[RecSys Implementation] TypeError: 'Box' object is not subscriptable

adhisetiawan · June 21, 2023, 6:07am

How severe does this issue affect your experience of using Ray?

High: It blocks me to complete my task.

Hello everyone,

I would like to seek assistance regarding an issue I encountered while running an example code from a GitHub repository. The code is related to a recommender system with RecSim, and the specific file I’m referring to is located here:

github.com

ray-project/ray/blob/master/rllib/examples/recommender_system_with_recsim_and_slateq.py

"""Using an RLlib-ready RecSim environment and the SlateQ algorithm
for solving recommendation system problems.

This example supports three different RecSim (RLlib-ready) environments,
configured via the --env option:
- "long-term-satisfaction"
- "interest-exploration"
- "interest-evolution"
"""

import argparse
import numpy as np
from scipy.stats import sem

import ray
from ray import air, tune
from ray.rllib.algorithms.algorithm_config import AlgorithmConfig
from ray.rllib.examples.env.recommender_system_envs_with_recsim import (
    InterestEvolutionRecSimEnv,
    InterestExplorationRecSimEnv,

This file has been truncated. show original

When I ran the code without making any modifications, I encountered the following error:

File "D:\Documents\Project\rllib\.env-rllib\lib\site-packages\ray\rllib\algorithms\slateq\slateq_torch_model.py", line 34, in __init__ self.embedding_size = self.orig_obs_space["doc"]["0"].shape[0] TypeError: 'Box' object is not subscriptable

I haven’t made any changes to the code and I tried running it with both PyTorch and TensorFlow with same command, but I still encountered the same error.
python .\examples\recommender_system_with_recsim_and_slateq.py --local-mode

I curious, is the error related to the SlateQ impelementation slateq_torch_model.py and slateq_tf_model.py file, because when I used the DQN model, it worked without any issues. However, I’m having trouble understanding the exact cause of the error.
I think this part cause the error.

github.com

ray-project/ray/blob/7f6fbf03436634657d4131a979a9fc118717725f/rllib/algorithms/slateq/slateq_torch_model.py#L33


      
          Each document candidate receives one full Q-value stack, defined by
          `fcnet_hiddens_per_candidate`. The input to each of these Q-value stacks
          is always {[user] concat [document[i]] for i in document_candidates}.
          
          
Extra model kwargs:
              fcnet_hiddens_per_candidate: List of layer-sizes for each(!) of the
                  candidate documents.
          """
          super().__init__()
          
          
self.orig_obs_space = obs_space
          self.embedding_size = self.orig_obs_space["doc"]["0"].shape[0]
          self.num_candidates = len(self.orig_obs_space["doc"])
          assert self.orig_obs_space["user"].shape[0] == self.embedding_size
          
          
self.q_nets = nn.ModuleList()
          for i in range(self.num_candidates):
              layers = nn.Sequential()
              ins = 2 * self.embedding_size
              for j, h in enumerate(fcnet_hiddens_per_candidate):
                  layers.add_module(

I would greatly appreciate any help or suggestions to resolve this issue.
Thank you in advance.

Topic		Replies	Views
How to use recsim env with SlateQ and Bandits? RLlib	0	258	August 12, 2022
[Rllib] File "/opt/conda/lib/python3.8/site-packages/ray/util/tracing/tracing_helper.py", line 461, in _resume_span TypeError: 'NoneType' object is not callable RLlib	5	663	July 29, 2022
Environment check throws Entire first structure: (., {}) Entire second structure: RLlib	4	312	May 24, 2023
TypeError: can't multiply sequence by non-int of type 'EnvContext' RLlib	2	246	June 21, 2023
Error: Custom observation Space not treated correctly RLlib	5	773	July 15, 2021

[RecSys Implementation] TypeError: 'Box' object is not subscriptable

Related Topics