Instructions to use Jakevin/d3-mini-ternary-mlx with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use Jakevin/d3-mini-ternary-mlx with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] hf download Jakevin/d3-mini-ternary-mlx --local-dir d3-mini-ternary-mlx
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
File size: 2,632 Bytes
0823df0 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 | #!/usr/bin/env python3
"""Score text questions with this MLX checkpoint.
The vision tower is stored at full precision in ``model.safetensors`` and is not
used here. Load requires ``mlx``, ``mlx-lm`` and ``transformers``.
"""
from __future__ import annotations
import json
import sys
from pathlib import Path
sys.path.insert(0, str(Path(__file__).resolve().parent))
import d3_mlx # noqa: E402
EXAMPLE = {
"state": "The order arrived damaged yesterday. The customer has a receipt and asks for a replacement today.",
"questions": {
"route": {
"type": "choice",
"instructions": "Which team should handle this request?",
"criteria": {
"returns": "Refunds, replacements and damaged deliveries",
"billing": "Payments, invoices and charges",
"technical": "Product setup and faults",
},
},
"receipt": {"type": "noul", "instructions": "Does the customer have a receipt?"},
"urgency": {
"type": "score",
"instructions": "How urgent is this request?",
"criteria": ["Routine", "Soon", "Today"],
},
},
}
def option_labels(question: dict) -> list[str]:
rendered = d3_mlx.normalize_question(question)
if rendered["type"] == "noul":
return ["false", "true"]
return list(rendered["criteria"])
def predict(model, readout, tokenizer, codes, state, questions: dict) -> dict:
answers = {}
for key, question in questions.items():
text, n_options = d3_mlx.render_prompt(tokenizer, state, question, codes)
token_ids = tokenizer.encode(text, add_special_tokens=False)
probabilities = d3_mlx.probs_from_ids(model, readout, token_ids, n_options)
labels = option_labels(question)
pairs = {label: float(value) for label, value in zip(labels, probabilities)}
best = max(pairs, key=pairs.get)
answers[key] = {"best": best, "probabilities": pairs}
return answers
def main() -> None:
from transformers import AutoTokenizer
root = Path(__file__).resolve().parent
model, _config, decision, readout = d3_mlx.load_model(root)
tokenizer = AutoTokenizer.from_pretrained(str(root))
payload = EXAMPLE
if len(sys.argv) > 1:
payload = json.loads(Path(sys.argv[1]).read_text(encoding="utf-8"))
answers = predict(
model,
readout,
tokenizer,
decision["codes"],
payload["state"],
payload["questions"],
)
print(json.dumps(answers, ensure_ascii=False, indent=2))
if __name__ == "__main__":
main()
|