Skip to content

[Tasks] Sample keyboard reset-buffer batches across all environments #8047

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
ooctipus merged 1 commit into from
Sep 27, 2026
Merged

[Tasks] Sample keyboard reset-buffer batches across all environments #8047

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 5 additions & 0 deletions source/isaaclab_tasks/changelog.d/keyboard-reset-buffer-coverage.rst
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,5 @@
Fixed
^^^^^

* Fixed keyboard reset-buffer partial batches to sample distinct environments across the full scene
instead of favoring its first clone variants. Buffer capacity was unchanged.
15 changes: 8 additions & 7 deletions source/isaaclab_tasks/isaaclab_tasks/contrib/keyboard/mdp/commands/typing_commands.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -531,13 +531,14 @@ def _build_buffer(self):
with tqdm(total=cap, desc="[typing] building reset-curriculum buffer (IK)", unit="snap") as pbar:
for start in range(0, cap, self.num_envs):
n = min(self.num_envs, cap - start)
ids = all_ids[:n]
# Partial batches must not favor the first clone variants when envs are grouped.
ids = all_ids if n == self.num_envs else torch.randperm(self.num_envs, device=self.device)[:n]
# Load this batch's cached command into the live state so the reset-IK aims at the right key.
self.target[:n] = tgt[start : start + n]
self.typed[:n] = typd[start : start + n]
self.target_len[:n] = tlen[start : start + n]
self.typed_len[:n] = typlen[start : start + n]
self.prefix_len[:n] = self._prefix_len()[:n]
self.target[ids] = tgt[start : start + n]
self.typed[ids] = typd[start : start + n]
self.target_len[ids] = tlen[start : start + n]
self.typed_len[ids] = typlen[start : start + n]
self.prefix_len[ids] = self._prefix_len()[ids]
self._solve_reset_pose(ids)
state = get_reset_state(self._env, ids, self._cur_reset_assets, is_relative=True)
if self._buf_state is None:
Expand All @@ -548,7 +549,7 @@ def _build_buffer(self):
ee_quat = self.robot.data.body_quat_w.torch[:, self._ik_body_idx]
tip = ee_pos + quat_apply(ee_quat, self._ik_offset)
reach = torch.linalg.norm(tip - (self.target_key_pos_w() + self._ik_hover), dim=-1)
self._buf_reach[start : start + n] = reach[:n]
self._buf_reach[start : start + n] = reach[ids]
pbar.update(n)
self._buffer_built = True
self._log_buffer_stats()
Expand Down
Loading