I have seen continiously failed tasks starting today. According to the stderr_txt file I reckon there might be at least two, possibly related, errors.

File "C:\ProgramData\BOINC\slots\5\python_dependencies\buffer.py", line 794, in insert_transition
state_embeds = [i["StateEmbeddings"] for i in sample[prl.INFO]]
File "C:\ProgramData\BOINC\slots\5\python_dependencies\buffer.py", line 794, in <listcomp>
state_embeds = [i["StateEmbeddings"] for i in sample[prl.INFO]]
KeyError: 'StateEmbeddings'
Traceback (most recent call last):
File "C:\ProgramData\BOINC\slots\5\lib\site-packages\pytorchrl\scheme\gradients\g_worker.py", line 196, in get_data
self.next_batch = self.batches.__next__()
AttributeError: 'GWorker' object has no attribute 'batches'

*KeyError: 'StateEmbeddings'
*AttributeError: 'GWorker' object has no attribute 'batches'

*KeyError: 'StateEmbeddings'
*AttributeError: 'GWorker' object has no attribute 'batches'

exactly same thing I notice on all my failed tasks.

Same here.

AttributeError: 'GWorker' object has no attribute 'batches'

my latest WU end with a computation error

---

Your first task link shows 4 attempts at retrieving the necessary python libraries and failing.

But instead of just stopping right there it looks like it tried to compute anyway with the missing 'batches' library and all the subsequent tasks failed also becauses of the missing batches element.

Seems that the error flow map is not branching out to a proper halt early enough in the task to stop the computation and waste anymore time.

Six tasks, all in a row. Errored out. Seven now and another in the works.

now the same problem on another host :-(
https://www.gpugrid.net/result.php?resultid=33101249

so, as seen by other members, too: all tasks which were downloaded within the past several hours seem to be faulty.

I joined yesterday and have 13 tasks failed in a row, all with the
AttributeError: 'GWorker' object has no attribute 'batches'.

Is this a failed installation? Should I try to reinstall this BOINC project from scratch?

Is this a failed installation? Should I try to reinstall this BOINC project from scratch?

in view of the above said, the current tasks are probably faulty.
No need to reinstall, I guess

Yes - just received and returned result 33101290, on a machine which regularly returns good results.

That was replication _6 of a WU which everyone else had failed - a sure sign that the problem was with the workunit, not the host processing it.

Forty-six failed WU"s? Please stop sending them until the problem is resolved.

Forty-six failed WU"s? Please stop sending them until the problem is resolved.

+ 1

Forty-six failed WU"s? Please stop sending them until the problem is resolved.

+ 1

Sorry. After writing the post I looked at the other computer and it had downloaded another. It lasted three minutes or so. It was still in the unzipping process. I cannot understand the txt files so can someone who can check the files to see what is going on?

+1

33 fails in a row. I'll set this project to inactive and wait for a solution.

I cannot understand the txt files so can someone who can check the files to see what is going on?

the task are wrongly configured. Don't download them for the time being.
I guess we will get some kind of "go ahead" here once the problem is solved on the project-side.

Hello, thanks you for reporting the job errors. Sorry to all, there was an error on my side setting up a batch of experiment agents. The errors is due to the specific python script of this batch, not related to the application itself. I have just fixed it, and the new jobs should be running correctly. Unfortunately, some already submitted jobs are bound to fail… I apologise for the inconvenience. They will fail briefly after starting as reported, so not a lot of compute will be wasted.

---

abouh,

could you also please make an adjustment (downwards) to the free disk space requirement of 33GB when downloading a Python task?

see my above Message 59449.

Many thanks :-)

Hello! I have checked and the disk space used by the jobs is set to 35e9 bytes.

<rsc_disk_bound>35e9</rsc_disk_bound>

I will change it first to 20e9, let me know if it helps. I can further decreased it in the future if necessary.

---

Hello! I have checked and the disk space used by the jobs is set to 35e9 bytes.

<rsc_disk_bound>35e9</rsc_disk_bound>

I will change it first to 20e9, let me know if it helps. I can further decreased it in the future if necessary.

Thanks, Abouh, for your quick reaction. The change will definitely help - at least in my case with limited disk space due to Ramdisk.

Hello, thanks you for reporting the job errors. Sorry to all, there was an error I have just fixed it, and the new jobs should be running correctly. Unfortunately, some already submitted jobs are bound to fail…

The problem is not fixed, I still get tasks that fail:
AttributeError: 'GWorker' object has no attribute 'batches'