Revert "fix: extend nccl timeout"

Signed-off-by: Charlie Doern <cdoern@redhat.com>

Author: cdoern

README.md

@@ -370,9 +370,3 @@ run_training(
     train_args=training_args,
 )
 ```
-
-## Environment variables
-
-Below is a list of custom environment variables users can set in the training library.
-
-1. `INSTRUCTLAB_NCCL_TIMEOUT_MS`, this environment variable controls the NCCL timeout in milliseconds. Consider increasing if seeing FSDP related NCCL errors.

src/instructlab/training/main_ds.py

@@ -4,7 +4,6 @@
 from copy import deepcopy
 from pathlib import Path
 import argparse
-import datetime
 import math
 import os
 import re
@@ -566,15 +565,10 @@ def main(args):
     model_conf = AutoConfig.from_pretrained(args.model_name_or_path)
     args.model_type = model_conf.model_type
 
-    # solution discovered from torchtune https://github.com/pytorch/torchtune/issues/2093
-    # gets converted to a timedelta of 1:40:00 if the default is kept
-    nccl_timeout = int(os.getenv("INSTRUCTLAB_NCCL_TIMEOUT_MS", "6000000"))
     #### distributed init #####
     torch.cuda.set_device(int(os.environ["LOCAL_RANK"]))
     args.local_rank = int(os.environ["LOCAL_RANK"])
-    torch.distributed.init_process_group(
-        "nccl", timeout=datetime.timedelta(milliseconds=nccl_timeout)
-    )
+    torch.distributed.init_process_group("nccl")
     args.global_rank = torch.distributed.get_rank()
     tensor = torch.ByteTensor([False]).cuda()
     torch.distributed.all_reduce(tensor)

tox.ini

@@ -18,7 +18,6 @@ basepython = python3.11
 description = run unit tests with pytest
 passenv =
 	HF_HOME
-        INSTRUCTLAB_NCCL_TIMEOUT_MS
 deps = 
     pytest
     -r requirements-dev.txt
@@ -32,7 +31,6 @@ commands = {envpython} -m pytest tests/unit {posargs}
 description = run accelerated smoke tests with pytest
 passenv =
 	HF_HOME
-        INSTRUCTLAB_NCCL_TIMEOUT_MS
 deps = 
     pytest
     -r requirements-dev.txt