Require setuptools>=70 and update deprecated api

Signed-off-by: Dong Hyuk Chang <donghyukc@nvidia.com>

examples/nlp/language_modeling/megatron_gpt_distillation.py

@@ -38,7 +38,7 @@
 from megatron.core.dist_checkpointing.mapping import ShardedStateDict
 from megatron.core.parallel_state import get_tensor_model_parallel_group
 from megatron.core.transformer import TransformerConfig
-from pkg_resources import packaging
+import packaging
 from torch import Tensor
 from torch.nn.modules.loss import _Loss
 

nemo/collections/multimodal/models/multimodal_llm/neva/neva_model.py

@@ -22,7 +22,7 @@
 import torch.nn.functional as F
 from einops import rearrange, reduce, repeat
 from omegaconf import DictConfig, ListConfig, OmegaConf
-from pkg_resources import packaging
+import packaging
 from pytorch_lightning.trainer.trainer import Trainer
 from transformers import CLIPVisionModel, SiglipVisionModel
 

nemo/collections/nlp/models/language_modeling/megatron/gpt_full_te_layer_autocast_spec.py

@@ -16,7 +16,7 @@
 from typing import Any, Callable, Optional
 
 import torch
-from pkg_resources import packaging
+import packaging
 
 from nemo.collections.nlp.modules.common.megatron.utils import ApexGuardDefaults
 from nemo.collections.nlp.parts import utils_funcs

nemo/collections/nlp/modules/common/megatron/attention.py

@@ -67,10 +67,9 @@
 
 try:
     # Flash Attention Triton
-    import pkg_resources
     from flash_attn.flash_attn_triton import flash_attn_func as flash_attn_func_triton
 
-except (ImportError, ModuleNotFoundError, pkg_resources.DistributionNotFound):
+except (ImportError, ModuleNotFoundError):
 
     flash_attn_func_triton = None
 

requirements/requirements.txt

@@ -5,7 +5,7 @@ onnx>=1.7.0
 python-dateutil
 ruamel.yaml
 scikit-learn
-setuptools>=65.5.1
+setuptools>=70.0.0
 tensorboard
 text-unidecode
 torch

tests/collections/nlp/test_flash_attention.py

@@ -39,7 +39,6 @@
     HAVE_FA = False
 
 try:
-    import pkg_resources
     import triton
 
     HAVE_TRITON = True