mirror of
https://github.com/hanzoai/trolley-bench.git
synced 2026-10-30 15:00:49 +00:00
Zachary Landes
247bd2614b
* Packaged the benchmark's working copy with its frozen prompts, runners, raw answers, analysis and built page as committed evidence * Rewrote run-machine paths and left out share credentials, drafts and operator notes; PACKAGING.md lists every exclusion and change * Fixed the models.yaml generator listing rerun models twice, and let the Jev runner take its key from TYPESAFE_API_KEY
43 lines
2.2 KiB
Python
43 lines
2.2 KiB
Python
"""Nimble's documented model preparation (README, "Download the model"), with the v2 revision pinned.
|
|
|
|
It downloads the adapter and its pinned base, merges them once on the CPU in bf16, and records the
|
|
merged folder in .cache/nimble-model.json for the MLX scorer. Run from the nimble clone in its prep venv.
|
|
"""
|
|
import hashlib
|
|
import json
|
|
from pathlib import Path
|
|
|
|
from huggingface_hub import snapshot_download
|
|
|
|
repo = "bespokelabs/Bespoke-Nimble-9B-v2"
|
|
revision = "4b8c04d1ac2cea3e41e5e3c4d2130bcead2c0abe"
|
|
snapshot = Path(snapshot_download(repo, revision=revision, cache_dir=".cache/huggingface/hub"))
|
|
contract_file = snapshot / "schema_config.json"
|
|
contract = json.loads(contract_file.read_text()) if contract_file.exists() else {}
|
|
if contract:
|
|
from transformers import AutoTokenizer
|
|
from nimble.training.candidate_schema import validate_contract
|
|
validate_contract(contract, AutoTokenizer.from_pretrained(snapshot))
|
|
|
|
model_path = snapshot
|
|
if (snapshot / "adapter_config.json").exists():
|
|
import torch
|
|
from peft import PeftModel
|
|
from transformers import AutoTokenizer, Qwen3_5ForConditionalGeneration
|
|
|
|
base = Qwen3_5ForConditionalGeneration.from_pretrained(contract["model"], revision=contract["revision"], dtype=torch.bfloat16, device_map="cpu", cache_dir=".cache/huggingface/hub")
|
|
adapter = PeftModel.from_pretrained(base, snapshot)
|
|
merged = adapter.merge_and_unload(safe_merge=True)
|
|
model_path = Path(".cache/models") / ("nimble-9b-" + snapshot.name)
|
|
merged.save_pretrained(model_path)
|
|
(model_path / "schema_config.json").write_text(json.dumps(contract, indent=2))
|
|
AutoTokenizer.from_pretrained(snapshot).save_pretrained(model_path)
|
|
(model_path / "READY.json").write_text(json.dumps({
|
|
"model": repo, "revision": snapshot.name,
|
|
"adapter_sha256": hashlib.sha256((snapshot / "adapter_model.safetensors").read_bytes()).hexdigest(),
|
|
}, indent=2))
|
|
|
|
config = {"model_path": str(model_path.resolve()), "model_id": repo, "revision": snapshot.name, "max_input_tokens": contract.get("max_length", 2048)}
|
|
Path(".cache/nimble-model.json").write_text(json.dumps(config, indent=2))
|
|
print("Ready:", model_path, json.dumps({"base": contract.get("model"), "base_revision": contract.get("revision")}))
|