nle-ttyplay crash

[StephenOman]

This is a migration copy of the original bug report here: facebookresearch#353

🐛 Bug

Crash when running nle-ttyplay (or nle-ttyplay2) on the ttyrec3 file generated by the test function in agent.

NetHack, Copyright 1985-2020
         By Stichting Mathematisch Centrum and M. Stephenson.
         Version 3.6.6-0 Unix post-release, built Mar 20 20:58:46 2023.
         See license for details.
Hello Agent, welcome to NetHack!  You are a neutral male human Monk.

                                                         -------------
                                                         ............|
                                                         |...........|
                                                         ......d.....|
                                                         |......Traceback (most recent call last):
  File "/opt/conda/bin/nle-ttyplay2", line 8, in <module>-----+----- -
    sys.exit(main())
  File "/opt/conda/lib/python3.8/site-packages/nle/scripts/ttyplay2.py", line 10, in main
    ttyplay.main()
  File "/opt/conda/lib/python3.8/site-packages/nle/scripts/ttyplay.py", line 212, in main
    process(f)
  File "/opt/conda/lib/python3.8/site-packages/nle/scripts/ttyplay.py", line 128, in process
    for timestamp, length, channel in read_header(
  File "/opt/conda/lib/python3.8/site-packages/nle/scripts/ttyplay.py", line 107, in read_header
    raise IOError("Illegal header %s" % ((sec, usec, length, channel),))
OSError: Illegal header (1679347987, 464255, 4, 2)

To Reproduce

Steps to reproduce the behavior:

1. Run the example agent training (python -m nle.agent.agent --num_actors 5 --batch_size 32 --unroll_length 80 --learning_rate 0.0001 --entropy_cost 0.0001 --use_lstm --total_steps 1000000000)
2. Run test mode
   python -m nle.agent.agent --mode test --use_lstm --save_ttyrec_every 1
3. Run ttyplay on generated ttyrec3 file, in my case this is something like
   nle-ttyplay nle_data/20230320-213238_4evrptzm/nle.256.1.ttyrec3.bz2

Expected behavior

Expected to see a ttyplay replay.

Environment

NLE version: 0.9.0+4f5b57e
PyTorch version: 2.0.0+cu117
Is debug build: No
CUDA used to build PyTorch: 11.7

OS: Ubuntu 18.04.6 LTS
GCC version: (Ubuntu 7.5.0-3ubuntu1~18.04) 7.5.0
CMake version: version 3.26.0

Python version: 3.8
Is CUDA available: Yes
CUDA runtime version: 10.2.89
GPU models and configuration: GPU 0: NVIDIA GeForce GTX 1070
Nvidia driver version: 520.56.06
cuDNN version: /usr/lib/x86_64-linux-gnu/libcudnn.so.7.6.5

Versions of relevant libraries:
[pip3] numpy==1.24.2
[pip3] torch==2.0.0
[conda] torch 2.0.0 pypi_0 pypi

Additional context

Running in a docker image built from the provided Dockerfile.bionic.