yasserfarouk / negmas Goto Github PK

Hello,

I'd like to extend the anac2019_std tournament deadline and for that purpose, set total_time_out to 72000, but it appears that it still considers the default parameter 7200 and the tournament finishes in two hours, I've found that anac_2019_config_generator in scml.utils always assigns total_time_out to 7200 which may be potentially the cause of the bug.

Describe the bug
Github workflow fails with

urllib.error.URLError: <urlopen error [Errno 110] Connection timed out>
Downloading: http://www.yasserm.com/scml/geniusbridge.jar

Rerunning it manually fixes it.
To Reproduce
Steps to reproduce the behavior:
Commit and Push any update

Expected behavior
Pass all the CI/CD pipelines

Screenshots
I am attaching the logs

logs.txt

call init function of monitor at the step 0, not step function
monitor.step(self) -> monitor.init(self)

Describe the bug

There are some bugs about logging.

• self.awi.logdebug_agent()` does not work
• World logs are stored in the log files of each agent.

To Reproduce

Run this script

from negmas import Outcome, MechanismState, ResponseType
from scml import SCML2023OneShotWorld, GreedySyncAgent
from scml.oneshot import OneShotAgent


class MyAgent(OneShotAgent):
    def init(self):
        super().init()
        self.awi.logdebug_agent("agent!")
        self.awi.logdebug("world!")

    def propose(self, negotiator_id: str, state: MechanismState) -> Outcome:
        return tuple([1, 1, 1])

    def respond(
        self, negotiator_id: str, state: MechanismState, offer: Outcome
    ) -> ResponseType:
        return ResponseType.ACCEPT_OFFER


world = SCML2023OneShotWorld(
    **SCML2023OneShotWorld.generate(
        agent_types=[MyAgent] * 2,
        agent_processes=[0, 1],
        n_steps=3,
        n_processes=2,
        name="test",
    )
)

world.run()

Expected behavior

• "agent!" is logged in ~/negmas/logs/test/_agent_logs/[email protected] and [email protected]
• world logs are logged in ~/negmas/logs/test/log.txt

actual behaviour

• "agent!" is not logged in any log files.
• [email protected] and [email protected] have logs as same as ~/negmas/logs/test/log.txt

([email protected])

2023-04-04 21:54:36,439 - DEBUG - 0/3 [0.00%]: world!
2023-04-04 21:54:36,439 - DEBUG - 0/3 [0.00%]: world!
2023-04-04 21:54:36,439 - INFO - 0/3 [0.00%]: 0 Negotiations/4 Agents
2023-04-04 21:54:36,440 - DEBUG - 0/3 [0.00%]: 01My@1 requested to buy 1 to 00My@0 q: (1, 10), u: (15, 16), t: (0, 0)
...

(log.txt)

2023-04-04 21:54:36,437 - INFO - 0/3 [0.00%]: test: World Created
2023-04-04 21:54:36,438 - INFO - 0/3 [0.00%]: 00My@0 joined
2023-04-04 21:54:36,438 - INFO - 0/3 [0.00%]: 01My@1 joined
2023-04-04 21:54:36,438 - INFO - 0/3 [0.00%]: SELLER joined
2023-04-04 21:54:36,438 - INFO - 0/3 [0.00%]: BUYER joined
2023-04-04 21:54:36,439 - DEBUG - 0/3 [0.00%]: world!
2023-04-04 21:54:36,439 - DEBUG - 0/3 [0.00%]: world!
2023-04-04 21:54:36,439 - INFO - 0/3 [0.00%]: 0 Negotiations/4 Agents
2023-04-04 21:54:36,440 - DEBUG - 0/3 [0.00%]: 01My@1 requested to buy 1 to 00My@0 q: (1, 10), u: (15, 16), t: (0, 0)

Desktop (please complete the following information):

• OS: macOS Ventura 13.2.1 (with M1)
• Python version: 3.10.10
• package versions:
  • negmas: 0.9.8
  • scml: 0.5.6
• Browser [e.g. chrome, safari]
• Version [e.g. 22]

Additional context

I think negmas.helpers.logging.create_loggers() (here) has some problems.

• Logs from each agent are sent to the root logger, which has the default logging level (=WARNING).
  • Debug level logs are ignored.
• World logs are sent to the "logging" logger, whose parent logger is the root logger.
  • World logs are propagated to the root logger. As the root logger has file log handlers associated with each agent, world logs are written to each agent's log file.

@yasserfarouk

Sir, Thanks for developing this excellent library with good documentation.

I would like to know
(a) what are some of the other business areas where this negotiation agents paradigm can be used (I mean other than the typical factory or supply chain process ). For instance, can this be used for financial portfolio optimization for example as a model, a seller offers to construct financial portfolio for the buyer with a certain monetary return value and risk where as the buyer expects the seller to construct a low risk portfolio with a high return over a certain tenure. ?

(b) Can this approach be used to solve operations research problems like route planning or optimization?

(c) How does Negmas' approach differ from typical constraint based optimization methods?

Would be great if you can share your thoughts and inputs. Thank you in advance.