docs: fixed deps version and added server run cmds

Author: juelg

README.md

@@ -38,31 +38,40 @@ conda create -n octo python=3.10
 conda activate octo
 conda install nvidia/label/cuda-11.8.0::cuda --no-channel-priority
 conda install conda-forge::cudnn=8.9
+# octo dependencies
+pip install git+https://github.com/octo-models/octo.git@241fb3514b7c40957a86d869fecb7c7fc353f540
+pip install -r agents/utils/fixed_octo_requirements.txt
 # for gpu support:
 pip install --upgrade "jax[cuda11_pip]==0.4.20" -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html
 ```
 
 Verify that the jax installation was successful and that jax finds your gpu.
-Open a python terminal in the same conda env and type
+Open a python shell in the same conda env and type
 ```python
 from jax.lib import xla_bridge
-# this should output gpu if the installation was successful
+# this should output "gpu" if the gpu installation was successful
 print(xla_bridge.get_backend().platform)
 ```
 
-Install Octo and dependencies:
-```shell
-pip install -r agents/utils/fixed_octo_requirements.txt
-pip install git+https://github.com/octo-models/octo.git
-```
 
 Install the agents library on top:
 ```shell
-pip install git+https://github.com/octo-models/octo.git
+pip install git+https://github.com/juelg/agents.git
 ```
 
 For more details, see the [Octo github page](https://github.com/octo-models/octo).
 
+#### Troubleshooting
+If pip conplains about dependency issues than it might have happened that torch somehow slipped in.
+Check if you have any torch packages installed by
+```shell
+pip freeze | grep torch
+# if any, uninstall them e.g.
+pip uninstall arm_pytorch_utilities
+pip uninstall pytorch-seed
+pip uninstall pytorch_kinematics
+```
+
 ### OpenVLA
 To use OpenVLA, create a new conda environment:
 ```shell
@@ -81,15 +90,23 @@ pip install "flash-attn==2.5.5" --no-build-isolation
 
 Install OpenVLA
 ```shell
-pip install git+https://github.com/openvla/openvla.git
+pip install git+https://github.com/openvla/openvla.git@46b752f477cc5773cc1234b2e82c0e2130e4e890
+```
+
+Install the agents library on top:
+```shell
+pip install git+https://github.com/juelg/agents.git
 ```
 
 For more details, see the [OpenVLA github page](https://github.com/openvla/openvla).
 
 ## Usage
 To start an agents server use the `start-server` command where `kwargs` is a dictionary of the constructor arguments of the policy you want to start e.g.
 ```shell
-python -m agents start-server octo --host localhost --port 8080 --kwargs '{"checkpoint_path": "hf://Juelg/octo-base-1.5-finetuned-maniskill", "checkpoint_step": 60000, "horizon": 1, "unnorm_key": []}'
+# octo
+python -m agents start-server octo --host localhost --port 8080 --kwargs '{"checkpoint_path": "hf://Juelg/octo-base-1.5-finetuned-maniskill", "checkpoint_step": None, "horizon": 1, "unnorm_key": []}'
+# openvla
+python -m agents start-server openvla --host localhost --port 8080 --kwargs '{"checkpoint_path": "Juelg/openvla-7b-finetuned-maniskill", "device": "cuda:0", "attn_implementation": "flash_attention_2", "unnorm_key": "maniskill_human:7.0.0", "checkpoint_step": 40000}'
 ```
 
 There is also the `run-eval-during-training` command to evaluate a model during training, so a single checkpoint.