search

opentensor / bittensor

Internet-scale Neural Networks
https://www.bittensor.com/
MIT License
929 stars 322 forks source link

btlogging weights #2441

Closed thewhaleking closed 2 days ago

thewhaleking commented 1 week ago

From Eclipse Vortex in Discord:

Hi everyone,

Has anyone encountered this error before using bittensor 8.3.1 with metagraph using torch (USE_TORCH == "1")?

TypeError: not all arguments converted during string formatting
Call stack:
  File "/root/SubVortex/neurons/validator.py", line 315, in run
    set_weights_for_validator(
  File "/root/SubVortex/subnet/validator/weights.py", line 117, in set_weights_for_validator
    ) = btuw.process_weights_for_netuid(
  File "/usr/local/lib/python3.10/dist-packages/bittensor/utils/weight_utils.py", line 259, in process_weights_for_netuid
    logging.debug("weights", *weights)
  File "/usr/local/lib/python3.10/dist-packages/bittensor/utils/btlogging/loggingmachine.py", line 438, in debug
    self._logger.debug(msg, *args, **kwargs)

I'm calling the method as follows: 

bt.process_weights_for_netuid(
    uids=metagraph.uids,
    weights=raw_weights,
    netuid=netuid,
    subtensor=subtensor,
    metagraph=metagraph,
)
Shr1ftyy commented 6 days ago

ditto. I too have this issue - I haven't tested with USE_TORCH=1 but it also happens when using numpy!

ibraheem-opentensor commented 5 days ago

This will be resolved in the upcoming release on Wednesday!