Skip to content
Open
Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
74 commits
Select commit Hold shift + click to select a range
bb00d91
add README
NicolasAG Jun 16, 2025
dc81770
increase env session inactivity timout
NicolasAG Jun 17, 2025
e60d4c1
update readme
NicolasAG Jun 17, 2025
f9e45c2
move miniwob to domains/
NicolasAG Jun 18, 2025
8cdbd06
fix
NicolasAG Jul 7, 2025
5510982
fix path
NicolasAG Jul 7, 2025
07e858c
return RuntimeError instead of HTTPException because not pickable
NicolasAG Jul 7, 2025
5e56896
add env_call_timeout
NicolasAG Jul 8, 2025
c06b768
update gpu fractions
NicolasAG Jul 8, 2025
b1ad285
set kl coef to 0
NicolasAG Jul 8, 2025
6bbe977
Merge remote-tracking branch 'origin/main' into debug_miniwob
NicolasAG Jul 8, 2025
c8ac64d
update max seq len
NicolasAG Jul 8, 2025
b87a6d1
revert to json instead of tool use agent
NicolasAG Jul 9, 2025
824d841
update README
NicolasAG Jul 9, 2025
8d170ec
debug overflow counter
NicolasAG Jul 10, 2025
21a1b2a
fix prompts
NicolasAG Jul 10, 2025
05b6794
update readme
NicolasAG Jul 11, 2025
ef6b2b0
flag tape as invalid instead of raising http errors
NicolasAG Jul 21, 2025
0abc2b0
use redis
NicolasAG Jul 21, 2025
d3f6889
track task names instead of data splits
NicolasAG Jul 21, 2025
9c319e3
fix
NicolasAG Jul 21, 2025
92c8a93
remove unused var in new tapeagent remote_env
NicolasAG Jul 22, 2025
edf4d00
use BaseMetrics
NicolasAG Jul 23, 2025
28749e0
fix
NicolasAG Jul 23, 2025
a4f9f79
keep track of time taken
NicolasAG Jul 23, 2025
8a6120f
send per step times to wandb
ollmer Jul 24, 2025
d1d1836
Merge remote-tracking branch 'origin/main' into debug_miniwob
NicolasAG Jul 25, 2025
5eb3a4e
use all miniwob tasks
NicolasAG Jul 25, 2025
75d3c9c
default save checkpoints
NicolasAG Jul 28, 2025
6b97c7b
update vllm max tokens
NicolasAG Jul 28, 2025
d3cf30b
assert group size is as expected
NicolasAG Jul 28, 2025
4c50f1f
assert finetuning length is as much as vllm max length
NicolasAG Jul 28, 2025
ff61d73
update finetuning & vllm max lengths
NicolasAG Jul 28, 2025
a00e6e6
debug agent
NicolasAG Jul 28, 2025
6f149c8
use ppo & upd config
NicolasAG Aug 8, 2025
2ae2dd8
update readme
NicolasAG Aug 8, 2025
913c8e2
stop training after 1k steps
NicolasAG Aug 11, 2025
402eeb2
scale up env servers by llm_servers
NicolasAG Aug 20, 2025
58f31cc
reweight actor/trainer
NicolasAG Aug 20, 2025
4101d77
add massimo miniwob split
NicolasAG Aug 20, 2025
b00e476
cleanup
NicolasAG Aug 20, 2025
0b56125
update agent reflection node
NicolasAG Aug 21, 2025
9b0a74c
towards massimo setup
NicolasAG Aug 22, 2025
e6e735d
Merge remote-tracking branch 'origin/main' into debug_miniwob
NicolasAG Aug 22, 2025
ef46f39
upd configs
NicolasAG Aug 28, 2025
1274748
upd
NicolasAG Aug 28, 2025
b16d45c
revert reward calculation
NicolasAG Aug 28, 2025
9e61c35
update massimo cfg to grpo
NicolasAG Aug 28, 2025
ef884f2
test with ppo
NicolasAG Aug 28, 2025
537ec7a
update configs
NicolasAG Sep 2, 2025
7a4e73f
add retry mechanism for agent loop
NicolasAG Sep 2, 2025
42e811e
add 30min timeout to rollout function
NicolasAG Sep 3, 2025
a4e8f5f
upd configs
NicolasAG Sep 5, 2025
95b735b
upd
NicolasAG Sep 5, 2025
8616303
upd configs
NicolasAG Sep 5, 2025
923cf6a
reduce n_env
NicolasAG Sep 6, 2025
44a033f
boost preprocess power
NicolasAG Sep 6, 2025
2918d1f
pop old data
NicolasAG Sep 6, 2025
dacaa1f
do not save playwright traces & screenshots
NicolasAG Sep 7, 2025
fcee5ee
return empty aggregate stats if empty stats
NicolasAG Sep 7, 2025
631389f
increase preprocessor power
NicolasAG Sep 7, 2025
f791211
better error handling
NicolasAG Sep 8, 2025
c54d900
fix
NicolasAG Sep 8, 2025
ea4918a
reduce timeouts
NicolasAG Sep 9, 2025
e5fca10
log number of groups done so far
NicolasAG Sep 12, 2025
df66a88
log everything if populate_rl_data fails
NicolasAG Sep 12, 2025
c8d0171
monitor env servers and reset if needed
NicolasAG Sep 12, 2025
981cd85
better health message
NicolasAG Sep 12, 2025
9c755ed
small fix
NicolasAG Sep 13, 2025
0b8a24d
better logs
NicolasAG Sep 26, 2025
cd27e30
always check the worker before launching the agent on it + more detai…
NicolasAG Sep 26, 2025
f9ce99e
log stack trace
NicolasAG Sep 29, 2025
60fb042
small cleanup
NicolasAG Sep 29, 2025
122db3c
add massimo heldout goals
NicolasAG Oct 14, 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
2 changes: 1 addition & 1 deletion pipelinerl/actor.py
Original file line number Diff line number Diff line change
Expand Up @@ -196,7 +196,7 @@ async def rollout_and_maybe_produce_result(
f"groups in progress: {len(group_rollouts)}, "
f"rollouts started so far: {started_rollouts}, "
f"rollouts finished so far: {finished_rollouts}, "
f"groups finished so far: {group_id}, "
f"groups started so far: {group_id}, "
f"max group size in bytes: {result_queue.max_actual_entry_size()}, "
)
last_logged = time.time()
Expand Down
5 changes: 4 additions & 1 deletion pipelinerl/domains/miniwob/rollouts.py
Original file line number Diff line number Diff line change
Expand Up @@ -73,7 +73,10 @@ async def check_env_server_health(env_job: Job, session: aiohttp.ClientSession)
error_text = await response.text()
return {"healthy": False, "error_status": f"HTTP {response.status}", "error_message": error_text}
except Exception as e:
return {"healthy": False, "error_status": "Unknown", "error_message": str(e)}
exception_type = type(e).__name__
exception_message = str(e) if str(e) else "No message available"
logger.exception(f"Error checking environment server health: {exception_type}: {exception_message}", stack_info=True)
return {"healthy": False, "error_status": f"Exception: {exception_type}", "error_message": exception_message}


async def reset_env_server(env_job: Job, session: aiohttp.ClientSession) -> bool:
Expand Down