60,000 tool-calling samples generated by Salesforce's APIGen pipeline across 3,673 executable APIs in 21 categories. Every sample passed three verification stages — format checking, actual function execution, and semantic verification — with >95% human-audited correctness. The quality benchmark for function-calling data, and the training set of the xLAM action-model family.
| Provider | Salesforce |
| Category | 智能体与函数调用 |
| Size | 60k Rows |
| License | CC BY 4.0 |
| Downloads | n/a |
| Tags | Function-Calling, APIGen, Execution-Verified, Agents, 2024 |
from datasets import load_dataset
ds = load_dataset("Salesforce/xlam-function-calling-60k")
使用 QLoRA(4-bit 基础模型 + LoRA 适配器)微调的预计显存需求(保守默认参数):
| 7B QLoRA | ~6GB VRAM |
| 13B QLoRA | ~10GB VRAM |