Template updates

FLUX 3 Action SO-101

flux-3-action-so101 · policy template by Black Forest Labs · SO-101 · 6 joints · 30 Hz · A100-40GB. Each version below is fixed once published, newest first.

seq init flux-3-action-so101 gives the stable version, seq init flux-3-action-so101@beta the beta, and seq init flux-3-action-so101@v<N> any version below that is not withdrawn. A deployment made from this template names the version it runs in seq policy versions and on its page in the console; seq policy upgrade moves it to the newer stable one.

Performance

No runs yet.

Versions

  1. v9stablebeta

    FLUX 3 Action as v8; policy.py says, as the README does, how a worker's CPU and memory are billed.

    Needs sequence-ai 0.56.0 or later, and LeRobot 0.6.1 or later beside sequence-ai[observe].

    • v8's files; only policy.py's comment on CPU and memory changed: the machine its card sets (an A100-40GB's 12 cores and 85 GiB), or what it declares where that is more.
    Use v9
    flux-3-action-so101 v9shell
    # start a policy from this version, then deploy it
    # my_policy_user: the name you choose
    seq init flux-3-action-so101@v9 my_policy_user
    seq deploy my_policy_user
    
    # or move a deployment made from this template to it
    seq policy upgrade my_policy_user --to flux-3-action-so101@v9
    
    # on the robot's machine: this version's client.py and its README
    pip install -U sequence-ai
    seq init flux-3-action-so101@v9 my_policy_user

    In the console, Deploy deploys this version: the console deploys the stable one.

  2. v8

    FLUX 3 Action as v7; the README gives its card's price, and its whole machine's CPU and memory apart.

    Needs sequence-ai 0.56.0 or later, and LeRobot 0.6.1 or later beside sequence-ai[observe].

    • v7's files; only the README's price changed: $4.19 an hour for the A100-40GB, and its machine's 12 cores and 85 GiB at the serving rates for CPU and memory.
    Use v8
    flux-3-action-so101 v8shell
    # start a policy from this version, then deploy it
    # my_policy_user: the name you choose
    seq init flux-3-action-so101@v8 my_policy_user
    seq deploy my_policy_user
    
    # or move a deployment made from this template to it
    seq policy upgrade my_policy_user --to flux-3-action-so101@v8
    
    # on the robot's machine: this version's client.py and its README
    pip install -U sequence-ai
    seq init flux-3-action-so101@v8 my_policy_user
  3. v7

    FLUX 3 Action on flux-action's prepared path, and your limits may change a command without ending the episode.

    Needs sequence-ai 0.56.0 or later, and LeRobot 0.6.1 or later beside sequence-ai[observe].

    • Each act runs flux-action's prepared path instead of its reference model code, not compiled, so a worker's start compiles nothing, and its answers are not exactly those of v6.
    • preprocess(..., sent=...) sends the command your arm was given last tick, and the next answer carries on from it.
    • The loop in the README and client.py sends it, so it ends an episode only when you move the arm by hand.
    Use v7
    flux-3-action-so101 v7shell
    # start a policy from this version, then deploy it
    # my_policy_user: the name you choose
    seq init flux-3-action-so101@v7 my_policy_user
    seq deploy my_policy_user
    
    # or move a deployment made from this template to it
    seq policy upgrade my_policy_user --to flux-3-action-so101@v7
    
    # on the robot's machine: this version's client.py and its README
    pip install -U sequence-ai
    seq init flux-3-action-so101@v7 my_policy_user
  4. v6

    FLUX 3 Action as before; the README says when an episode must end: whenever your limits change a command.

    Needs sequence-ai 0.56.0 or later, and LeRobot 0.6.1 or later beside sequence-ai[observe].

    • Each answer continues from the last command, so call end_episode() before the next act() once your limits change one or you move the arm.
    • The loop in the README and client.py does this: robot_user.step() returns whether your limits changed the command.
    Use v6
    flux-3-action-so101 v6shell
    # start a policy from this version, then deploy it
    # my_policy_user: the name you choose
    seq init flux-3-action-so101@v6 my_policy_user
    seq deploy my_policy_user
    
    # or move a deployment made from this template to it
    seq policy upgrade my_policy_user --to flux-3-action-so101@v6
    
    # on the robot's machine: this version's client.py and its README
    pip install -U sequence-ai
    seq init flux-3-action-so101@v6 my_policy_user
  5. v5

    FLUX 3 Action as before; its README priced an A100-40GB with its CPU and memory included, as prices were then.

    Needs sequence-ai 0.56.0 or later, and LeRobot 0.6.1 or later beside sequence-ai[observe].

    • A worker was billed its card alone then, its price holding its CPU and memory; v8's README gives the prices that replaced it.
    Use v5
    flux-3-action-so101 v5shell
    # start a policy from this version, then deploy it
    # my_policy_user: the name you choose
    seq init flux-3-action-so101@v5 my_policy_user
    seq deploy my_policy_user
    
    # or move a deployment made from this template to it
    seq policy upgrade my_policy_user --to flux-3-action-so101@v5
    
    # on the robot's machine: this version's client.py and its README
    pip install -U sequence-ai
    seq init flux-3-action-so101@v5 my_policy_user
  6. v4

    FLUX 3 Action as before; the README says which LeRobot to run it beside.

    Needs sequence-ai 0.56.0 or later, and LeRobot 0.6.1 or later beside sequence-ai[observe].

    • LeRobot 0.5 and earlier require a version of av that cannot be installed beside sequence-ai[observe]; 0.6.1 and later install with it.
    • LeRobot 0.7.0 reads the arm as 0.5 does, so from_lerobot() and to_lerobot() are unchanged.
    Use v4
    flux-3-action-so101 v4shell
    # start a policy from this version, then deploy it
    # my_policy_user: the name you choose
    seq init flux-3-action-so101@v4 my_policy_user
    seq deploy my_policy_user
    
    # or move a deployment made from this template to it
    seq policy upgrade my_policy_user --to flux-3-action-so101@v4
    
    # on the robot's machine: this version's client.py and its README
    pip install -U sequence-ai
    seq init flux-3-action-so101@v4 my_policy_user
  7. v3

    FLUX 3 Action as before, with from_lerobot() and to_lerobot() in client.py, as molmoact2-so101 has them.

    Needs sequence-ai 0.56.0 or later.

    • Both pass LeRobot 0.5's numbers (use_degrees=True, its default) as they are, so one loop drives either SO-101 template.
    • The NOTICE says what v2 does: each deployment reads the weights itself, with no Hugging Face token of yours.
    Use v3
    flux-3-action-so101 v3shell
    # start a policy from this version, then deploy it
    # my_policy_user: the name you choose
    seq init flux-3-action-so101@v3 my_policy_user
    seq deploy my_policy_user
    
    # or move a deployment made from this template to it
    seq policy upgrade my_policy_user --to flux-3-action-so101@v3
    
    # on the robot's machine: this version's client.py and its README
    pip install -U sequence-ai
    seq init flux-3-action-so101@v3 my_policy_user
  8. v2

    FLUX 3 Action as before, with no Hugging Face token of yours: each deployment reads the weights itself.

    Needs sequence-ai 0.56.0 or later.

    Use v2
    flux-3-action-so101 v2shell
    # start a policy from this version, then deploy it
    # my_policy_user: the name you choose
    seq init flux-3-action-so101@v2 my_policy_user
    seq deploy my_policy_user
    
    # or move a deployment made from this template to it
    seq policy upgrade my_policy_user --to flux-3-action-so101@v2
    
    # on the robot's machine: this version's client.py and its README
    pip install -U sequence-ai
    seq init flux-3-action-so101@v2 my_policy_user
  9. v1

    Black Forest Labs' FLUX 3 Action for an SO-101 arm, on an A100, with client.py for the robot.

    Needs sequence-ai 0.56.0 or later, and a Hugging Face token of your own that has accepted the model's license: seq secret set hf --env HF_TOKEN.

    Use v1
    flux-3-action-so101 v1shell
    # start a policy from this version, then deploy it
    # my_policy_user: the name you choose
    seq init flux-3-action-so101@v1 my_policy_user
    seq deploy my_policy_user
    
    # or move a deployment made from this template to it
    seq policy upgrade my_policy_user --to flux-3-action-so101@v1
    
    # on the robot's machine: this version's client.py and its README
    pip install -U sequence-ai
    seq init flux-3-action-so101@v1 my_policy_user