Skip to content

[Feature] Add training code for InternVLA-N1 #184

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
Tai-Wang merged 20 commits into from
Dec 15, 2025
Merged

[Feature] Add training code for InternVLA-N1 #184

Show file tree
Hide file tree
Changes from 13 commits
Commits
Show all changes
20 commits
Select commit Hold shift + click to select a range
2f3e1c0
[feat] Add training code for InternVLA-N1
kellyiss Dec 4, 2025
ee161af
merge dev and seperate utils
kew6688 Dec 5, 2025
152daf1
merge refactored evaluator with decoupled system2/dual_system eval
kew6688 Dec 5, 2025
ed992d9
update unused code; update setup.py and requirements for dtw
kew6688 Dec 5, 2025
2bc96cc
update requirements
kew6688 Dec 5, 2025
0f1d8a7
update diffuser version
kew6688 Dec 8, 2025
cf00d45
test r2r, fix import utils
kew6688 Dec 8, 2025
4dc059d
update ndtw json path
kew6688 Dec 8, 2025
4291aea
test rxr
kew6688 Dec 8, 2025
59a3b93
fix internvla_n1_policy with new model updates; remove use_async flag…
kew6688 Dec 8, 2025
f1d1d9f
Merge pull request #1 from kew6688/update_habitat_evaluator
kellyiss Dec 8, 2025
3b9f6a3
update README.md
kellyiss Dec 10, 2025
cc66ebe
Merge branch 'dev' into main
kew6688 Dec 10, 2025
d0563ee
fix some typos
kellyiss Dec 10, 2025
1f43712
Merge branch 'main' of github.com:kellyiss/InternNav into main
kellyiss Dec 10, 2025
4f6500e
Minor code cleanup
kellyiss Dec 12, 2025
4ef6856
remove unused code
kellyiss Dec 15, 2025
69413ff
rename dataset
kellyiss Dec 15, 2025
a9f0f48
reorganize trainer structure
kellyiss Dec 15, 2025
bdbe4cf
fix a bug in reorganization
kellyiss Dec 15, 2025
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
184 changes: 104 additions & 80 deletions README.md
View file
Open in desktop

Large diffs are not rendered by default.

13 changes: 5 additions & 8 deletions internnav/agent/internvla_n1_agent.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -207,7 +207,7 @@ def s2_thread_func():
self.s2_thread.daemon = True
self.s2_thread.start()

def should_infer_s2(self, mode="sync"):
def should_infer_s2(self, mode="partial_async"):
"""Function: Enables the sys2 inference thread depending on the mode.
mode: just support 2 modes: "sync" and "partial_async".
"sync": Synchronous mode (navdp_version >= 0.0), Sys1 and Sys2 execute in a sequential inference chain.
Expand Down Expand Up @@ -298,7 +298,7 @@ def step(self, obs):
if self.sys1_infer_times > 0:
self.dual_forward_step += 1

# print('Output action:', output, self.dual_forward_step)
print('Output action:', output, self.dual_forward_step)

else:
self.look_down = False
Expand Down Expand Up @@ -333,13 +333,9 @@ def step(self, obs):
.unsqueeze(-1)
.to(self.device)
) # [1, 2, 224, 224, 1]
self.s1_output = self.policy.s1_step_latent(
rgbs, depths, self.s2_output.output_latent, use_async=True
)
self.s1_output = self.policy.s1_step_latent(rgbs, depths, self.s2_output.output_latent)
else:
self.s1_output = self.policy.s1_step_latent(
rgb, depth * 10000.0, self.s2_output.output_latent, use_async=False
)
self.s1_output = self.policy.s1_step_latent(rgb, depth * 10000.0, self.s2_output.output_latent)

else:
assert False, f"S2 output should be either action or latent, but got neither! {self.s2_output}"
Expand Down Expand Up @@ -372,6 +368,7 @@ def step(self, obs):
if self.dual_forward_step > self.sys2_max_forward_step:
print("!!!!!!!!!!!!")
print("ERR: self.dual_forward_step ", self.dual_forward_step, " > ", self.sys2_max_forward_step)
print("Potential reason: sys1 infers empty trajectory list []")
print("!!!!!!!!!!!!")

print('Output discretized traj:', output['action'], self.dual_forward_step)
Expand Down
Loading