Qrzysztof/functiongemma-270m-it-prepaid-cards-v2-mlx

FunctionGemma 270M IT — Prepaid Cards Tool-Calling (v2, MLX 8-bit)

Model description

MLX-optimized 8-bit conversion of Qrzysztof/functiongemma-270m-it-prepaid-cards-v2 (a google/functiongemma-270m-it fine-tune for prepaid-card tool calling in 107 languages with noisy/multi-turn input).

Runs natively and fast on Apple Silicon (M1/M2/M3/M4) with mlx-lm; 304 MB on disk, fits easily in 8 GB-RAM Macs.

Files

File	Description
model.safetensors (+ .index.json)	MLX weights, 8-bit quantized (≈8.5 bits/weight)
config.json	MLX config incl. quantization info
tokenizer.json, tokenizer_config.json	same tokenizer as the parent model
chat_template.jinja	FunctionGemma chat template

How to use

pip install mlx-lm
python3 -c "
from mlx_lm import load, generate
from mlx_lm.sample_utils import make_sampler
model, tokenizer = load('Qrzysztof/functiongemma-270m-it-prepaid-cards-v2-mlx')
prompt = '<bos><start_of_turn>developer...'   # tokenizer.apply_chat_template(messages, tools=tools, add_generation_prompt=True)
print(generate(model, tokenizer, prompt=prompt, max_tokens=96, sampler=make_sampler(temp=0.0)))
"

API note: mlx-lm ≥ 0.30 removed temperature= in favour of a sampler object (make_sampler(temp=0.0)).

Intended uses & limitations

Same as the parent model (see the SafeTensors card): synthetic data, uneven language quality, no backend — plus 8-bit quantization caveats.

How it was made

pip install mlx-lm
python3 -m mlx_lm convert --hf-path <hf_model_dir> -q --q-bits 8
# output lands in ./mlx_model/ (no --output-dir in mlx-lm 0.31)

Full-precision (bf16) MLX version: same command without -q.

Evaluation

Same prompts & greedy decoding as the other formats, over the held-out v2 test subset (N=40).

Format	Success rate
SafeTensors (reference)	40/40 = 100%
MLX 8-bit	40/40 = 100%

8-bit quantization showed no drop in tool-name selection on this subset; argument wording can differ marginally from bf16.

Fine-tuning from this model

This model was fine-tuned with the tutorial below; you can use it as the starting point for a new tool set (or fine-tune google/functiongemma-270m-it directly).

Fine-tuning tutorial

A complete, minimal fine-tune of a FunctionGemma-class model on this data (follows the official FunctionGemma fine-tuning guide).

1. Setup

pip install torch transformers trl datasets accelerate
huggingface-cli login   # accept the gemma license for google/functiongemma-270m-it

2. Load the dataset and normalize messages

The Hub dataset stores messages/tools as JSON strings (Arrow cannot infer the nested schema), and TRL's SFTTrainer needs a uniform struct schema, so normalize first:

import json
from datasets import load_dataset
from transformers import AutoModelForCausalLM, AutoTokenizer

def normalize_messages(msgs):
    out = []
    for m in msgs:
        n = {"role": m["role"], "content": m.get("content") or "", "name": None,
             "tool_call_id": m.get("tool_call_id"), "tool_calls": None}
        if m["role"] == "tool":
            n["name"] = m["content"]["name"]
            n["content"] = json.dumps(m["content"]["response"], ensure_ascii=False)
        if m.get("tool_calls"):
            n["tool_calls"] = [{"id": tc.get("id"), "type": tc.get("type", "function"),
                                "function": {"name": tc["function"]["name"],
                                             "arguments": json.dumps(tc["function"]["arguments"], ensure_ascii=False)}}
                               for tc in m["tool_calls"]]
        out.append(n)
    return out

def rows_to_dataset(rows):
    from datasets import Dataset
    return Dataset.from_list([{
        "messages": normalize_messages(r["messages"]),
        "tools": json.dumps(r["tools"], ensure_ascii=False),
    } for r in rows])

ds = load_dataset("Qrzysztof/ecommerce-chat-tool-calling", token=HF_TOKEN)["train"]
train_rows = [{"messages": json.loads(r["messages_json"]), "tools": json.loads(r["tools_json"])}
              for r in ds if r["split"] == "train"]
train_ds = rows_to_dataset(train_rows)

3. Train

import torch
from transformers import AutoModelForCausalLM
from trl import SFTConfig, SFTTrainer

model = AutoModelForCausalLM.from_pretrained("google/functiongemma-270m-it",
                                             dtype=torch.bfloat16, attn_implementation="eager")
tokenizer = AutoTokenizer.from_pretrained("google/functiongemma-270m-it")

trainer = SFTTrainer(
    model=model,
    args=SFTConfig(
        output_dir="functiongemma-ecommerce",
        max_length=1024,          # covers the longest sample + margin
        packing=False,            # keep tool calls intact (no cross-sample packing)
        num_train_epochs=3,
        per_device_train_batch_size=8,
        learning_rate=5e-5,
        lr_scheduler_type="constant",
        warmup_steps=50,
        bf16=True,                # or fp16 on non-Ampere GPUs
        eval_strategy="epoch",
        report_to="none",
    ),
    train_dataset=train_ds,
    processing_class=tokenizer,
)
trainer.train()

TRL applies the FunctionGemma chat template with the per-sample tools column; assistant_only_loss=True (default) masks everything but the model's own turns, so it learns to emit tool calls — not to copy the schema.

4. Evaluate (greedy success rate)

ok = 0
for item in test_rows:
    inputs = tokenizer.apply_chat_template(item["messages"][:-1], tools=item["tools"],
                                           add_generation_prompt=True, return_tensors="pt")
    out = model.generate(**inputs, max_new_tokens=256)
    output = tokenizer.decode(out[0][len(inputs["input_ids"][0]):], skip_special_tokens=False)
    expected = <expected tool name / args from expected_json>
    ok += expected-tool-in-output and no-other-tool-in-output

5. Push

trainer.push_to_hub("YOUR_USER/functiongemma-ecommerce")

Best practices

Data

• Keep noise digit-safe: never corrupt the values the model must extract (prices, ids). The noise.py engine skips any token containing digits.
• Use deterministic train/test splits (by template_id) and hold out whole languages + (for the e-commerce set) whole schemas — that is the only honest way to measure generalization.
• Balance the training subset per (language, intent) — cap the big buckets instead of letting English dominate.

Training

• packing=False for tool-calling data; packed sequences splice mid-call.
• max_length ≥ longest sample + a margin; ~1024 covers these datasets.
• Constant LR + short warmup (the official guide's defaults) work well.
• Upload a checkpoint to the Hub after every epoch — Colab VMs die mid-run, and the last good epoch is always recoverable.

Evaluation

• Always evaluate with greedy decoding for comparability across formats and runs.
• Score two things separately: tool-name selection and argument fidelity (query + every filter key:value pair).
• Compare every exported format (SafeTensors / GGUF / MLX / ONNX) on the same prompts — quantization changes results.

Deployment

• Validate tool arguments server-side before executing anything (a small model can garble a card number under heavy noise).
• In a live agent, follow the FunctionGemma full loop: model call → backend executes → tool response → model continues; never let the model see or emit secrets.
• For browser deployment use the fp16 ONNX file; for low-end hardware the Q8_0 GGUF or MLX 8-bit; for exact reference behavior the SafeTensors model.

Related

• Parent: ...-v2 (SafeTensors)
• Siblings: GGUF f16+Q8_0 · ONNX
• Dataset: ...-tool-calling-v2