Skip to content

full

Every family in one extra. Use it when your models span more than one family extra.

extra CPU tag GPU tag families models example
full :full :full-gpu all of them 117 chronos_2

chronos + kronos + granite + moirai + tirex + tirex2 + toto + mantis + timesfm3 + t0 + tafsut, which pulls in hub and base. It is also the largest install: for one family, the extra on that family's page pulls far less.

Start a server

docker run --rm -p 8000:8000 sktime/tserve:full \
  chronos_2 tirex kronos

uv pip install "tserve[server,full]"

uv run tserve chronos_2 tirex kronos

pip install "tserve[server,full]"

tserve chronos_2 tirex kronos

Each model costs a download at first start and stays in memory.

Check what loaded:

curl -s http://127.0.0.1:8000/models

Predict

One process answers for every loaded model; switch by changing model. Python needs the client extra on the caller.

curl -s http://127.0.0.1:8000/predict -H "Content-Type: application/json" -d '{
  "past": {
    "timestamp": ["2024-01-01", "2024-01-02", "2024-01-03", "2024-01-04", "2024-01-05"],
    "sales": [120, 135, 128, 142, 138]
  },
  "time": "timestamp",
  "target": ["sales"],
  "fh": 3,
  "model": "chronos_2"
}'

curl.exe -s http://127.0.0.1:8000/predict -H "Content-Type: application/json" -d '{"past":{"timestamp":["2024-01-01","2024-01-02","2024-01-03","2024-01-04","2024-01-05"],"sales":[120,135,128,142,138]},"time":"timestamp","target":["sales"],"fh":3,"model":"chronos_2"}'

from tserve.client import Client

past = {
    "timestamp": ["2024-01-01", "2024-01-02", "2024-01-03", "2024-01-04", "2024-01-05"],
    "sales": [120, 135, 128, 142, 138],
}

with Client("http://127.0.0.1:8000") as client:
    for model in ("chronos_2", "tirex", "kronos"):
        result = client.predict(
            past=past,
            time="timestamp",
            target=["sales"],
            fh=3,
            model=model,
        )
        print(model, result.predictions)

Models

All 117 models load here. Checkpoints for all of them: All models.

extra families models catalog
server Naive 1 base
hub Chronos Bolt, Chronos T5, TTM, TimesFM 2.x 81 hub
chronos Chronos-2 3 chronos
kronos Kronos, WindFM 5 kronos
granite FlowState 2 granite
moirai Moirai 2, Moirai 1.x, Lag-Llama 8 moirai
tirex TiRex 2 tirex
tirex2 TiRex-2 4 tirex2
toto Toto-2 5 toto
mantis Mantis 3 mantis
timesfm3 TimesFM 3 1 timesfm3
t0 T0 1 t0
tafsut Tafsut 1 tafsut

Also loadable here

Combinations no single family extra allows. kronos alone cannot load a TTM model, tirex alone cannot load a Kronos model; full serves all three at once:

tserve kronos ttm_r3 tirex

Next steps