Hanzo AI
cd6a0c79c5
Hand-written package for the engine /v1/training API: ServiceClient + TrainingClient (create/list/get/delete/forward_backward/optim_step/ sample/save_weights), dataclasses, sync httpx, completed-future .result() wrapper so tinker fut.result() code ports 1:1. 22 pytest tests over httpx.MockTransport, no network. Wired like siblings: publish-pypi hanzo-train-* tag case, rye+uv workspace member, uv source, pyright exclude. uv.lock relock also catches the lockfile up to on-disk truth (hanzoai 3.1.1, hanzo-tools-browser 0.5.10) -- no package versions changed.
hanzo-train
Tinker-shaped Python client for the Hanzo Engine training API.
It mirrors the shapes of Thinking Machines' tinker
SDK, so training loops written against tinker port across unchanged, while using
Hanzo-canonical field names. The transport is synchronous httpx with no retries;
the four training ops return a completed future exposing .result(timeout=None)
so fut.result() code ports 1:1.
Install
pip install hanzo-train
Quickstart
from hanzo_train import ServiceClient, LoraConfig, AdamParams, SamplingParams, Datum, ModelInput
sc = ServiceClient(base_url="http://localhost:1234", api_key=None) # api_key -> Authorization: Bearer
tc = sc.create_lora_training_client(
"HuggingFaceTB/SmolLM2-135M", lora_config=LoraConfig(rank=16), wait=True, timeout=600.0
)
# wait=True polls the client until status is ready (raises on failed); wait=False returns immediately.
out = tc.forward_backward([
{"prompt": "2+2=", "completion": "4"},
Datum(model_input=ModelInput(tokens=[1, 2, 3]), target_tokens=[2, 3, 4], weights=[0.0, 1.0, 1.0]),
]).result()
# out.loss, out.num_tokens, out.metrics
tc.optim_step(AdamParams(lr=1e-4)).result()
resp = tc.sample(
prompt="2+2=", sampling_params=SamplingParams(max_tokens=8, temperature=0.0), num_samples=1
).result()
# resp.sequences[0].tokens / .text
saved = tc.save_weights_and_get_sampling_client(name="my-adapter").result() # .path, .format == "peft"
# alias: tc.save_weights(name="my-adapter")
info = tc.get_info() # TrainingClientInfo incl. loss_history
sc.list_training_clients() # list[TrainingClientInfo]
tc.delete()
Errors
Every non-2xx response raises HanzoTrainError(status, message):
400— bad input404— unknown training client id409— client still loading, or failed to load
API
ServiceClient(base_url, api_key=None)
create_lora_training_client(base_model, lora_config=None, wait=True, timeout=600.0, poll_interval=1.0) -> TrainingClientlist_training_clients() -> list[TrainingClientInfo]close()
TrainingClient
forward_backward(data) -> Future[ForwardBackwardResult]—dataitems are either{"prompt": str, "completion": str}dicts orDatumvalues.optim_step(adam_params=None) -> Future[OptimResult]sample(prompt=None, tokens=None, sampling_params=None, num_samples=1) -> Future[SampleResult]— pass exactly one ofpromptortokens.save_weights_and_get_sampling_client(name, dir=None) -> Future[SaveResult](alias:save_weights)get_info() -> TrainingClientInfodelete()