jddqd/xlam-function-calling-60k-sft

xLAM function calling, prepared for SFT

Derived from Salesforce/xlam-function-calling-60k (revision 26d14ebfe18b1f7b524bd39b404b50af5dc97866), licensed CC-BY-4.0 by Salesforce. Please cite the original APIGen / xLAM work when using this data.

Each line is a single-turn conversation: messages in the Hugging Face chat format, tools in the exact function-doc format the BFCL evaluation puts in the prompt. Usable with tokenizer.apply_chat_template(messages, tools=tools):

{
  "messages": [
    {
      "role": "user",
      "content": "What is the weather in Paris?"
    },
    {
      "role": "assistant",
      "content": "",
      "tool_calls": [
        {
          "type": "function",
          "function": {
            "name": "get_weather",
            "arguments": {
              "city": "Paris"
            }
          }
        }
      ]
    }
  ],
  "tools": [
    {
      "name": "get_weather",
      "description": "Get the current weather of a city. Note that the provided function is in Python 3 syntax.",
      "parameters": {
        "type": "dict",
        "properties": {
          "city": {
            "type": "string",
            "description": "The city name."
          },
          "unit": {
            "type": "string",
            "description": "celsius or fahrenheit.",
            "default": "celsius"
          }
        },
        "required": [
          "city"
        ]
      }
    }
  ]
}

Processing

1. Tools converted to BFCL's function-doc format: no {"type": "function"} wrapper, BFCL's Python-3 hint appended to descriptions, parameters as a dict schema with properties / required and BFCL type names (string, integer, float, boolean, array, tuple, dict). Rows using types without an equivalent (Callable) are dropped. A parameter is optional if its type says so or it has a default. Defaults of the wrong type are converted when possible ("250" → 250), otherwise removed.
2. Calls validated against the tools: no unknown function, no unknown argument, no missing required argument.
3. Unicode NFC normalization and whitespace stripping.
4. Exact deduplication on (question, answer). Decontaminated against the BFCL questions: dropped rows whose user message is a BFCL question or shares 8-grams covering at least 50% of the shorter text.
5. Tool order shuffled (in the source, the called tool is listed first in 49% of rows).
6. Rows longer than 2048 tokens with the Qwen/Qwen3-0.6B-Base chat template dropped (never truncated).
7. Train / validation split (3%) stratified by BFCL-style category, seed 0.

step	rows
source rows	60,000
valid (converted + calls checked)	59,152
after exact dedup (question + answer)	57,537
after decontamination against BFCL	57,373
within max length	57,348
train	55,627
validation	1,721

rejection reason	rows
duplicate	1,615
unsupported_type	593
missing_required_argument	255
contaminated	164
too_long	25

rejected.jsonl contains every dropped source row with its reason.

Statistics

category	train	validation
parallel_multiple	19,451	602
multiple	18,004	557
parallel	10,569	327
simple	7,603	235

Train tokens (Qwen/Qwen3-0.6B-Base chat template): mean 545.2, median 509.0, p99 1271.72, max 2012, total 30,327,357.