diff --git a/megatron/core/optimizer/emerging_optimizers.py b/megatron/core/optimizer/emerging_optimizers.py
index 3cf36670fd3..25294beabdf 100644
--- a/megatron/core/optimizer/emerging_optimizers.py
+++ b/megatron/core/optimizer/emerging_optimizers.py
@@ -8,6 +8,7 @@
   3. Add an ``EmergingOptimizerEntry`` to ``_EMERGING_OPTIMIZERS`` at the bottom.
 """
 
+import inspect
 import logging
 from dataclasses import dataclass, field
 from typing import Any, Callable, Dict, List, Literal, Optional
@@ -21,15 +22,20 @@
 from .optimizer_config import ParamKey, ParamPredicate
 
 try:
+    from emerging_optimizers import registry
     from emerging_optimizers.orthogonalized_optimizers import (
         OrthogonalizedOptimizer,
         get_muon_scale_factor,
     )
     from emerging_optimizers.orthogonalized_optimizers.muon_utils import newton_schulz_tp
 
+    # It is necessary to import SOAP for the registry to work.
+    from emerging_optimizers.soap import SOAP  # pylint: disable=unused-import
+
     HAVE_EMERGING_OPTIMIZERS = True
 except ImportError:
     HAVE_EMERGING_OPTIMIZERS = False
+    OrthogonalizedOptimizer = object
 
 
 logger = logging.getLogger(__name__)
@@ -54,14 +60,19 @@ class EmergingOptimizerEntry:
 
     optimizer_cls: type
     init_state_fn: Callable
-    config_to_kwargs: Callable
+    config_to_kwargs: Callable | None
     default_param_overrides: Dict[ParamKey, Dict[str, Any]] = field(default_factory=dict)
 
 
 def _create_emerging_optimizer(config, param_groups, eopt_name, model_chunks, pg_collection):
     """Instantiate an emerging optimizer and return it with its init_state_fn."""
     entry = _EMERGING_OPTIMIZERS[eopt_name]
-    eopt_kwargs = entry.config_to_kwargs(config, model_chunks, pg_collection)
+    if entry.config_to_kwargs is not None:
+        eopt_kwargs = entry.config_to_kwargs(config, model_chunks, pg_collection)
+    else:
+        eopt_kwargs = _default_adam_based_eopt_config_to_kwargs(
+            eopt_name, config, model_chunks, pg_collection
+        )
     optimizer = entry.optimizer_cls(param_groups, **eopt_kwargs)
     return optimizer, entry.init_state_fn
 
@@ -96,159 +107,180 @@ def _get_qkv_split_shapes(model_cfg) -> List[int]:
 # Muon
 # ===========================================================================
 
-if HAVE_EMERGING_OPTIMIZERS:
 
-    class TensorParallelMuon(OrthogonalizedOptimizer):
-        """Tensor Parallel Muon optimizer."""
-
-        def __init__(
-            self,
-            params: ParamsT,
-            lr: float = 3e-4,
-            momentum_beta: float = 0.95,
-            use_nesterov: bool = True,
-            weight_decay: float = 0.01,
-            use_decoupled_weight_decay: bool = True,
-            split_qkv: bool = False,
-            is_qkv_fn: Callable[[torch.Tensor], bool] | None = None,
-            qkv_split_shapes: tuple[int, int, int] | None = None,
-            fp32_matmul_prec: str = "medium",
-            coefficient_type: str = "quintic",
-            num_ns_steps: int = 5,
-            scale_mode: str = "spectral",
-            extra_scale_factor: float = 1.0,
-            pg_collection: Optional[ProcessGroupCollection] = None,
-            mode: Literal["blockwise", "duplicated", "distributed"] = "duplicated",
-        ) -> None:
-            if num_ns_steps < 1:
-                raise ValueError(f"num_ns_steps must be at least 1, got {num_ns_steps}")
-
-            def scaled_orthogonalize_fn(
-                grad: torch.Tensor,
-                tp_group: torch.distributed.ProcessGroup,
-                partition_dim: int | None = None,
-            ) -> torch.Tensor:
-                log_single_rank(
-                    logger,
-                    logging.DEBUG,
-                    f'Orthogonalizing grad with {num_ns_steps} steps, '
-                    f'{coefficient_type} coefficient, '
-                    f'{scale_mode} scale mode, extra_scale_factor={extra_scale_factor}',
-                )
-                size = [grad.size(-2), grad.size(-1)]
-                if partition_dim is not None:
-                    size[partition_dim] *= get_pg_size(tp_group)
-                orth_grad = newton_schulz_tp(
-                    grad,
-                    steps=num_ns_steps,
-                    coefficient_type=coefficient_type,
-                    tp_group=tp_group,
-                    partition_dim=partition_dim,
-                    mode="duplicated" if mode == "blockwise" else mode,
-                )
-                scale_factor = get_muon_scale_factor(size[0], size[1], mode=scale_mode)
-                return orth_grad * scale_factor * extra_scale_factor
-
-            self.pg_collection = pg_collection
-            self.mode = mode
-            self.split_qkv = split_qkv
-            self.is_qkv_fn = is_qkv_fn
-            self.qkv_split_shapes = qkv_split_shapes
-
-            weight_decay_method = "decoupled" if use_decoupled_weight_decay else "l2"
-            super().__init__(
-                params,
-                lr,
-                momentum_beta,
-                use_nesterov=use_nesterov,
-                weight_decay=weight_decay,
-                weight_decay_method=weight_decay_method,
-                fp32_matmul_prec=fp32_matmul_prec,
-                scaled_orthogonalize_fn=scaled_orthogonalize_fn,
+class TensorParallelMuon(OrthogonalizedOptimizer):
+    """Tensor Parallel Muon optimizer."""
+
+    def __init__(
+        self,
+        params: ParamsT,
+        lr: float = 3e-4,
+        momentum: float = 0.95,
+        nesterov: bool = True,
+        weight_decay: float = 0.01,
+        use_decoupled_weight_decay: bool = True,
+        split_qkv: bool = False,
+        is_qkv_fn: Callable[[torch.Tensor], bool] | None = None,
+        qkv_split_shapes: tuple[int, int, int] | None = None,
+        fp32_matmul_prec: str = "medium",
+        coefficient_type: str = "quintic",
+        num_ns_steps: int = 5,
+        scale_mode: str = "spectral",
+        extra_scale_factor: float = 1.0,
+        pg_collection: Optional[ProcessGroupCollection] = None,
+        tp_mode: Literal["blockwise", "duplicated", "distributed"] = "duplicated",
+    ) -> None:
+        if num_ns_steps < 1:
+            raise ValueError(f"num_ns_steps must be at least 1, got {num_ns_steps}")
+
+        def scaled_orthogonalize_fn(
+            grad: torch.Tensor,
+            tp_group: torch.distributed.ProcessGroup,
+            partition_dim: int | None = None,
+        ) -> torch.Tensor:
+            log_single_rank(
+                logger,
+                logging.DEBUG,
+                f'Orthogonalizing grad with {num_ns_steps} steps, '
+                f'{coefficient_type} coefficient, '
+                f'{scale_mode} scale mode, extra_scale_factor={extra_scale_factor}',
+            )
+            size = [grad.size(-2), grad.size(-1)]
+            if partition_dim is not None:
+                size[partition_dim] *= get_pg_size(tp_group)
+            orth_grad = newton_schulz_tp(
+                grad,
+                steps=num_ns_steps,
+                coefficient_type=coefficient_type,
+                tp_group=tp_group,
+                partition_dim=partition_dim,
+                tp_mode="duplicated" if tp_mode == "blockwise" else tp_mode,
             )
+            scale_factor = get_muon_scale_factor(size[0], size[1], mode=scale_mode)
+            return orth_grad * scale_factor * extra_scale_factor
+
+        self.pg_collection = pg_collection
+        self.tp_mode = tp_mode
+        self.split_qkv = split_qkv
+        self.is_qkv_fn = is_qkv_fn
+        self.qkv_split_shapes = qkv_split_shapes
+
+        weight_decay_method = "decoupled" if use_decoupled_weight_decay else "l2"
+        super().__init__(
+            params,
+            lr,
+            momentum,
+            nesterov=nesterov,
+            weight_decay=weight_decay,
+            weight_decay_method=weight_decay_method,
+            fp32_matmul_prec=fp32_matmul_prec,
+            scaled_orthogonalize_fn=scaled_orthogonalize_fn,
+        )
+
+    def orthogonalize(self, p: torch.Tensor, grad: torch.Tensor, **kwargs: Any) -> torch.Tensor:
+        """Orthogonalize the momentum.
+
+        Args:
+            p: The parameter tensor. i is necessary to pass param tensor in addition to
+                momentum because a lot of information is only available in the param tensor,
+                attributes for example.
+            grad: The momentum tensor.
+
+        Returns:
+            The orthogonalized gradient tensor.
+        """
+        # TODO(deyuf): switch to group
+        if self.pg_collection:
+            tp_group = (
+                self.pg_collection.expt_tp
+                if getattr(p, 'expert_tp', False)
+                else self.pg_collection.tp
+            )
+        else:
+            tp_group = None
+        partition_dim = None if self.tp_mode == "blockwise" else getattr(p, "partition_dim", None)
+        if partition_dim == -1:
+            partition_dim = None
+
+        if self.split_qkv and self.is_qkv_fn(p):  # type: ignore[misc]
+            grad_shape = grad.shape
+            log_single_rank(
+                logger,
+                logging.DEBUG,
+                f'qkv split grad shape {grad_shape}, ' f'split shapes {self.qkv_split_shapes}',
+            )
+            num_query_groups = grad_shape[0] // sum(self.qkv_split_shapes)
+            qkv_grads = torch.split(
+                grad.view(num_query_groups, sum(self.qkv_split_shapes), -1),
+                self.qkv_split_shapes,
+                dim=1,
+            )
+            qkv_grads = [g.reshape(-1, grad_shape[-1]) for g in qkv_grads]
 
-        def orthogonalize(self, p: torch.Tensor, grad: torch.Tensor, **kwargs: Any) -> torch.Tensor:
-            """Orthogonalize the momentum.
-
-            Args:
-                p: The parameter tensor. i is necessary to pass param tensor in addition to
-                    momentum because a lot of information is only available in the param tensor,
-                    attributes for example.
-                grad: The momentum tensor.
-
-            Returns:
-                The orthogonalized gradient tensor.
-            """
-            # TODO(deyuf): switch to group
-            if self.pg_collection:
-                tp_group = (
-                    self.pg_collection.expt_tp
-                    if getattr(p, 'expert_tp', False)
-                    else self.pg_collection.tp
-                )
-            else:
-                tp_group = None
-            partition_dim = None if self.mode == "blockwise" else getattr(p, "partition_dim", None)
-            if partition_dim == -1:
-                partition_dim = None
-
-            if self.split_qkv and self.is_qkv_fn(p):  # type: ignore[misc]
-                grad_shape = grad.shape
-                log_single_rank(
-                    logger,
-                    logging.DEBUG,
-                    f'qkv split grad shape {grad_shape}, ' f'split shapes {self.qkv_split_shapes}',
-                )
-                num_query_groups = grad_shape[0] // sum(self.qkv_split_shapes)
-                qkv_grads = torch.split(
-                    grad.view(num_query_groups, sum(self.qkv_split_shapes), -1),
-                    self.qkv_split_shapes,
-                    dim=1,
+            qkv_grads = [
+                self.scaled_orthogonalize_fn(g, tp_group, partition_dim).view(
+                    num_query_groups, -1, grad_shape[-1]
                 )
-                qkv_grads = [g.reshape(-1, grad_shape[-1]) for g in qkv_grads]
+                for g in qkv_grads
+            ]
+            grad = torch.cat(qkv_grads, dim=1).view(grad_shape)
+        else:
+            grad = self.scaled_orthogonalize_fn(grad, tp_group, partition_dim)
+        return grad
 
-                qkv_grads = [
-                    self.scaled_orthogonalize_fn(g, tp_group, partition_dim).view(
-                        num_query_groups, -1, grad_shape[-1]
-                    )
-                    for g in qkv_grads
-                ]
-                grad = torch.cat(qkv_grads, dim=1).view(grad_shape)
-            else:
-                grad = self.scaled_orthogonalize_fn(grad, tp_group, partition_dim)
-            return grad
-
-    def _muon_init_state_fn(opt, config=None):
-        """Initialize Muon optimizer state for torch_dist checkpoint format."""
-        for group in opt.param_groups:
-            for p in group['params']:
-                if len(opt.state[p]) == 0:
-                    opt.state[p]['momentum_buffer'] = torch.zeros_like(p.data)
-
-    def _muon_config_to_kwargs(config, model_chunks, pg_collection) -> Dict[str, Any]:
-        """Convert OptimizerConfig to TensorParallelMuon constructor kwargs."""
-        return {
-            "lr": config.lr,
-            "weight_decay": config.weight_decay,
-            "momentum_beta": config.muon_momentum,
-            "use_nesterov": config.muon_use_nesterov,
-            "fp32_matmul_prec": config.muon_fp32_matmul_prec,
-            "num_ns_steps": config.muon_num_ns_steps,
-            "scale_mode": config.muon_scale_mode,
-            "extra_scale_factor": config.muon_extra_scale_factor,
-            "mode": config.muon_tp_mode,
-            "split_qkv": config.muon_split_qkv,
-            "is_qkv_fn": lambda p: getattr(p, "is_qkv", False),
-            "qkv_split_shapes": _get_qkv_split_shapes(model_chunks[0].config),
-            "pg_collection": pg_collection,
-        }
-
-    # -----------------------------------------------------------------------
-    # Register Muon
-    # -----------------------------------------------------------------------
-    _EMERGING_OPTIMIZERS['muon'] = EmergingOptimizerEntry(
+
+def _eopt_init_state_fn(opt, config=None):
+    """Initialize emerging optimizer state for torch_dist checkpoint format."""
+    for group in opt.param_groups:
+        opt._init_group(group)
+
+
+def _kwargs_from_config(optimizer_cls: type, prefix: str, config) -> Dict[str, Any]:
+    """Match ``optimizer_cls.__init__`` parameters to config attributes.
+
+    For each init parameter, looks for ``{prefix}_{name}`` on *config* first,
+    then falls back to ``{name}`` (unprefixed).  ``self`` and ``params`` are
+    always skipped.
+    """
+    skip_params = {"self", "params"}
+    sig = inspect.signature(optimizer_cls.__init__)
+    kwargs: Dict[str, Any] = {}
+    for name in sig.parameters:
+        if name in skip_params:
+            continue
+        prefixed = f"{prefix}_{name}"
+        if hasattr(config, prefixed):
+            kwargs[name] = getattr(config, prefixed)
+        elif hasattr(config, name):
+            kwargs[name] = getattr(config, name)
+    return kwargs
+
+
+def _muon_config_to_kwargs(config, model_chunks, pg_collection) -> Dict[str, Any]:
+    """Convert OptimizerConfig to TensorParallelMuon constructor kwargs."""
+    kwargs = _kwargs_from_config(TensorParallelMuon, "muon", config)
+    kwargs["is_qkv_fn"] = lambda p: getattr(p, "is_qkv", False)
+    kwargs["qkv_split_shapes"] = _get_qkv_split_shapes(model_chunks[0].config)
+    kwargs["pg_collection"] = pg_collection
+    return kwargs
+
+
+def _default_adam_based_eopt_config_to_kwargs(
+    eopt_name, config, model_chunks, pg_collection
+) -> Dict[str, Any]:
+    """Convert OptimizerConfig to default emerging optimizer constructor kwargs."""
+    kwargs = _kwargs_from_config(registry.get_optimizer_cls(eopt_name), eopt_name, config)
+    kwargs["betas"] = (config.adam_beta1, config.adam_beta2)
+    return kwargs
+
+
+# -----------------------------------------------------------------------
+# Register emerging optimizers
+# -----------------------------------------------------------------------
+_EMERGING_OPTIMIZERS = {
+    'muon': EmergingOptimizerEntry(
         optimizer_cls=TensorParallelMuon,
-        init_state_fn=_muon_init_state_fn,
+        init_state_fn=_eopt_init_state_fn,
         config_to_kwargs=_muon_config_to_kwargs,
         default_param_overrides={
             ParamKey(
@@ -258,3 +290,23 @@ def _muon_config_to_kwargs(config, model_chunks, pg_collection) -> Dict[str, Any
             ): {'optimizer': 'adam'}
         },
     )
+}
+
+# Register soap with default config
+# TODO(skyw): register all emerging optimizers.
+if HAVE_EMERGING_OPTIMIZERS:
+    for eopt_name in ["soap"]:
+        if eopt_name in _EMERGING_OPTIMIZERS:
+            continue
+        _EMERGING_OPTIMIZERS[eopt_name] = EmergingOptimizerEntry(
+            optimizer_cls=registry.get_optimizer_cls(eopt_name),
+            init_state_fn=_eopt_init_state_fn,
+            config_to_kwargs=None,
+            default_param_overrides={
+                ParamKey(
+                    predicate=ParamPredicate(
+                        name="nonlinear_or_embedding", fn=_is_nonlinear_or_embedding
+                    )
+                ): {'optimizer': 'adam'}
+            },
+        )
diff --git a/megatron/core/optimizer/optimizer_config.py b/megatron/core/optimizer/optimizer_config.py
index 4b43e7b5c08..e10fd7852c7 100644
--- a/megatron/core/optimizer/optimizer_config.py
+++ b/megatron/core/optimizer/optimizer_config.py
@@ -255,14 +255,14 @@ class OptimizerConfig:
     sgd_momentum: float = 0.9
     """Momentum factor for SGD optimizer."""
 
-    # Muon / emerging optimizers.
+    # emerging optimizers.
     muon_momentum: float = 0.95
     """The momentum used by the internal SGD in Muon optimizer."""
 
     muon_split_qkv: bool = True
     """Whether to split QKV parameters for Muon optimizer."""
 
-    muon_use_nesterov: bool = False
+    muon_nesterov: bool = False
     """Whether to use Nesterov-style momentum in the internal SGD."""
 
     muon_scale_mode: str = "spectral"
@@ -280,6 +280,15 @@ class OptimizerConfig:
     muon_extra_scale_factor: float = 1.0
     """Additional scale factor for the muon update."""
 
+    soap_shampoo_beta: float = 0.95
+    """The beta parameter for the Shampoo preconditioner."""
+
+    soap_precondition_frequency: int = 1
+    """The frequency of the Shampoo preconditioner."""
+
+    soap_use_kl_shampoo: bool = True
+    """Whether to use the KL-Shampoo preconditioner."""
+
     #######################
     # Distributed optimizer
     #######################
diff --git a/megatron/training/arguments.py b/megatron/training/arguments.py
index b4691091be9..eb91fa11cc0 100644
--- a/megatron/training/arguments.py
+++ b/megatron/training/arguments.py
@@ -1323,8 +1323,8 @@ def validate_args(args, defaults={}):
         args.no_load_optim = True
         warn_rank_0('enabling --no-load-optim when skipping training.')
 
-    # Muon / emerging optimizer check
-    if args.optimizer in ('muon', 'dist_muon'):
+    # emerging optimizer check
+    if args.optimizer not in ('sgd', 'adam'):
         if args.optimizer == 'dist_muon':
             warn_rank_0(
                 "optimizer='dist_muon' is deprecated. "
@@ -2047,7 +2047,7 @@ def _add_regularization_args(parser):
     group.add_argument('--muon-no-split-qkv', action='store_false', default=True,
                        dest='muon_split_qkv',
                        help='Whether to split QKV parameters for Muon optimizer')
-    group.add_argument('--muon-use-nesterov', action='store_true',
+    group.add_argument('--muon-nesterov', action='store_true',
                        help='Whether to use Nesterov-style momentum in the internal SGD')
     group.add_argument('--muon-scale-mode', type=str, default='spectral',
                        choices=['spectral', 'unit_rms_norm', 'shape_scaling'],
@@ -2256,7 +2256,7 @@ def _add_training_args(parser):
                        help='use FlashAttention implementation of attention. '
                        'https://arxiv.org/abs/2205.14135')
     group.add_argument('--optimizer', type=str, default='adam',
-                       choices=['adam', 'sgd', 'muon', 'dist_muon'],
+                       choices=['adam', 'sgd', 'muon', 'dist_muon', 'soap'],
                        help='Optimizer function. '
                             'Note: dist_muon is deprecated; use --optimizer muon '
                             'with --use-distributed-optimizer instead.')
diff --git a/pyproject.toml b/pyproject.toml
index d39c9a011fc..52a168aaa3a 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -93,7 +93,7 @@ dev = [
     "onnxscript",
     "fastapi~=0.50",                                          # Forcing a little bit more recent version of fastapi to be compatible with pydantic 2.0
     "datasets",
-    "emerging_optimizers",
+    "emerging_optimizers; python_version >= '3.12'",
     "flask[async]",
     "hypercorn",
     "openai",
@@ -116,7 +116,7 @@ lts = [
     "onnxscript",
     "fastapi~=0.50",                      # Forcing a little bit more recent version of fastapi to be compatible with pydantic 2.0
     "datasets",
-    "emerging_optimizers",
+    "emerging_optimizers; python_version >= '3.12'",
 ]
 
 [dependency-groups]
@@ -160,7 +160,7 @@ linting = [
     "pylint==3.2.6",
 ]
 ci = ["python-gitlab", "slack-sdk", "pandas"]
-no_pypi_wheels = ["emerging_optimizers", "fast-hadamard-transform"]
+no_pypi_wheels = ["emerging_optimizers; python_version >= '3.12'", "fast-hadamard-transform"]
 
 [tool.uv]
 default-groups = ["linting", "build", "test"]
@@ -190,7 +190,7 @@ flash_mla = [
 ]
 transformer-engine = { git = "https://github.com/NVIDIA/TransformerEngine.git", rev = "5671fd3675906cda1ade26c24a65d3dedd88eb89" }
 nemo-run = { git = "https://github.com/NVIDIA-NeMo/Run.git", rev = "01a9a8ba360f7b2908728ad0516e0ad9d936966d" }
-emerging_optimizers = { git = "https://github.com/NVIDIA-NeMo/Emerging-Optimizers.git", rev = "v0.1.0" }
+emerging_optimizers = { git = "https://github.com/NVIDIA-NeMo/Emerging-Optimizers.git", rev = "a8faf641d5fca6a0515cfc010b6cedbf488cc33a" }
 fast-hadamard-transform = { git = "https://github.com/Dao-AILab/fast-hadamard-transform.git", rev = "f134af63deb2df17e1171a9ec1ea4a7d8604d5ca" }
 
 [tool.isort]
diff --git a/tests/unit_tests/dist_checkpointing/utils.py b/tests/unit_tests/dist_checkpointing/utils.py
index cf6662c72bf..5d7d42d9152 100644
--- a/tests/unit_tests/dist_checkpointing/utils.py
+++ b/tests/unit_tests/dist_checkpointing/utils.py
@@ -212,7 +212,10 @@ def setup_model_and_optimizer(
 
     if isinstance(optimizer, ChainedOptimizer):
         for opt in optimizer.chained_optimizers:
-            opt.init_state_fn(opt)
+            if not hasattr(opt, 'optimizer'):
+                opt.init_state_fn(opt)
+            else:
+                opt.init_state_fn(opt.optimizer)
     else:
         for group in optimizer.optimizer.param_groups:
             for p in group['params']:
@@ -308,7 +311,10 @@ def setup_moe_model_and_optimizer(
 
     if optimizer_type in ('muon', 'dist_muon'):
         for opt in optimizer.chained_optimizers:
-            opt.init_state_fn(opt)
+            if not hasattr(opt, 'optimizer'):
+                opt.init_state_fn(opt)
+            else:
+                opt.init_state_fn(opt.optimizer)
     else:
         for opt in optimizer.chained_optimizers:
             for group in opt.param_groups:
diff --git a/tests/unit_tests/test_muon_optimizer.py b/tests/unit_tests/test_emerging_optimizers.py
similarity index 66%
rename from tests/unit_tests/test_muon_optimizer.py
rename to tests/unit_tests/test_emerging_optimizers.py
index 86d75ee7a49..8fbac85c99f 100644
--- a/tests/unit_tests/test_muon_optimizer.py
+++ b/tests/unit_tests/test_emerging_optimizers.py
@@ -11,15 +11,20 @@
 from megatron.core import parallel_state
 from megatron.core.distributed import DistributedDataParallel, DistributedDataParallelConfig
 from megatron.core.optimizer import OptimizerConfig, get_megatron_optimizer
-from megatron.core.optimizer.emerging_optimizers import TensorParallelMuon
+from megatron.core.optimizer.emerging_optimizers import HAVE_EMERGING_OPTIMIZERS, TensorParallelMuon
 from megatron.core.process_groups_config import ProcessGroupCollection
 from megatron.core.transformer import TransformerConfig
 from tests.unit_tests.test_utilities import Utils
 
+if HAVE_EMERGING_OPTIMIZERS:
+    from emerging_optimizers.soap import SOAP
+else:
+    SOAP = None
+
 # Skip all tests in this file for LTS versions
 pytestmark = pytest.mark.skipif(
     Version(os.getenv('NVIDIA_PYTORCH_VERSION', "24.01")) <= Version("25.05"),
-    reason="Skip muon optimizer for LTS test",
+    reason="Skip emerging optimizer tests for LTS test",
 )
 
 
@@ -41,6 +46,11 @@ def forward(self, x):
         return x
 
 
+# ===========================================================================
+# Muon optimizer tests
+# ===========================================================================
+
+
 def test_muon_optimizer_smoke():
     """Smoke test for TensorParallelMuon optimizer."""
     # Create a simple linear model for testing
@@ -52,8 +62,8 @@ def test_muon_optimizer_smoke():
     optimizer = TensorParallelMuon(
         params=[model.weight],
         lr=0.01,
-        momentum_beta=0.95,
-        use_nesterov=True,
+        momentum=0.95,
+        nesterov=True,
         weight_decay=0.01,
         use_decoupled_weight_decay=True,
         split_qkv=False,
@@ -62,7 +72,7 @@ def test_muon_optimizer_smoke():
         scale_mode="spectral",
         extra_scale_factor=1.0,
         pg_collection=None,
-        mode="duplicated",
+        tp_mode="duplicated",
     )
 
     # Test basic properties
@@ -147,7 +157,7 @@ def test_get_megatron_optimizer_smoke(self):
             bf16=True,
             use_distributed_optimizer=False,  # Muon doesn't support distributed optimizer
             muon_momentum=0.95,
-            muon_use_nesterov=True,
+            muon_nesterov=True,
             muon_fp32_matmul_prec="medium",
             muon_num_ns_steps=5,
             muon_scale_mode="spectral",
@@ -243,7 +253,7 @@ def test_get_megatron_optimizer_layer_wise(self):
             bf16=True,
             use_layer_wise_distributed_optimizer=True,
             muon_momentum=0.95,
-            muon_use_nesterov=True,
+            muon_nesterov=True,
             muon_fp32_matmul_prec="medium",
             muon_num_ns_steps=5,
             muon_scale_mode="spectral",
@@ -292,11 +302,11 @@ def test_muon_optimizer_different_modes_single_rank(mode):
     optimizer = TensorParallelMuon(
         params=[model.weight],
         lr=0.01,
-        momentum_beta=0.95,
+        momentum=0.95,
         weight_decay=0.0,  # Disable weight decay for deterministic comparison
         num_ns_steps=5,
         pg_collection=None,
-        mode=mode,
+        tp_mode=mode,
     )
 
     # Use fixed input for deterministic results
@@ -352,11 +362,11 @@ def create_tp_model_and_optimizer(self, mode):
         optimizer = TensorParallelMuon(
             params=[model.weight],
             lr=0.01,
-            momentum_beta=0.95,
+            momentum=0.95,
             weight_decay=0.0,
             num_ns_steps=5,
             pg_collection=pg_collection,
-            mode=mode,
+            tp_mode=mode,
         )
 
         return model, optimizer
@@ -418,7 +428,7 @@ def test_muon_optimizer_coefficient_types(coefficient_type_and_steps):
         coefficient_type=coefficient_type_and_steps[0],
         num_ns_steps=coefficient_type_and_steps[1],
         pg_collection=None,
-        mode="duplicated",
+        tp_mode="duplicated",
     )
 
     input_tensor = torch.randn(16, 80, dtype=torch.float32, device='cuda')
@@ -447,7 +457,7 @@ def test_muon_optimizer_scale_modes(scale_mode):
         scale_mode=scale_mode,
         num_ns_steps=5,
         pg_collection=None,
-        mode="duplicated",
+        tp_mode="duplicated",
     )
 
     input_tensor = torch.randn(16, 60, dtype=torch.float32, device='cuda')
@@ -463,8 +473,8 @@ def test_muon_optimizer_scale_modes(scale_mode):
     ), f"Weight should be updated with scale_mode={scale_mode}"
 
 
-@pytest.mark.parametrize("use_nesterov", [True, False])
-def test_muon_optimizer_nesterov(use_nesterov):
+@pytest.mark.parametrize("nesterov", [True, False])
+def test_muon_optimizer_nesterov(nesterov):
     """Test TensorParallelMuon optimizer with and without Nesterov momentum."""
     model = torch.nn.Linear(50, 25, bias=False, dtype=torch.float32, device='cuda')
     model.requires_grad_(True)
@@ -473,11 +483,11 @@ def test_muon_optimizer_nesterov(use_nesterov):
     optimizer = TensorParallelMuon(
         params=[model.weight],
         lr=0.01,
-        momentum_beta=0.9,
-        use_nesterov=use_nesterov,
+        momentum=0.9,
+        nesterov=nesterov,
         num_ns_steps=5,
         pg_collection=None,
-        mode="duplicated",
+        tp_mode="duplicated",
     )
 
     input_tensor = torch.randn(16, 50, dtype=torch.float32, device='cuda')
@@ -490,7 +500,7 @@ def test_muon_optimizer_nesterov(use_nesterov):
 
     assert not torch.equal(
         model.weight.data, original_weight
-    ), f"Weight should be updated with use_nesterov={use_nesterov}"
+    ), f"Weight should be updated with nesterov={nesterov}"
 
 
 def test_muon_optimizer_multiple_steps():
@@ -502,11 +512,11 @@ def test_muon_optimizer_multiple_steps():
     optimizer = TensorParallelMuon(
         params=[model.weight],
         lr=0.01,
-        momentum_beta=0.95,
+        momentum=0.95,
         weight_decay=0.01,
         num_ns_steps=5,
         pg_collection=None,
-        mode="duplicated",
+        tp_mode="duplicated",
     )
 
     weights_history = [model.weight.data.clone()]
@@ -552,7 +562,7 @@ def test_muon_optimizer_qkv_split():
         qkv_split_shapes=qkv_split_shapes,
         num_ns_steps=5,
         pg_collection=None,
-        mode="duplicated",
+        tp_mode="duplicated",
     )
 
     input_tensor = torch.randn(16, hidden_size, dtype=torch.float32, device='cuda')
@@ -576,7 +586,7 @@ def test_muon_optimizer_qkv_split():
         split_qkv=False,
         num_ns_steps=5,
         pg_collection=None,
-        mode="duplicated",
+        tp_mode="duplicated",
     )
 
     output = model(input_tensor)
@@ -608,7 +618,7 @@ def test_muon_optimizer_extra_scale_factor():
         extra_scale_factor=2.0,
         num_ns_steps=5,
         pg_collection=None,
-        mode="duplicated",
+        tp_mode="duplicated",
     )
 
     input_tensor = torch.randn(16, 80, dtype=torch.float32, device='cuda')
@@ -637,7 +647,7 @@ def test_muon_optimizer_num_ns_steps(num_ns_steps):
         coefficient_type="quintic",
         num_ns_steps=num_ns_steps,
         pg_collection=None,
-        mode="duplicated",
+        tp_mode="duplicated",
     )
 
     input_tensor = torch.randn(16, 60, dtype=torch.float32, device='cuda')
@@ -651,3 +661,290 @@ def test_muon_optimizer_num_ns_steps(num_ns_steps):
     assert not torch.equal(
         model.weight.data, original_weight
     ), f"Weight should be updated with num_ns_steps={num_ns_steps}"
+
+
+# ===========================================================================
+# SOAP optimizer tests
+# ===========================================================================
+
+skip_no_soap = pytest.mark.skipif(
+    not HAVE_EMERGING_OPTIMIZERS, reason="emerging_optimizers package not installed"
+)
+
+
+@skip_no_soap
+def test_soap_optimizer_smoke():
+    """Smoke test for SOAP optimizer."""
+
+    model = torch.nn.Linear(100, 50, bias=False, dtype=torch.float32, device='cuda')
+    model.requires_grad_(True)
+    model.weight.data.fill_(1.0)
+
+    optimizer = SOAP(
+        params=[model.weight],
+        lr=0.01,
+        betas=(0.9, 0.999),
+        shampoo_beta=0.95,
+        weight_decay=0.01,
+        precondition_frequency=1,
+    )
+
+    # Test basic properties
+    assert optimizer is not None, "Optimizer should not be None"
+    assert hasattr(optimizer, 'param_groups'), "Optimizer should have param_groups"
+    assert len(optimizer.param_groups) > 0, "Optimizer should have at least one parameter group"
+
+    # Test forward and backward pass
+    input_tensor = torch.randn(32, 100, dtype=torch.float32, device='cuda')
+    output = model(input_tensor)
+    loss = output.sum()
+    loss.backward()
+
+    # Store original weight
+    original_weight = model.weight.data.clone()
+
+    # Test optimizer step
+    optimizer.step()
+
+    # Verify weight was updated
+    assert not torch.equal(
+        model.weight.data, original_weight
+    ), "Weight should be updated after optimizer step"
+
+    # Test zero_grad
+    optimizer.zero_grad()
+    assert model.weight.grad is None or torch.all(
+        model.weight.grad == 0
+    ), "Gradients should be zeroed"
+
+    # Test state_dict and load_state_dict
+    state_dict = optimizer.state_dict()
+    assert 'state' in state_dict, "State dict should contain state"
+    assert 'param_groups' in state_dict, "State dict should contain param_groups"
+
+    # Load state dict should not raise error
+    optimizer.load_state_dict(state_dict)
+
+
+@skip_no_soap
+def test_soap_optimizer_multiple_steps():
+    """Test SOAP optimizer across multiple optimization steps."""
+    model = torch.nn.Linear(100, 50, bias=False, dtype=torch.float32, device='cuda')
+    model.requires_grad_(True)
+    model.weight.data.fill_(1.0)
+
+    optimizer = SOAP(
+        params=[model.weight],
+        lr=0.01,
+        betas=(0.9, 0.999),
+        shampoo_beta=0.95,
+        weight_decay=0.01,
+        precondition_frequency=1,
+    )
+
+    weights_history = [model.weight.data.clone()]
+
+    for i in range(3):
+        input_tensor = torch.randn(32, 100, dtype=torch.float32, device='cuda')
+        output = model(input_tensor)
+        loss = output.sum()
+        loss.backward()
+
+        optimizer.step()
+        optimizer.zero_grad()
+        weights_history.append(model.weight.data.clone())
+
+    # Verify weights changed at each step
+    for i in range(len(weights_history) - 1):
+        assert not torch.equal(
+            weights_history[i], weights_history[i + 1]
+        ), f"Weight should change at step {i}"
+
+
+@skip_no_soap
+@pytest.mark.parametrize("precondition_frequency", [1, 5, 10])
+def test_soap_optimizer_precondition_frequency(precondition_frequency):
+    """Test SOAP optimizer with different precondition frequencies."""
+
+    model = torch.nn.Linear(60, 30, bias=False, dtype=torch.float32, device='cuda')
+    model.requires_grad_(True)
+    model.weight.data.fill_(1.0)
+
+    optimizer = SOAP(
+        params=[model.weight],
+        lr=0.01,
+        betas=(0.9, 0.999),
+        shampoo_beta=0.95,
+        precondition_frequency=precondition_frequency,
+    )
+
+    input_tensor = torch.randn(16, 60, dtype=torch.float32, device='cuda')
+    output = model(input_tensor)
+    loss = output.sum()
+    loss.backward()
+
+    original_weight = model.weight.data.clone()
+    optimizer.step()
+
+    assert not torch.equal(
+        model.weight.data, original_weight
+    ), f"Weight should be updated with precondition_frequency={precondition_frequency}"
+
+
+@skip_no_soap
+@pytest.mark.parametrize("use_kl_shampoo", [True, False])
+def test_soap_optimizer_kl_shampoo(use_kl_shampoo):
+    """Test SOAP optimizer with and without KL-Shampoo preconditioner."""
+
+    model = torch.nn.Linear(60, 30, bias=False, dtype=torch.float32, device='cuda')
+    model.requires_grad_(True)
+    model.weight.data.fill_(1.0)
+
+    optimizer = SOAP(
+        params=[model.weight],
+        lr=0.01,
+        betas=(0.9, 0.999),
+        shampoo_beta=0.95,
+        use_kl_shampoo=use_kl_shampoo,
+        precondition_frequency=1,
+    )
+
+    input_tensor = torch.randn(16, 60, dtype=torch.float32, device='cuda')
+    output = model(input_tensor)
+    loss = output.sum()
+    loss.backward()
+
+    original_weight = model.weight.data.clone()
+    optimizer.step()
+
+    assert not torch.equal(
+        model.weight.data, original_weight
+    ), f"Weight should be updated with use_kl_shampoo={use_kl_shampoo}"
+
+
+@skip_no_soap
+@pytest.mark.parametrize("shampoo_beta", [0.5, 0.9, 0.99])
+def test_soap_optimizer_shampoo_beta(shampoo_beta):
+    """Test SOAP optimizer with different shampoo_beta values."""
+
+    model = torch.nn.Linear(60, 30, bias=False, dtype=torch.float32, device='cuda')
+    model.requires_grad_(True)
+    model.weight.data.fill_(1.0)
+
+    optimizer = SOAP(
+        params=[model.weight],
+        lr=0.01,
+        betas=(0.9, 0.999),
+        shampoo_beta=shampoo_beta,
+        precondition_frequency=1,
+    )
+
+    input_tensor = torch.randn(16, 60, dtype=torch.float32, device='cuda')
+    output = model(input_tensor)
+    loss = output.sum()
+    loss.backward()
+
+    original_weight = model.weight.data.clone()
+    optimizer.step()
+
+    assert not torch.equal(
+        model.weight.data, original_weight
+    ), f"Weight should be updated with shampoo_beta={shampoo_beta}"
+
+
+@pytest.mark.skipif(
+    int(os.getenv('WORLD_SIZE', '1')) == 1, reason="Multi-rank test requires WORLD_SIZE > 1"
+)
+class TestSoapOptimizerMultiRank:
+    """Test class for SOAP optimizer with multi-rank setup."""
+
+    @pytest.fixture(autouse=True)
+    def setup_and_teardown(self):
+        """Setup and teardown for each test."""
+        Utils.initialize_model_parallel()
+        yield
+        Utils.destroy_model_parallel()
+
+    def create_ddp_model(self, model):
+        """Wrap model in DDP."""
+        ddp_config = DistributedDataParallelConfig(use_distributed_optimizer=False)
+        return DistributedDataParallel(
+            TransformerConfig(num_attention_heads=1, num_layers=1), ddp_config, model
+        )
+
+    def test_get_megatron_optimizer_soap_smoke(self):
+        """Smoke test for get_megatron_optimizer with SOAP."""
+        model = Net().bfloat16().cuda()
+        model.requires_grad_(True)
+        model = self.create_ddp_model(model)
+
+        for param in model.parameters():
+            assert param.requires_grad, "All parameters should require gradients"
+
+        optimizer_config = OptimizerConfig(
+            optimizer='soap',
+            lr=0.01,
+            weight_decay=0.01,
+            bf16=True,
+            use_distributed_optimizer=False,
+            soap_shampoo_beta=0.95,
+            soap_precondition_frequency=1,
+            soap_use_kl_shampoo=True,
+        )
+
+        optimizer = get_megatron_optimizer(
+            config=optimizer_config, model_chunks=[model], use_gloo_process_groups=True
+        )
+
+        assert optimizer is not None, "Optimizer should not be None"
+        assert hasattr(optimizer, 'param_groups'), "Optimizer should have param_groups"
+        assert hasattr(optimizer, 'chained_optimizers'), "Should be a ChainedOptimizer"
+        assert len(optimizer.chained_optimizers) >= 1, "Should have at least one chained optimizer"
+
+        # Test forward and backward pass
+        input_tensor = torch.randn(16, 80, dtype=torch.bfloat16, device='cuda')
+        output = model(input_tensor)
+        loss = output.sum()
+        loss.backward()
+
+        # Store original parameters
+        original_params = {}
+        for name, param in model.named_parameters():
+            original_params[name] = param.data.clone()
+
+        # Test optimizer step
+        optimizer.step()
+
+        # Verify at least some parameters were updated
+        params_updated = 0
+        for name, param in model.named_parameters():
+            if not torch.equal(param.data, original_params[name]):
+                params_updated += 1
+
+        assert params_updated > 0, "At least some parameters should be updated after optimizer step"
+
+        # Test zero_grad
+        optimizer.zero_grad()
+        for param in model.parameters():
+            assert param.grad is None or torch.all(
+                param.grad == 0
+            ), "Gradients should be zeroed for all parameters"
+
+        # Test state_dict and load_state_dict
+        state_dict = optimizer.state_dict()
+        assert isinstance(state_dict, list), "State dict should be a list"
+        optimizer.load_state_dict(state_dict)
+
+    def test_get_megatron_optimizer_soap_validation(self):
+        """Test validation logic for get_megatron_optimizer with SOAP."""
+        model = torch.nn.Linear(100, 50, bias=False, dtype=torch.bfloat16, device='cuda')
+        model.requires_grad_(True)
+        model = self.create_ddp_model(model)
+
+        # FP16 should raise exception
+        optimizer_config_fp16 = OptimizerConfig(
+            optimizer='soap', lr=0.01, fp16=True, use_distributed_optimizer=False
+        )
+
+        with pytest.raises(Exception, match='emerging optimizer with fp16 is not supported'):
+            get_megatron_optimizer(config=optimizer_config_fp16, model_chunks=[model])
diff --git a/uv.lock b/uv.lock
index 433e8b3ea8e..08482b4b7b8 100644
--- a/uv.lock
+++ b/uv.lock
@@ -1,5 +1,5 @@
 version = 1
-revision = 2
+revision = 3
 requires-python = ">=3.10"
 resolution-markers = [
     "python_full_version >= '3.14' and sys_platform == 'linux'",
@@ -107,7 +107,7 @@ source = { registry = "https://pypi.org/simple" }
 dependencies = [
     { name = "aiohappyeyeballs" },
     { name = "aiosignal" },
-    { name = "async-timeout", marker = "python_full_version < '3.11'" },
+    { name = "async-timeout", marker = "python_full_version < '3.11' or (extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
     { name = "attrs" },
     { name = "frozenlist" },
     { name = "multidict" },
@@ -247,7 +247,7 @@ version = "1.4.0"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
     { name = "frozenlist" },
-    { name = "typing-extensions", marker = "python_full_version < '3.13'" },
+    { name = "typing-extensions", marker = "python_full_version < '3.13' or (extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
 ]
 sdist = { url = "https://files.pythonhosted.org/packages/61/62/06741b579156360248d1ec624842ad0edf697050bbaf7c3e46394e106ad1/aiosignal-1.4.0.tar.gz", hash = "sha256:f47eecd9468083c2029cc99945502cb7708b082c232f9aca65da147157b251c7", size = 25007, upload-time = "2025-07-03T22:54:43.528Z" }
 wheels = [
@@ -301,10 +301,10 @@ name = "anyio"
 version = "4.9.0"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
-    { name = "exceptiongroup", marker = "python_full_version < '3.11'" },
+    { name = "exceptiongroup", marker = "python_full_version < '3.11' or (extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
     { name = "idna" },
     { name = "sniffio" },
-    { name = "typing-extensions", marker = "python_full_version < '3.13'" },
+    { name = "typing-extensions", marker = "python_full_version < '3.13' or (extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
 ]
 sdist = { url = "https://files.pythonhosted.org/packages/95/7d/4c1bd541d4dffa1b52bd83fb8527089e097a106fc90b467a7313b105f840/anyio-4.9.0.tar.gz", hash = "sha256:673c0c244e15788651a4ff38710fea9675823028a6f08a5eda409e0c9840a028", size = 190949, upload-time = "2025-03-17T00:02:54.77Z" }
 wheels = [
@@ -749,7 +749,7 @@ name = "cffi"
 version = "2.0.0"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
-    { name = "pycparser", marker = "implementation_name != 'PyPy'" },
+    { name = "pycparser", marker = "implementation_name != 'PyPy' or (extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
 ]
 sdist = { url = "https://files.pythonhosted.org/packages/eb/56/b1ba7935a17738ae8453301356628e8147c79dbb825bcbc73dc7401f9846/cffi-2.0.0.tar.gz", hash = "sha256:44d1b5909021139fe36001ae048dbdde8214afa20200eda0f64c068cac5d5529", size = 523588, upload-time = "2025-09-08T23:24:04.541Z" }
 wheels = [
@@ -920,7 +920,7 @@ name = "click"
 version = "8.3.1"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
-    { name = "colorama", marker = "sys_platform == 'win32'" },
+    { name = "colorama", marker = "sys_platform == 'win32' or (extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
 ]
 sdist = { url = "https://files.pythonhosted.org/packages/3d/fa/656b739db8587d7b5dfa22e22ed02566950fbfbcdc20311993483657a5c0/click-8.3.1.tar.gz", hash = "sha256:12ff4785d337a1bb490bb7e9c2b1ee5da3112e94a8622f26a6c77f5d2fc6842a", size = 295065, upload-time = "2025-11-15T20:45:42.706Z" }
 wheels = [
@@ -1410,12 +1410,11 @@ wheels = [
 
 [[package]]
 name = "emerging-optimizers"
-version = "0.1.0"
-source = { git = "https://github.com/NVIDIA-NeMo/Emerging-Optimizers.git?rev=v0.1.0#d5363b4a418128cd8111983b191c4b8869a9766b" }
+version = "0.2.0"
+source = { git = "https://github.com/NVIDIA-NeMo/Emerging-Optimizers.git?rev=a8faf641d5fca6a0515cfc010b6cedbf488cc33a#a8faf641d5fca6a0515cfc010b6cedbf488cc33a" }
 dependencies = [
-    { name = "absl-py" },
-    { name = "torch", marker = "sys_platform == 'never'" },
-    { name = "typing-extensions" },
+    { name = "absl-py", marker = "python_full_version >= '3.12' or (extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
+    { name = "torch", marker = "(python_full_version >= '3.12' and sys_platform == 'never') or (python_full_version < '3.12' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (sys_platform != 'never' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
 ]
 
 [[package]]
@@ -1423,7 +1422,7 @@ name = "exceptiongroup"
 version = "1.3.1"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
-    { name = "typing-extensions", marker = "python_full_version < '3.13'" },
+    { name = "typing-extensions", marker = "python_full_version < '3.13' or (extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
 ]
 sdist = { url = "https://files.pythonhosted.org/packages/50/79/66800aadf48771f6b62f7eb014e352e5d06856655206165d775e675a02c9/exceptiongroup-1.3.1.tar.gz", hash = "sha256:8b412432c6055b0b7d14c310000ae93352ed6754f70fa8f7c34141f91c4e3219", size = 30371, upload-time = "2025-11-21T23:01:54.787Z" }
 wheels = [
@@ -2027,7 +2026,7 @@ dependencies = [
     { name = "filelock" },
     { name = "fsspec", version = "2025.10.0", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version < '3.14' or sys_platform != 'linux' or extra == 'extra-13-megatron-core-dev' or extra == 'extra-13-megatron-core-lts'" },
     { name = "fsspec", version = "2026.2.0", source = { registry = "https://pypi.org/simple" }, marker = "(python_full_version >= '3.14' and sys_platform == 'linux' and extra != 'extra-13-megatron-core-dev' and extra != 'extra-13-megatron-core-lts') or (extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
-    { name = "hf-xet", marker = "platform_machine == 'aarch64' or platform_machine == 'amd64' or platform_machine == 'arm64' or platform_machine == 'x86_64'" },
+    { name = "hf-xet", marker = "platform_machine == 'aarch64' or platform_machine == 'amd64' or platform_machine == 'arm64' or platform_machine == 'x86_64' or (extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
     { name = "packaging" },
     { name = "pyyaml" },
     { name = "requests" },
@@ -2462,7 +2461,7 @@ resolution-markers = [
     "python_full_version < '3.11' and sys_platform != 'linux'",
 ]
 dependencies = [
-    { name = "mdurl", marker = "python_full_version < '3.11'" },
+    { name = "mdurl", marker = "python_full_version < '3.11' or (extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
 ]
 sdist = { url = "https://files.pythonhosted.org/packages/38/71/3b932df36c1a044d397a1f92d1cf91ee0a503d91e470cbd670aa66b07ed0/markdown-it-py-3.0.0.tar.gz", hash = "sha256:e3f60a94fa066dc52ec76661e37c851cb232d92f9886b15cb560aaada2df8feb", size = 74596, upload-time = "2023-06-03T06:41:14.443Z" }
 wheels = [
@@ -2492,7 +2491,7 @@ resolution-markers = [
     "python_full_version == '3.11.*' and sys_platform != 'emscripten' and sys_platform != 'linux' and sys_platform != 'win32'",
 ]
 dependencies = [
-    { name = "mdurl", marker = "python_full_version >= '3.11'" },
+    { name = "mdurl", marker = "python_full_version >= '3.11' or (extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
 ]
 sdist = { url = "https://files.pythonhosted.org/packages/5b/f5/4ec618ed16cc4f8fb3b701563655a69816155e79e24a17b651541804721d/markdown_it_py-4.0.0.tar.gz", hash = "sha256:cb0a2b4aa34f932c007117b194e945bd74e0ec24133ceb5bac59009cda1cb9f3", size = 73070, upload-time = "2025-08-11T12:57:52.854Z" }
 wheels = [
@@ -2631,7 +2630,7 @@ dev = [
     { name = "causal-conv1d" },
     { name = "datasets" },
     { name = "einops" },
-    { name = "emerging-optimizers" },
+    { name = "emerging-optimizers", marker = "python_full_version >= '3.12'" },
     { name = "fastapi" },
     { name = "flash-linear-attention" },
     { name = "flashinfer-python" },
@@ -2658,7 +2657,7 @@ lts = [
     { name = "causal-conv1d" },
     { name = "datasets" },
     { name = "einops" },
-    { name = "emerging-optimizers" },
+    { name = "emerging-optimizers", marker = "python_full_version >= '3.12'" },
     { name = "fastapi" },
     { name = "flashinfer-python" },
     { name = "mamba-ssm" },
@@ -2718,7 +2717,7 @@ linting = [
     { name = "ruff" },
 ]
 no-pypi-wheels = [
-    { name = "emerging-optimizers" },
+    { name = "emerging-optimizers", marker = "python_full_version >= '3.12' or (extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
     { name = "fast-hadamard-transform" },
 ]
 test = [
@@ -2749,8 +2748,8 @@ requires-dist = [
     { name = "datasets", marker = "extra == 'lts'" },
     { name = "einops", marker = "extra == 'dev'", specifier = "~=0.8" },
     { name = "einops", marker = "extra == 'lts'", specifier = "~=0.8" },
-    { name = "emerging-optimizers", marker = "extra == 'dev'", git = "https://github.com/NVIDIA-NeMo/Emerging-Optimizers.git?rev=v0.1.0" },
-    { name = "emerging-optimizers", marker = "extra == 'lts'", git = "https://github.com/NVIDIA-NeMo/Emerging-Optimizers.git?rev=v0.1.0" },
+    { name = "emerging-optimizers", marker = "python_full_version >= '3.12' and extra == 'dev'", git = "https://github.com/NVIDIA-NeMo/Emerging-Optimizers.git?rev=a8faf641d5fca6a0515cfc010b6cedbf488cc33a" },
+    { name = "emerging-optimizers", marker = "python_full_version >= '3.12' and extra == 'lts'", git = "https://github.com/NVIDIA-NeMo/Emerging-Optimizers.git?rev=a8faf641d5fca6a0515cfc010b6cedbf488cc33a" },
     { name = "fastapi", marker = "extra == 'dev'", specifier = "~=0.50" },
     { name = "fastapi", marker = "extra == 'lts'", specifier = "~=0.50" },
     { name = "flash-linear-attention", marker = "extra == 'dev'", specifier = "~=0.4.0" },
@@ -2824,7 +2823,7 @@ linting = [
     { name = "ruff", specifier = "~=0.9.0" },
 ]
 no-pypi-wheels = [
-    { name = "emerging-optimizers", git = "https://github.com/NVIDIA-NeMo/Emerging-Optimizers.git?rev=v0.1.0" },
+    { name = "emerging-optimizers", marker = "python_full_version >= '3.12'", git = "https://github.com/NVIDIA-NeMo/Emerging-Optimizers.git?rev=a8faf641d5fca6a0515cfc010b6cedbf488cc33a" },
     { name = "fast-hadamard-transform", git = "https://github.com/Dao-AILab/fast-hadamard-transform.git?rev=f134af63deb2df17e1171a9ec1ea4a7d8604d5ca" },
 ]
 test = [
@@ -3030,7 +3029,7 @@ name = "multidict"
 version = "6.7.1"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
-    { name = "typing-extensions", marker = "python_full_version < '3.11'" },
+    { name = "typing-extensions", marker = "python_full_version < '3.11' or (extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
 ]
 sdist = { url = "https://files.pythonhosted.org/packages/1a/c2/c2d94cbe6ac1753f3fc980da97b3d930efe1da3af3c9f5125354436c073d/multidict-6.7.1.tar.gz", hash = "sha256:ec6652a1bee61c53a3e5776b6049172c53b6aaba34f18c9ad04f82712bac623d", size = 102010, upload-time = "2026-01-26T02:46:45.979Z" }
 wheels = [
@@ -3572,6 +3571,59 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/8a/86/94188e03e5d4dd7b73c390b0cddcde5618b3799c18e327b2bf15763f6137/nvdlfw_inspect-0.2.2-py3-none-any.whl", hash = "sha256:8a4dc2814c5a4cd19ae304170b9bfa514538ef3c3eb243a45a82404ec3cb279d", size = 30964, upload-time = "2025-12-03T10:52:01.933Z" },
 ]
 
+[[package]]
+name = "nvidia-cublas-cu12"
+version = "12.8.4.1"
+source = { registry = "https://pypi.org/simple" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/29/99/db44d685f0e257ff0e213ade1964fc459b4a690a73293220e98feb3307cf/nvidia_cublas_cu12-12.8.4.1-py3-none-manylinux_2_27_aarch64.whl", hash = "sha256:b86f6dd8935884615a0683b663891d43781b819ac4f2ba2b0c9604676af346d0", size = 590537124, upload-time = "2025-03-07T01:43:53.556Z" },
+    { url = "https://files.pythonhosted.org/packages/dc/61/e24b560ab2e2eaeb3c839129175fb330dfcfc29e5203196e5541a4c44682/nvidia_cublas_cu12-12.8.4.1-py3-none-manylinux_2_27_x86_64.whl", hash = "sha256:8ac4e771d5a348c551b2a426eda6193c19aa630236b418086020df5ba9667142", size = 594346921, upload-time = "2025-03-07T01:44:31.254Z" },
+    { url = "https://files.pythonhosted.org/packages/70/61/7d7b3c70186fb651d0fbd35b01dbfc8e755f69fd58f817f3d0f642df20c3/nvidia_cublas_cu12-12.8.4.1-py3-none-win_amd64.whl", hash = "sha256:47e9b82132fa8d2b4944e708049229601448aaad7e6f296f630f2d1a32de35af", size = 567544208, upload-time = "2025-03-07T01:53:30.535Z" },
+]
+
+[[package]]
+name = "nvidia-cuda-cupti-cu12"
+version = "12.8.90"
+source = { registry = "https://pypi.org/simple" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/d5/1f/b3bd73445e5cb342727fd24fe1f7b748f690b460acadc27ea22f904502c8/nvidia_cuda_cupti_cu12-12.8.90-py3-none-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:4412396548808ddfed3f17a467b104ba7751e6b58678a4b840675c56d21cf7ed", size = 9533318, upload-time = "2025-03-07T01:40:10.421Z" },
+    { url = "https://files.pythonhosted.org/packages/f8/02/2adcaa145158bf1a8295d83591d22e4103dbfd821bcaf6f3f53151ca4ffa/nvidia_cuda_cupti_cu12-12.8.90-py3-none-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:ea0cb07ebda26bb9b29ba82cda34849e73c166c18162d3913575b0c9db9a6182", size = 10248621, upload-time = "2025-03-07T01:40:21.213Z" },
+    { url = "https://files.pythonhosted.org/packages/41/bc/83f5426095d93694ae39fe1311431b5d5a9bb82e48bf0dd8e19be2765942/nvidia_cuda_cupti_cu12-12.8.90-py3-none-win_amd64.whl", hash = "sha256:bb479dcdf7e6d4f8b0b01b115260399bf34154a1a2e9fe11c85c517d87efd98e", size = 7015759, upload-time = "2025-03-07T01:51:11.355Z" },
+]
+
+[[package]]
+name = "nvidia-cuda-nvrtc-cu12"
+version = "12.8.93"
+source = { registry = "https://pypi.org/simple" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/05/6b/32f747947df2da6994e999492ab306a903659555dddc0fbdeb9d71f75e52/nvidia_cuda_nvrtc_cu12-12.8.93-py3-none-manylinux2010_x86_64.manylinux_2_12_x86_64.whl", hash = "sha256:a7756528852ef889772a84c6cd89d41dfa74667e24cca16bb31f8f061e3e9994", size = 88040029, upload-time = "2025-03-07T01:42:13.562Z" },
+    { url = "https://files.pythonhosted.org/packages/eb/d1/e50d0acaab360482034b84b6e27ee83c6738f7d32182b987f9c7a4e32962/nvidia_cuda_nvrtc_cu12-12.8.93-py3-none-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:fc1fec1e1637854b4c0a65fb9a8346b51dd9ee69e61ebaccc82058441f15bce8", size = 43106076, upload-time = "2025-03-07T01:41:59.817Z" },
+    { url = "https://files.pythonhosted.org/packages/45/51/52a3d84baa2136cc8df15500ad731d74d3a1114d4c123e043cb608d4a32b/nvidia_cuda_nvrtc_cu12-12.8.93-py3-none-win_amd64.whl", hash = "sha256:7a4b6b2904850fe78e0bd179c4b655c404d4bb799ef03ddc60804247099ae909", size = 73586838, upload-time = "2025-03-07T01:52:13.483Z" },
+]
+
+[[package]]
+name = "nvidia-cuda-runtime-cu12"
+version = "12.8.90"
+source = { registry = "https://pypi.org/simple" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/7c/75/f865a3b236e4647605ea34cc450900854ba123834a5f1598e160b9530c3a/nvidia_cuda_runtime_cu12-12.8.90-py3-none-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:52bf7bbee900262ffefe5e9d5a2a69a30d97e2bc5bb6cc866688caa976966e3d", size = 965265, upload-time = "2025-03-07T01:39:43.533Z" },
+    { url = "https://files.pythonhosted.org/packages/0d/9b/a997b638fcd068ad6e4d53b8551a7d30fe8b404d6f1804abf1df69838932/nvidia_cuda_runtime_cu12-12.8.90-py3-none-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:adade8dcbd0edf427b7204d480d6066d33902cab2a4707dcfc48a2d0fd44ab90", size = 954765, upload-time = "2025-03-07T01:40:01.615Z" },
+    { url = "https://files.pythonhosted.org/packages/30/a5/a515b7600ad361ea14bfa13fb4d6687abf500adc270f19e89849c0590492/nvidia_cuda_runtime_cu12-12.8.90-py3-none-win_amd64.whl", hash = "sha256:c0c6027f01505bfed6c3b21ec546f69c687689aad5f1a377554bc6ca4aa993a8", size = 944318, upload-time = "2025-03-07T01:51:01.794Z" },
+]
+
+[[package]]
+name = "nvidia-cudnn-cu12"
+version = "9.10.2.21"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "nvidia-cublas-cu12" },
+]
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/fa/41/e79269ce215c857c935fd86bcfe91a451a584dfc27f1e068f568b9ad1ab7/nvidia_cudnn_cu12-9.10.2.21-py3-none-manylinux_2_27_aarch64.whl", hash = "sha256:c9132cc3f8958447b4910a1720036d9eff5928cc3179b0a51fb6d167c6cc87d8", size = 705026878, upload-time = "2025-06-06T21:52:51.348Z" },
+    { url = "https://files.pythonhosted.org/packages/ba/51/e123d997aa098c61d029f76663dedbfb9bc8dcf8c60cbd6adbe42f76d049/nvidia_cudnn_cu12-9.10.2.21-py3-none-manylinux_2_27_x86_64.whl", hash = "sha256:949452be657fa16687d0930933f032835951ef0892b37d2d53824d1a84dc97a8", size = 706758467, upload-time = "2025-06-06T21:54:08.597Z" },
+    { url = "https://files.pythonhosted.org/packages/3d/90/0bd6e586701b3a890fd38aa71c387dab4883d619d6e5ad912ccbd05bfd67/nvidia_cudnn_cu12-9.10.2.21-py3-none-win_amd64.whl", hash = "sha256:c6288de7d63e6cf62988f0923f96dc339cea362decb1bf5b3141883392a7d65e", size = 692992268, upload-time = "2025-06-06T21:55:18.114Z" },
+]
+
 [[package]]
 name = "nvidia-cudnn-frontend"
 version = "1.18.0"
@@ -3594,6 +3646,76 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/67/53/df2810b56d259ef96fa6beaa1381bd14c29fbe82836b409516e864c5e177/nvidia_cudnn_frontend-1.18.0-cp314-cp314-win_amd64.whl", hash = "sha256:5053b473fa74168b5fbf35934cd6187f88aa03b8447b9f2cd417332d5e5c9569", size = 1592759, upload-time = "2026-02-11T21:32:33.87Z" },
 ]
 
+[[package]]
+name = "nvidia-cufft-cu12"
+version = "11.3.3.83"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "nvidia-nvjitlink-cu12" },
+]
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/60/bc/7771846d3a0272026c416fbb7e5f4c1f146d6d80704534d0b187dd6f4800/nvidia_cufft_cu12-11.3.3.83-py3-none-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:848ef7224d6305cdb2a4df928759dca7b1201874787083b6e7550dd6765ce69a", size = 193109211, upload-time = "2025-03-07T01:44:56.873Z" },
+    { url = "https://files.pythonhosted.org/packages/1f/13/ee4e00f30e676b66ae65b4f08cb5bcbb8392c03f54f2d5413ea99a5d1c80/nvidia_cufft_cu12-11.3.3.83-py3-none-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:4d2dd21ec0b88cf61b62e6b43564355e5222e4a3fb394cac0db101f2dd0d4f74", size = 193118695, upload-time = "2025-03-07T01:45:27.821Z" },
+    { url = "https://files.pythonhosted.org/packages/7d/ec/ce1629f1e478bb5ccd208986b5f9e0316a78538dd6ab1d0484f012f8e2a1/nvidia_cufft_cu12-11.3.3.83-py3-none-win_amd64.whl", hash = "sha256:7a64a98ef2a7c47f905aaf8931b69a3a43f27c55530c698bb2ed7c75c0b42cb7", size = 192216559, upload-time = "2025-03-07T01:53:57.106Z" },
+]
+
+[[package]]
+name = "nvidia-cufile-cu12"
+version = "1.13.1.3"
+source = { registry = "https://pypi.org/simple" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/bb/fe/1bcba1dfbfb8d01be8d93f07bfc502c93fa23afa6fd5ab3fc7c1df71038a/nvidia_cufile_cu12-1.13.1.3-py3-none-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:1d069003be650e131b21c932ec3d8969c1715379251f8d23a1860554b1cb24fc", size = 1197834, upload-time = "2025-03-07T01:45:50.723Z" },
+    { url = "https://files.pythonhosted.org/packages/1e/f5/5607710447a6fe9fd9b3283956fceeee8a06cda1d2f56ce31371f595db2a/nvidia_cufile_cu12-1.13.1.3-py3-none-manylinux_2_27_aarch64.whl", hash = "sha256:4beb6d4cce47c1a0f1013d72e02b0994730359e17801d395bdcbf20cfb3bb00a", size = 1120705, upload-time = "2025-03-07T01:45:41.434Z" },
+]
+
+[[package]]
+name = "nvidia-curand-cu12"
+version = "10.3.9.90"
+source = { registry = "https://pypi.org/simple" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/45/5e/92aa15eca622a388b80fbf8375d4760738df6285b1e92c43d37390a33a9a/nvidia_curand_cu12-10.3.9.90-py3-none-manylinux_2_27_aarch64.whl", hash = "sha256:dfab99248034673b779bc6decafdc3404a8a6f502462201f2f31f11354204acd", size = 63625754, upload-time = "2025-03-07T01:46:10.735Z" },
+    { url = "https://files.pythonhosted.org/packages/fb/aa/6584b56dc84ebe9cf93226a5cde4d99080c8e90ab40f0c27bda7a0f29aa1/nvidia_curand_cu12-10.3.9.90-py3-none-manylinux_2_27_x86_64.whl", hash = "sha256:b32331d4f4df5d6eefa0554c565b626c7216f87a06a4f56fab27c3b68a830ec9", size = 63619976, upload-time = "2025-03-07T01:46:23.323Z" },
+    { url = "https://files.pythonhosted.org/packages/b9/75/70c05b2f3ed5be3bb30b7102b6eb78e100da4bbf6944fd6725c012831cab/nvidia_curand_cu12-10.3.9.90-py3-none-win_amd64.whl", hash = "sha256:f149a8ca457277da854f89cf282d6ef43176861926c7ac85b2a0fbd237c587ec", size = 62765309, upload-time = "2025-03-07T01:54:20.478Z" },
+]
+
+[[package]]
+name = "nvidia-cusolver-cu12"
+version = "11.7.3.90"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "nvidia-cublas-cu12" },
+    { name = "nvidia-cusparse-cu12" },
+    { name = "nvidia-nvjitlink-cu12" },
+]
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/c8/32/f7cd6ce8a7690544d084ea21c26e910a97e077c9b7f07bf5de623ee19981/nvidia_cusolver_cu12-11.7.3.90-py3-none-manylinux_2_27_aarch64.whl", hash = "sha256:db9ed69dbef9715071232caa9b69c52ac7de3a95773c2db65bdba85916e4e5c0", size = 267229841, upload-time = "2025-03-07T01:46:54.356Z" },
+    { url = "https://files.pythonhosted.org/packages/85/48/9a13d2975803e8cf2777d5ed57b87a0b6ca2cc795f9a4f59796a910bfb80/nvidia_cusolver_cu12-11.7.3.90-py3-none-manylinux_2_27_x86_64.whl", hash = "sha256:4376c11ad263152bd50ea295c05370360776f8c3427b30991df774f9fb26c450", size = 267506905, upload-time = "2025-03-07T01:47:16.273Z" },
+    { url = "https://files.pythonhosted.org/packages/13/c0/76ca8551b8a84146ffa189fec81c26d04adba4bc0dbe09cd6e6fd9b7de04/nvidia_cusolver_cu12-11.7.3.90-py3-none-win_amd64.whl", hash = "sha256:4a550db115fcabc4d495eb7d39ac8b58d4ab5d8e63274d3754df1c0ad6a22d34", size = 256720438, upload-time = "2025-03-07T01:54:39.898Z" },
+]
+
+[[package]]
+name = "nvidia-cusparse-cu12"
+version = "12.5.8.93"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "nvidia-nvjitlink-cu12" },
+]
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/bc/f7/cd777c4109681367721b00a106f491e0d0d15cfa1fd59672ce580ce42a97/nvidia_cusparse_cu12-12.5.8.93-py3-none-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:9b6c161cb130be1a07a27ea6923df8141f3c295852f4b260c65f18f3e0a091dc", size = 288117129, upload-time = "2025-03-07T01:47:40.407Z" },
+    { url = "https://files.pythonhosted.org/packages/c2/f5/e1854cb2f2bcd4280c44736c93550cc300ff4b8c95ebe370d0aa7d2b473d/nvidia_cusparse_cu12-12.5.8.93-py3-none-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:1ec05d76bbbd8b61b06a80e1eaf8cf4959c3d4ce8e711b65ebd0443bb0ebb13b", size = 288216466, upload-time = "2025-03-07T01:48:13.779Z" },
+    { url = "https://files.pythonhosted.org/packages/62/07/f3b2ad63f8e3d257a599f422ae34eb565e70c41031aecefa3d18b62cabd1/nvidia_cusparse_cu12-12.5.8.93-py3-none-win_amd64.whl", hash = "sha256:9a33604331cb2cac199f2e7f5104dfbb8a5a898c367a53dfda9ff2acb6b6b4dd", size = 284937404, upload-time = "2025-03-07T01:55:07.742Z" },
+]
+
+[[package]]
+name = "nvidia-cusparselt-cu12"
+version = "0.7.1"
+source = { registry = "https://pypi.org/simple" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/73/b9/598f6ff36faaece4b3c50d26f50e38661499ff34346f00e057760b35cc9d/nvidia_cusparselt_cu12-0.7.1-py3-none-manylinux2014_aarch64.whl", hash = "sha256:8878dce784d0fac90131b6817b607e803c36e629ba34dc5b433471382196b6a5", size = 283835557, upload-time = "2025-02-26T00:16:54.265Z" },
+    { url = "https://files.pythonhosted.org/packages/56/79/12978b96bd44274fe38b5dde5cfb660b1d114f70a65ef962bcbbed99b549/nvidia_cusparselt_cu12-0.7.1-py3-none-manylinux2014_x86_64.whl", hash = "sha256:f1bb701d6b930d5a7cea44c19ceb973311500847f81b634d802b7b539dc55623", size = 287193691, upload-time = "2025-02-26T00:15:44.104Z" },
+    { url = "https://files.pythonhosted.org/packages/2f/d8/a6b0d0d0c2435e9310f3e2bb0d9c9dd4c33daef86aa5f30b3681defd37ea/nvidia_cusparselt_cu12-0.7.1-py3-none-win_amd64.whl", hash = "sha256:f67fbb5831940ec829c9117b7f33807db9f9678dc2a617fbe781cac17b4e1075", size = 271020911, upload-time = "2025-02-26T00:14:47.204Z" },
+]
+
 [[package]]
 name = "nvidia-cutlass-dsl"
 version = "4.4.0"
@@ -3667,6 +3789,44 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/16/09/30147ab0d0409d3492f1d37469fe0586c82aeec6eec9a907f59d24094516/nvidia_modelopt-0.41.0-py3-none-any.whl", hash = "sha256:ffa5f903d22653649318831a470550ae55ee04716c068d5ade61c3176fdc1d7d", size = 934582, upload-time = "2026-01-20T17:21:28.494Z" },
 ]
 
+[[package]]
+name = "nvidia-nccl-cu12"
+version = "2.27.5"
+source = { registry = "https://pypi.org/simple" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/bb/1c/857979db0ef194ca5e21478a0612bcdbbe59458d7694361882279947b349/nvidia_nccl_cu12-2.27.5-py3-none-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:31432ad4d1fb1004eb0c56203dc9bc2178a1ba69d1d9e02d64a6938ab5e40e7a", size = 322400625, upload-time = "2025-06-26T04:11:04.496Z" },
+    { url = "https://files.pythonhosted.org/packages/6e/89/f7a07dc961b60645dbbf42e80f2bc85ade7feb9a491b11a1e973aa00071f/nvidia_nccl_cu12-2.27.5-py3-none-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:ad730cf15cb5d25fe849c6e6ca9eb5b76db16a80f13f425ac68d8e2e55624457", size = 322348229, upload-time = "2025-06-26T04:11:28.385Z" },
+]
+
+[[package]]
+name = "nvidia-nvjitlink-cu12"
+version = "12.8.93"
+source = { registry = "https://pypi.org/simple" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/f6/74/86a07f1d0f42998ca31312f998bd3b9a7eff7f52378f4f270c8679c77fb9/nvidia_nvjitlink_cu12-12.8.93-py3-none-manylinux2010_x86_64.manylinux_2_12_x86_64.whl", hash = "sha256:81ff63371a7ebd6e6451970684f916be2eab07321b73c9d244dc2b4da7f73b88", size = 39254836, upload-time = "2025-03-07T01:49:55.661Z" },
+    { url = "https://files.pythonhosted.org/packages/2a/a2/8cee5da30d13430e87bf99bb33455d2724d0a4a9cb5d7926d80ccb96d008/nvidia_nvjitlink_cu12-12.8.93-py3-none-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:adccd7161ace7261e01bb91e44e88da350895c270d23f744f0820c818b7229e7", size = 38386204, upload-time = "2025-03-07T01:49:43.612Z" },
+    { url = "https://files.pythonhosted.org/packages/ed/d7/34f02dad2e30c31b10a51f6b04e025e5dd60e5f936af9045a9b858a05383/nvidia_nvjitlink_cu12-12.8.93-py3-none-win_amd64.whl", hash = "sha256:bd93fbeeee850917903583587f4fc3a4eafa022e34572251368238ab5e6bd67f", size = 268553710, upload-time = "2025-03-07T01:56:24.13Z" },
+]
+
+[[package]]
+name = "nvidia-nvshmem-cu12"
+version = "3.4.5"
+source = { registry = "https://pypi.org/simple" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/1d/6a/03aa43cc9bd3ad91553a88b5f6fb25ed6a3752ae86ce2180221962bc2aa5/nvidia_nvshmem_cu12-3.4.5-py3-none-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:0b48363fc6964dede448029434c6abed6c5e37f823cb43c3bcde7ecfc0457e15", size = 138936938, upload-time = "2025-09-06T00:32:05.589Z" },
+    { url = "https://files.pythonhosted.org/packages/b5/09/6ea3ea725f82e1e76684f0708bbedd871fc96da89945adeba65c3835a64c/nvidia_nvshmem_cu12-3.4.5-py3-none-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:042f2500f24c021db8a06c5eec2539027d57460e1c1a762055a6554f72c369bd", size = 139103095, upload-time = "2025-09-06T00:32:31.266Z" },
+]
+
+[[package]]
+name = "nvidia-nvtx-cu12"
+version = "12.8.90"
+source = { registry = "https://pypi.org/simple" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/10/c0/1b303feea90d296f6176f32a2a70b5ef230f9bdeb3a72bddb0dc922dc137/nvidia_nvtx_cu12-12.8.90-py3-none-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:d7ad891da111ebafbf7e015d34879f7112832fc239ff0d7d776b6cb685274615", size = 91161, upload-time = "2025-03-07T01:42:23.922Z" },
+    { url = "https://files.pythonhosted.org/packages/a2/eb/86626c1bbc2edb86323022371c39aa48df6fd8b0a1647bc274577f72e90b/nvidia_nvtx_cu12-12.8.90-py3-none-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:5b17e2001cc0d751a5bc2c6ec6d26ad95913324a4adb86788c944f8ce9ba441f", size = 89954, upload-time = "2025-03-07T01:42:44.131Z" },
+    { url = "https://files.pythonhosted.org/packages/9f/99/4c9c0c329bf9fc125008c3b54c7c94c0023518d06fc025ae36431375e1fe/nvidia_nvtx_cu12-12.8.90-py3-none-win_amd64.whl", hash = "sha256:619c8304aedc69f02ea82dd244541a83c3d9d40993381b3b590f1adaed3db41e", size = 56492, upload-time = "2025-03-07T01:52:24.69Z" },
+]
+
 [[package]]
 name = "nvidia-resiliency-ext"
 version = "0.5.0"
@@ -3895,7 +4055,7 @@ wheels = [
 
 [[package]]
 name = "opentelemetry-api"
-version = "1.40.0"
+version = "1.39.1"
 source = { registry = "https://pypi.org/simple" }
 resolution-markers = [
     "python_full_version >= '3.14' and sys_platform == 'linux'",
@@ -3921,9 +4081,9 @@ dependencies = [
     { name = "importlib-metadata", marker = "(extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (extra != 'extra-13-megatron-core-dev' and extra != 'extra-13-megatron-core-lts')" },
     { name = "typing-extensions", marker = "(extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (extra != 'extra-13-megatron-core-dev' and extra != 'extra-13-megatron-core-lts')" },
 ]
-sdist = { url = "https://files.pythonhosted.org/packages/2c/1d/4049a9e8698361cc1a1aa03a6c59e4fa4c71e0c0f94a30f988a6876a2ae6/opentelemetry_api-1.40.0.tar.gz", hash = "sha256:159be641c0b04d11e9ecd576906462773eb97ae1b657730f0ecf64d32071569f", size = 70851, upload-time = "2026-03-04T14:17:21.555Z" }
+sdist = { url = "https://files.pythonhosted.org/packages/97/b9/3161be15bb8e3ad01be8be5a968a9237c3027c5be504362ff800fca3e442/opentelemetry_api-1.39.1.tar.gz", hash = "sha256:fbde8c80e1b937a2c61f20347e91c0c18a1940cecf012d62e65a7caf08967c9c", size = 65767, upload-time = "2025-12-11T13:32:39.182Z" }
 wheels = [
-    { url = "https://files.pythonhosted.org/packages/5f/bf/93795954016c522008da367da292adceed71cca6ee1717e1d64c83089099/opentelemetry_api-1.40.0-py3-none-any.whl", hash = "sha256:82dd69331ae74b06f6a874704be0cfaa49a1650e1537d4a813b86ecef7d0ecf9", size = 68676, upload-time = "2026-03-04T14:17:01.24Z" },
+    { url = "https://files.pythonhosted.org/packages/cf/df/d3f1ddf4bb4cb50ed9b1139cc7b1c54c34a1e7ce8fd1b9a37c0d1551a6bd/opentelemetry_api-1.39.1-py3-none-any.whl", hash = "sha256:2edd8463432a7f8443edce90972169b195e7d6a05500cd29e6d13898187c9950", size = 66356, upload-time = "2025-12-11T13:32:17.304Z" },
 ]
 
 [[package]]
@@ -3962,7 +4122,7 @@ wheels = [
 
 [[package]]
 name = "opentelemetry-exporter-prometheus"
-version = "0.61b0"
+version = "0.60b1"
 source = { registry = "https://pypi.org/simple" }
 resolution-markers = [
     "python_full_version >= '3.14' and sys_platform == 'linux'",
@@ -3985,13 +4145,13 @@ resolution-markers = [
     "python_full_version < '3.11' and sys_platform != 'linux'",
 ]
 dependencies = [
-    { name = "opentelemetry-api", version = "1.40.0", source = { registry = "https://pypi.org/simple" }, marker = "(extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (extra != 'extra-13-megatron-core-dev' and extra != 'extra-13-megatron-core-lts')" },
-    { name = "opentelemetry-sdk", version = "1.40.0", source = { registry = "https://pypi.org/simple" }, marker = "(extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (extra != 'extra-13-megatron-core-dev' and extra != 'extra-13-megatron-core-lts')" },
+    { name = "opentelemetry-api", version = "1.39.1", source = { registry = "https://pypi.org/simple" }, marker = "(extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (extra != 'extra-13-megatron-core-dev' and extra != 'extra-13-megatron-core-lts')" },
+    { name = "opentelemetry-sdk", version = "1.39.1", source = { registry = "https://pypi.org/simple" }, marker = "(extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (extra != 'extra-13-megatron-core-dev' and extra != 'extra-13-megatron-core-lts')" },
     { name = "prometheus-client", marker = "(extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (extra != 'extra-13-megatron-core-dev' and extra != 'extra-13-megatron-core-lts')" },
 ]
-sdist = { url = "https://files.pythonhosted.org/packages/4a/20/9e818fd364d12e8d0cfdce4a3b2d82e24d98c4ceebb315de6b6770b5f214/opentelemetry_exporter_prometheus-0.61b0.tar.gz", hash = "sha256:7c4919bd8e79abd62b610767e80f42c9c3a06c5183f4dd9141eedeb57aea284b", size = 15136, upload-time = "2026-03-04T14:17:26.275Z" }
+sdist = { url = "https://files.pythonhosted.org/packages/14/39/7dafa6fff210737267bed35a8855b6ac7399b9e582b8cf1f25f842517012/opentelemetry_exporter_prometheus-0.60b1.tar.gz", hash = "sha256:a4011b46906323f71724649d301b4dc188aaa068852e814f4df38cc76eac616b", size = 14976, upload-time = "2025-12-11T13:32:42.944Z" }
 wheels = [
-    { url = "https://files.pythonhosted.org/packages/02/4a/b65d40e94d1d930aee73a1a2857211ee6ab10ce3686cbdae5eea78cd9d34/opentelemetry_exporter_prometheus-0.61b0-py3-none-any.whl", hash = "sha256:3013b41f4370143d48d219a2351473761423e5882fa4c213811eaefacba39cb7", size = 13149, upload-time = "2026-03-04T14:17:08.983Z" },
+    { url = "https://files.pythonhosted.org/packages/9b/0d/4be6bf5477a3eb3d917d2f17d3c0b6720cd6cb97898444a61d43cc983f5c/opentelemetry_exporter_prometheus-0.60b1-py3-none-any.whl", hash = "sha256:49f59178de4f4590e3cef0b8b95cf6e071aae70e1f060566df5546fad773b8fd", size = 13019, upload-time = "2025-12-11T13:32:23.974Z" },
 ]
 
 [[package]]
@@ -4042,7 +4202,7 @@ wheels = [
 
 [[package]]
 name = "opentelemetry-sdk"
-version = "1.40.0"
+version = "1.39.1"
 source = { registry = "https://pypi.org/simple" }
 resolution-markers = [
     "python_full_version >= '3.14' and sys_platform == 'linux'",
@@ -4065,13 +4225,13 @@ resolution-markers = [
     "python_full_version < '3.11' and sys_platform != 'linux'",
 ]
 dependencies = [
-    { name = "opentelemetry-api", version = "1.40.0", source = { registry = "https://pypi.org/simple" }, marker = "(extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (extra != 'extra-13-megatron-core-dev' and extra != 'extra-13-megatron-core-lts')" },
-    { name = "opentelemetry-semantic-conventions", version = "0.61b0", source = { registry = "https://pypi.org/simple" }, marker = "(extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (extra != 'extra-13-megatron-core-dev' and extra != 'extra-13-megatron-core-lts')" },
+    { name = "opentelemetry-api", version = "1.39.1", source = { registry = "https://pypi.org/simple" }, marker = "(extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (extra != 'extra-13-megatron-core-dev' and extra != 'extra-13-megatron-core-lts')" },
+    { name = "opentelemetry-semantic-conventions", version = "0.60b1", source = { registry = "https://pypi.org/simple" }, marker = "(extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (extra != 'extra-13-megatron-core-dev' and extra != 'extra-13-megatron-core-lts')" },
     { name = "typing-extensions", marker = "(extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (extra != 'extra-13-megatron-core-dev' and extra != 'extra-13-megatron-core-lts')" },
 ]
-sdist = { url = "https://files.pythonhosted.org/packages/58/fd/3c3125b20ba18ce2155ba9ea74acb0ae5d25f8cd39cfd37455601b7955cc/opentelemetry_sdk-1.40.0.tar.gz", hash = "sha256:18e9f5ec20d859d268c7cb3c5198c8d105d073714db3de50b593b8c1345a48f2", size = 184252, upload-time = "2026-03-04T14:17:31.87Z" }
+sdist = { url = "https://files.pythonhosted.org/packages/eb/fb/c76080c9ba07e1e8235d24cdcc4d125ef7aa3edf23eb4e497c2e50889adc/opentelemetry_sdk-1.39.1.tar.gz", hash = "sha256:cf4d4563caf7bff906c9f7967e2be22d0d6b349b908be0d90fb21c8e9c995cc6", size = 171460, upload-time = "2025-12-11T13:32:49.369Z" }
 wheels = [
-    { url = "https://files.pythonhosted.org/packages/2c/c5/6a852903d8bfac758c6dc6e9a68b015d3c33f2f1be5e9591e0f4b69c7e0a/opentelemetry_sdk-1.40.0-py3-none-any.whl", hash = "sha256:787d2154a71f4b3d81f20524a8ce061b7db667d24e46753f32a7bc48f1c1f3f1", size = 141951, upload-time = "2026-03-04T14:17:17.961Z" },
+    { url = "https://files.pythonhosted.org/packages/7c/98/e91cf858f203d86f4eccdf763dcf01cf03f1dae80c3750f7e635bfa206b6/opentelemetry_sdk-1.39.1-py3-none-any.whl", hash = "sha256:4d5482c478513ecb0a5d938dcc61394e647066e0cc2676bee9f3af3f3f45f01c", size = 132565, upload-time = "2025-12-11T13:32:35.069Z" },
 ]
 
 [[package]]
@@ -4109,7 +4269,7 @@ wheels = [
 
 [[package]]
 name = "opentelemetry-semantic-conventions"
-version = "0.61b0"
+version = "0.60b1"
 source = { registry = "https://pypi.org/simple" }
 resolution-markers = [
     "python_full_version >= '3.14' and sys_platform == 'linux'",
@@ -4132,12 +4292,12 @@ resolution-markers = [
     "python_full_version < '3.11' and sys_platform != 'linux'",
 ]
 dependencies = [
-    { name = "opentelemetry-api", version = "1.40.0", source = { registry = "https://pypi.org/simple" }, marker = "(extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (extra != 'extra-13-megatron-core-dev' and extra != 'extra-13-megatron-core-lts')" },
+    { name = "opentelemetry-api", version = "1.39.1", source = { registry = "https://pypi.org/simple" }, marker = "(extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (extra != 'extra-13-megatron-core-dev' and extra != 'extra-13-megatron-core-lts')" },
     { name = "typing-extensions", marker = "(extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (extra != 'extra-13-megatron-core-dev' and extra != 'extra-13-megatron-core-lts')" },
 ]
-sdist = { url = "https://files.pythonhosted.org/packages/6d/c0/4ae7973f3c2cfd2b6e321f1675626f0dab0a97027cc7a297474c9c8f3d04/opentelemetry_semantic_conventions-0.61b0.tar.gz", hash = "sha256:072f65473c5d7c6dc0355b27d6c9d1a679d63b6d4b4b16a9773062cb7e31192a", size = 145755, upload-time = "2026-03-04T14:17:32.664Z" }
+sdist = { url = "https://files.pythonhosted.org/packages/91/df/553f93ed38bf22f4b999d9be9c185adb558982214f33eae539d3b5cd0858/opentelemetry_semantic_conventions-0.60b1.tar.gz", hash = "sha256:87c228b5a0669b748c76d76df6c364c369c28f1c465e50f661e39737e84bc953", size = 137935, upload-time = "2025-12-11T13:32:50.487Z" }
 wheels = [
-    { url = "https://files.pythonhosted.org/packages/b2/37/cc6a55e448deaa9b27377d087da8615a3416d8ad523d5960b78dbeadd02a/opentelemetry_semantic_conventions-0.61b0-py3-none-any.whl", hash = "sha256:fa530a96be229795f8cef353739b618148b0fe2b4b3f005e60e262926c4d38e2", size = 231621, upload-time = "2026-03-04T14:17:19.33Z" },
+    { url = "https://files.pythonhosted.org/packages/7a/5e/5958555e09635d09b75de3c4f8b9cae7335ca545d77392ffe7331534c402/opentelemetry_semantic_conventions-0.60b1-py3-none-any.whl", hash = "sha256:9fa8c8b0c110da289809292b0591220d3a7b53c1526a23021e977d68597893fb", size = 219982, upload-time = "2025-12-11T13:32:36.955Z" },
 ]
 
 [[package]]
@@ -4167,10 +4327,10 @@ resolution-markers = [
     "python_full_version < '3.11' and sys_platform != 'linux'",
 ]
 dependencies = [
-    { name = "numpy", version = "2.2.6", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version < '3.11'" },
-    { name = "python-dateutil", marker = "python_full_version < '3.11'" },
-    { name = "pytz", marker = "python_full_version < '3.11'" },
-    { name = "tzdata", marker = "python_full_version < '3.11'" },
+    { name = "numpy", version = "2.2.6", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version < '3.11' or (extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
+    { name = "python-dateutil", marker = "python_full_version < '3.11' or (extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
+    { name = "pytz", marker = "python_full_version < '3.11' or (extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
+    { name = "tzdata", marker = "python_full_version < '3.11' or (extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
 ]
 sdist = { url = "https://files.pythonhosted.org/packages/33/01/d40b85317f86cf08d853a4f495195c73815fdf205eef3993821720274518/pandas-2.3.3.tar.gz", hash = "sha256:e05e1af93b977f7eafa636d043f9f94c7ee3ac81af99c13508215942e64c993b", size = 4495223, upload-time = "2025-09-29T23:34:51.853Z" }
 wheels = [
@@ -4246,9 +4406,9 @@ resolution-markers = [
     "python_full_version == '3.11.*' and sys_platform != 'emscripten' and sys_platform != 'linux' and sys_platform != 'win32'",
 ]
 dependencies = [
-    { name = "numpy", version = "2.4.2", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version >= '3.11'" },
-    { name = "python-dateutil", marker = "python_full_version >= '3.11'" },
-    { name = "tzdata", marker = "(python_full_version >= '3.11' and sys_platform == 'emscripten') or (python_full_version >= '3.11' and sys_platform == 'win32')" },
+    { name = "numpy", version = "2.4.2", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version >= '3.11' or (extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
+    { name = "python-dateutil", marker = "python_full_version >= '3.11' or (extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
+    { name = "tzdata", marker = "(python_full_version >= '3.11' and sys_platform == 'emscripten') or (python_full_version >= '3.11' and sys_platform == 'win32') or (sys_platform != 'emscripten' and sys_platform != 'win32' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (sys_platform == 'emscripten' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (sys_platform == 'win32' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
 ]
 sdist = { url = "https://files.pythonhosted.org/packages/2e/0c/b28ed414f080ee0ad153f848586d61d1878f91689950f037f976ce15f6c8/pandas-3.0.1.tar.gz", hash = "sha256:4186a699674af418f655dbd420ed87f50d56b4cd6603784279d9eef6627823c8", size = 4641901, upload-time = "2026-02-17T22:20:16.434Z" }
 wheels = [
@@ -5335,10 +5495,10 @@ default = [
     { name = "grpcio" },
     { name = "opencensus" },
     { name = "opentelemetry-exporter-prometheus", version = "0.54b1", source = { registry = "https://pypi.org/simple" }, marker = "extra == 'extra-13-megatron-core-dev' or extra == 'extra-13-megatron-core-lts'" },
-    { name = "opentelemetry-exporter-prometheus", version = "0.61b0", source = { registry = "https://pypi.org/simple" }, marker = "(extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (extra != 'extra-13-megatron-core-dev' and extra != 'extra-13-megatron-core-lts')" },
+    { name = "opentelemetry-exporter-prometheus", version = "0.60b1", source = { registry = "https://pypi.org/simple" }, marker = "(extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (extra != 'extra-13-megatron-core-dev' and extra != 'extra-13-megatron-core-lts')" },
     { name = "opentelemetry-proto" },
     { name = "opentelemetry-sdk", version = "1.33.1", source = { registry = "https://pypi.org/simple" }, marker = "extra == 'extra-13-megatron-core-dev' or extra == 'extra-13-megatron-core-lts'" },
-    { name = "opentelemetry-sdk", version = "1.40.0", source = { registry = "https://pypi.org/simple" }, marker = "(extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (extra != 'extra-13-megatron-core-dev' and extra != 'extra-13-megatron-core-lts')" },
+    { name = "opentelemetry-sdk", version = "1.39.1", source = { registry = "https://pypi.org/simple" }, marker = "(extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (extra != 'extra-13-megatron-core-dev' and extra != 'extra-13-megatron-core-lts')" },
     { name = "prometheus-client" },
     { name = "py-spy" },
     { name = "pydantic" },
@@ -5354,7 +5514,7 @@ source = { registry = "https://pypi.org/simple" }
 dependencies = [
     { name = "attrs" },
     { name = "rpds-py" },
-    { name = "typing-extensions", marker = "python_full_version < '3.13'" },
+    { name = "typing-extensions", marker = "python_full_version < '3.13' or (extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
 ]
 sdist = { url = "https://files.pythonhosted.org/packages/22/f5/df4e9027acead3ecc63e50fe1e36aca1523e1719559c499951bb4b53188f/referencing-0.37.0.tar.gz", hash = "sha256:44aefc3142c5b842538163acb373e24cce6632bd54bdb01b21ad5863489f50d8", size = 78036, upload-time = "2025-10-13T15:30:48.871Z" }
 wheels = [
@@ -6324,7 +6484,7 @@ version = "0.52.1"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
     { name = "anyio" },
-    { name = "typing-extensions", marker = "python_full_version < '3.13'" },
+    { name = "typing-extensions", marker = "python_full_version < '3.13' or (extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
 ]
 sdist = { url = "https://files.pythonhosted.org/packages/c4/68/79977123bb7be889ad680d79a40f339082c1978b5cfcf62c2d8d196873ac/starlette-0.52.1.tar.gz", hash = "sha256:834edd1b0a23167694292e94f597773bc3f89f362be6effee198165a35d62933", size = 2653702, upload-time = "2026-01-18T13:34:11.062Z" }
 wheels = [
@@ -6663,21 +6823,45 @@ name = "torch"
 version = "2.10.0"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
-    { name = "filelock", marker = "(python_full_version < '3.11' and sys_platform == 'emscripten') or (python_full_version < '3.11' and sys_platform == 'win32') or (sys_platform != 'emscripten' and sys_platform != 'linux' and sys_platform != 'win32')" },
-    { name = "fsspec", version = "2025.10.0", source = { registry = "https://pypi.org/simple" }, marker = "(python_full_version < '3.11' and sys_platform == 'emscripten') or (python_full_version < '3.11' and sys_platform == 'win32') or (sys_platform != 'emscripten' and sys_platform != 'linux' and sys_platform != 'win32')" },
-    { name = "jinja2", marker = "(python_full_version < '3.11' and sys_platform == 'emscripten') or (python_full_version < '3.11' and sys_platform == 'win32') or (sys_platform != 'emscripten' and sys_platform != 'linux' and sys_platform != 'win32')" },
-    { name = "networkx", version = "3.4.2", source = { registry = "https://pypi.org/simple" }, marker = "(python_full_version < '3.11' and sys_platform != 'linux') or (python_full_version >= '3.11' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (sys_platform == 'linux' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
-    { name = "networkx", version = "3.6.1", source = { registry = "https://pypi.org/simple" }, marker = "(python_full_version >= '3.11' and sys_platform != 'emscripten' and sys_platform != 'linux' and sys_platform != 'win32') or (python_full_version < '3.11' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (sys_platform == 'emscripten' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (sys_platform == 'linux' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (sys_platform == 'win32' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
-    { name = "setuptools", marker = "python_full_version >= '3.12' and sys_platform != 'emscripten' and sys_platform != 'linux' and sys_platform != 'win32'" },
-    { name = "sympy", marker = "(python_full_version < '3.11' and sys_platform == 'emscripten') or (python_full_version < '3.11' and sys_platform == 'win32') or (sys_platform != 'emscripten' and sys_platform != 'linux' and sys_platform != 'win32')" },
-    { name = "triton", marker = "sys_platform == 'never'" },
-    { name = "typing-extensions", marker = "(python_full_version < '3.11' and sys_platform == 'emscripten') or (python_full_version < '3.11' and sys_platform == 'win32') or (sys_platform != 'emscripten' and sys_platform != 'linux' and sys_platform != 'win32')" },
+    { name = "cuda-bindings", marker = "(platform_machine == 'x86_64' and sys_platform == 'linux') or (platform_machine != 'x86_64' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (sys_platform != 'linux' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
+    { name = "filelock" },
+    { name = "fsspec", version = "2025.10.0", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version < '3.14' or sys_platform != 'linux' or extra == 'extra-13-megatron-core-dev' or extra == 'extra-13-megatron-core-lts'" },
+    { name = "fsspec", version = "2026.2.0", source = { registry = "https://pypi.org/simple" }, marker = "(python_full_version >= '3.14' and sys_platform == 'linux' and extra != 'extra-13-megatron-core-dev' and extra != 'extra-13-megatron-core-lts') or (extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
+    { name = "jinja2" },
+    { name = "networkx", version = "3.4.2", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version < '3.11' or (extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
+    { name = "networkx", version = "3.6.1", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version >= '3.11' or (extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
+    { name = "nvidia-cublas-cu12", marker = "(platform_machine == 'x86_64' and sys_platform == 'linux') or (platform_machine != 'x86_64' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (sys_platform != 'linux' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
+    { name = "nvidia-cuda-cupti-cu12", marker = "(platform_machine == 'x86_64' and sys_platform == 'linux') or (platform_machine != 'x86_64' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (sys_platform != 'linux' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
+    { name = "nvidia-cuda-nvrtc-cu12", marker = "(platform_machine == 'x86_64' and sys_platform == 'linux') or (platform_machine != 'x86_64' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (sys_platform != 'linux' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
+    { name = "nvidia-cuda-runtime-cu12", marker = "(platform_machine == 'x86_64' and sys_platform == 'linux') or (platform_machine != 'x86_64' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (sys_platform != 'linux' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
+    { name = "nvidia-cudnn-cu12", marker = "(platform_machine == 'x86_64' and sys_platform == 'linux') or (platform_machine != 'x86_64' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (sys_platform != 'linux' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
+    { name = "nvidia-cufft-cu12", marker = "(platform_machine == 'x86_64' and sys_platform == 'linux') or (platform_machine != 'x86_64' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (sys_platform != 'linux' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
+    { name = "nvidia-cufile-cu12", marker = "(platform_machine == 'x86_64' and sys_platform == 'linux') or (platform_machine != 'x86_64' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (sys_platform != 'linux' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
+    { name = "nvidia-curand-cu12", marker = "(platform_machine == 'x86_64' and sys_platform == 'linux') or (platform_machine != 'x86_64' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (sys_platform != 'linux' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
+    { name = "nvidia-cusolver-cu12", marker = "(platform_machine == 'x86_64' and sys_platform == 'linux') or (platform_machine != 'x86_64' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (sys_platform != 'linux' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
+    { name = "nvidia-cusparse-cu12", marker = "(platform_machine == 'x86_64' and sys_platform == 'linux') or (platform_machine != 'x86_64' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (sys_platform != 'linux' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
+    { name = "nvidia-cusparselt-cu12", marker = "(platform_machine == 'x86_64' and sys_platform == 'linux') or (platform_machine != 'x86_64' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (sys_platform != 'linux' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
+    { name = "nvidia-nccl-cu12", marker = "(platform_machine == 'x86_64' and sys_platform == 'linux') or (platform_machine != 'x86_64' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (sys_platform != 'linux' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
+    { name = "nvidia-nvjitlink-cu12", marker = "(platform_machine == 'x86_64' and sys_platform == 'linux') or (platform_machine != 'x86_64' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (sys_platform != 'linux' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
+    { name = "nvidia-nvshmem-cu12", marker = "(platform_machine == 'x86_64' and sys_platform == 'linux') or (platform_machine != 'x86_64' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (sys_platform != 'linux' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
+    { name = "nvidia-nvtx-cu12", marker = "(platform_machine == 'x86_64' and sys_platform == 'linux') or (platform_machine != 'x86_64' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts') or (sys_platform != 'linux' and extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
+    { name = "setuptools", marker = "python_full_version >= '3.12' or (extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
+    { name = "sympy" },
+    { name = "triton", marker = "sys_platform == 'never' or (extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
+    { name = "typing-extensions" },
 ]
 wheels = [
     { url = "https://files.pythonhosted.org/packages/5b/30/bfebdd8ec77db9a79775121789992d6b3b75ee5494971294d7b4b7c999bc/torch-2.10.0-2-cp310-none-macosx_11_0_arm64.whl", hash = "sha256:2b980edd8d7c0a68c4e951ee1856334a43193f98730d97408fbd148c1a933313", size = 79411457, upload-time = "2026-02-10T21:44:59.189Z" },
     { url = "https://files.pythonhosted.org/packages/0f/8b/4b61d6e13f7108f36910df9ab4b58fd389cc2520d54d81b88660804aad99/torch-2.10.0-2-cp311-none-macosx_11_0_arm64.whl", hash = "sha256:418997cb02d0a0f1497cf6a09f63166f9f5df9f3e16c8a716ab76a72127c714f", size = 79423467, upload-time = "2026-02-10T21:44:48.711Z" },
     { url = "https://files.pythonhosted.org/packages/d3/54/a2ba279afcca44bbd320d4e73675b282fcee3d81400ea1b53934efca6462/torch-2.10.0-2-cp312-none-macosx_11_0_arm64.whl", hash = "sha256:13ec4add8c3faaed8d13e0574f5cd4a323c11655546f91fbe6afa77b57423574", size = 79498202, upload-time = "2026-02-10T21:44:52.603Z" },
     { url = "https://files.pythonhosted.org/packages/ec/23/2c9fe0c9c27f7f6cb865abcea8a4568f29f00acaeadfc6a37f6801f84cb4/torch-2.10.0-2-cp313-none-macosx_11_0_arm64.whl", hash = "sha256:e521c9f030a3774ed770a9c011751fb47c4d12029a3d6522116e48431f2ff89e", size = 79498254, upload-time = "2026-02-10T21:44:44.095Z" },
+    { url = "https://files.pythonhosted.org/packages/16/ee/efbd56687be60ef9af0c9c0ebe106964c07400eade5b0af8902a1d8cd58c/torch-2.10.0-3-cp310-cp310-manylinux_2_28_x86_64.whl", hash = "sha256:a1ff626b884f8c4e897c4c33782bdacdff842a165fee79817b1dd549fdda1321", size = 915510070, upload-time = "2026-03-11T14:16:39.386Z" },
+    { url = "https://files.pythonhosted.org/packages/36/ab/7b562f1808d3f65414cd80a4f7d4bb00979d9355616c034c171249e1a303/torch-2.10.0-3-cp311-cp311-manylinux_2_28_x86_64.whl", hash = "sha256:ac5bdcbb074384c66fa160c15b1ead77839e3fe7ed117d667249afce0acabfac", size = 915518691, upload-time = "2026-03-11T14:15:43.147Z" },
+    { url = "https://files.pythonhosted.org/packages/b3/7a/abada41517ce0011775f0f4eacc79659bc9bc6c361e6bfe6f7052a6b9363/torch-2.10.0-3-cp312-cp312-manylinux_2_28_x86_64.whl", hash = "sha256:98c01b8bb5e3240426dcde1446eed6f40c778091c8544767ef1168fc663a05a6", size = 915622781, upload-time = "2026-03-11T14:17:11.354Z" },
+    { url = "https://files.pythonhosted.org/packages/ab/c6/4dfe238342ffdcec5aef1c96c457548762d33c40b45a1ab7033bb26d2ff2/torch-2.10.0-3-cp313-cp313-manylinux_2_28_x86_64.whl", hash = "sha256:80b1b5bfe38eb0e9f5ff09f206dcac0a87aadd084230d4a36eea5ec5232c115b", size = 915627275, upload-time = "2026-03-11T14:16:11.325Z" },
+    { url = "https://files.pythonhosted.org/packages/d8/f0/72bf18847f58f877a6a8acf60614b14935e2f156d942483af1ffc081aea0/torch-2.10.0-3-cp313-cp313t-manylinux_2_28_x86_64.whl", hash = "sha256:46b3574d93a2a8134b3f5475cfb98e2eb46771794c57015f6ad1fb795ec25e49", size = 915523474, upload-time = "2026-03-11T14:17:44.422Z" },
+    { url = "https://files.pythonhosted.org/packages/f4/39/590742415c3030551944edc2ddc273ea1fdfe8ffb2780992e824f1ebee98/torch-2.10.0-3-cp314-cp314-manylinux_2_28_x86_64.whl", hash = "sha256:b1d5e2aba4eb7f8e87fbe04f86442887f9167a35f092afe4c237dfcaaef6e328", size = 915632474, upload-time = "2026-03-11T14:15:13.666Z" },
+    { url = "https://files.pythonhosted.org/packages/b6/8e/34949484f764dde5b222b7fe3fede43e4a6f0da9d7f8c370bb617d629ee2/torch-2.10.0-3-cp314-cp314t-manylinux_2_28_x86_64.whl", hash = "sha256:0228d20b06701c05a8f978357f657817a4a63984b0c90745def81c18aedfa591", size = 915523882, upload-time = "2026-03-11T14:14:46.311Z" },
     { url = "https://files.pythonhosted.org/packages/0c/1a/c61f36cfd446170ec27b3a4984f072fd06dab6b5d7ce27e11adb35d6c838/torch-2.10.0-cp310-cp310-manylinux_2_28_aarch64.whl", hash = "sha256:5276fa790a666ee8becaffff8acb711922252521b28fbce5db7db5cf9cb2026d", size = 145992962, upload-time = "2026-01-21T16:24:14.04Z" },
     { url = "https://files.pythonhosted.org/packages/b5/60/6662535354191e2d1555296045b63e4279e5a9dbad49acf55a5d38655a39/torch-2.10.0-cp310-cp310-manylinux_2_28_x86_64.whl", hash = "sha256:aaf663927bcd490ae971469a624c322202a2a1e68936eb952535ca4cd3b90444", size = 915599237, upload-time = "2026-01-21T16:23:25.497Z" },
     { url = "https://files.pythonhosted.org/packages/40/b8/66bbe96f0d79be2b5c697b2e0b187ed792a15c6c4b8904613454651db848/torch-2.10.0-cp310-cp310-win_amd64.whl", hash = "sha256:a4be6a2a190b32ff5c8002a0977a25ea60e64f7ba46b1be37093c141d9c49aeb", size = 113720931, upload-time = "2026-01-21T16:24:23.743Z" },
@@ -6734,7 +6918,7 @@ name = "tqdm"
 version = "4.67.3"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
-    { name = "colorama", marker = "sys_platform == 'win32'" },
+    { name = "colorama", marker = "sys_platform == 'win32' or (extra == 'extra-13-megatron-core-dev' and extra == 'extra-13-megatron-core-lts')" },
 ]
 sdist = { url = "https://files.pythonhosted.org/packages/09/a9/6ba95a270c6f1fbcd8dac228323f2777d886cb206987444e4bce66338dd4/tqdm-4.67.3.tar.gz", hash = "sha256:7d825f03f89244ef73f1d4ce193cb1774a8179fd96f31d7e1dcde62092b960bb", size = 169598, upload-time = "2026-02-03T17:35:53.048Z" }
 wheels = [