Search results

Inquiry about Cosmos Policy RoboCasa task-specific success rates

Status: Open.

#24 In NVlabs/cosmos-policy;

· LPY1219 opened on May 2, 2026

Validation takes so much time.

Status: Open.

#23 In NVlabs/cosmos-policy;

· NKUShaw opened on Apr 30, 2026

dgx spark inference performance

Status: Open.

#22 In NVlabs/cosmos-policy;

· jcl2023 opened on Apr 26, 2026

Request: env_args / full robomimic-style metadata for success_only HDF5 (or release path for geometry / replay tooling)

Status: Open.

#21 In NVlabs/cosmos-policy;

· shiona733 opened on Apr 18, 2026

Action predictions are very jittery at inference

Status: Open.

#20 In NVlabs/cosmos-policy;

· TTZYTttzyt opened on Apr 16, 2026

How to finetune on a custom dataset(Lerobot V2)

Status: Open.

#19 In NVlabs/cosmos-policy;

· SuperDagger opened on Apr 15, 2026

Performance without success rollout, failure rollout, value prediction

Status: Open.

#17 In NVlabs/cosmos-policy;

· lzl2040 opened on Apr 3, 2026

Checkpoint saved as .distcp instead of .pt when training with official Aloha dataset and pipeline

Status: Open.

#16 In NVlabs/cosmos-policy;

· zbzyjya opened on Apr 2, 2026

upgrade to Cosmos Predict 2.5 and Cosmos Reason 2.5

Status: Open.

#15 In NVlabs/cosmos-policy;

· skylerz-julyapril opened on Mar 31, 2026

about the default max_iter in cosmos_policy_experiment_configs

Status: Open.

#14 In NVlabs/cosmos-policy;

· stellasummer opened on Mar 26, 2026

Generating diverse candidates for best-of-N search

Status: Open.

#12 In NVlabs/cosmos-policy;

· 2016choang opened on Mar 17, 2026

Ablating proprio result

Status: Open.

#10 In NVlabs/cosmos-policy;

· Luvata opened on Mar 9, 2026