive.

Collection Method

Processed from the original ToolBench data.zip.

Time Range

null

Freshness

Last updated 2024-01-03 09:58:25; freshness should be verified.

Geography

null