[Dev] Fix Qwen3-Next hang on Blackwell, add a flag to control torch.compile

megatron/core/jit.py

@@ -7,12 +7,27 @@
 jit_fuser = torch.jit.script
 # nvFuser is deprecated in PyTorch JIT starting from 2.2
 
-try:
-    if is_torch_min_version("2.2.0a0"):
-        jit_fuser = torch.compile
-except ImportError:
 
-    def noop_decorator(func):
-        return func
+def noop_decorator(func):
+    '''No-op decorator'''
+    return func
 
+
+def enable_jit_fuser():
+    '''Enable the JIT fuser'''
+    global jit_fuser
+    try:
+        if is_torch_min_version("2.2.0a0"):
+            jit_fuser = torch.compile
+    except ImportError:
+
+        jit_fuser = noop_decorator
+
+
+def disable_jit_fuser():
+    '''Disable the JIT fuser'''
+    global jit_fuser
     jit_fuser = noop_decorator
+
+
+enable_jit_fuser()

megatron/core/ssm/gated_delta_net.py

@@ -18,6 +18,7 @@
 from megatron.core.dist_checkpointing.mapping import ReplicaId, ShardedTensorFactory
 from megatron.core.fp8_utils import get_fp8_align_size
 from megatron.core.inference.contexts import BaseInferenceContext
+from megatron.core.jit import jit_fuser
 from megatron.core.packed_seq_params import PackedSeqParams
 from megatron.core.process_groups_config import ProcessGroupCollection
 from megatron.core.tensor_parallel import get_cuda_rng_tracker
@@ -384,7 +385,7 @@ def forward(
 
         # RMSNorm
         nvtx_range_push(suffix="gated_norm")
-        norm_out = self._torch_compiled_gated_norm(core_attn_out, gate)
+        norm_out = self._apply_gated_norm(core_attn_out, gate)
         nvtx_range_pop(suffix="gated_norm")
 
         # Transpose: b s x --> s b x
@@ -399,8 +400,8 @@ def forward(
 
         return out, out_bias
 
-    @torch.compile
-    def _torch_compiled_gated_norm(self, x, gate):
+    @jit_fuser
+    def _apply_gated_norm(self, x, gate):
         # Output Norm
         x_dtype = x.dtype
         x = x.reshape(-1, x.shape[-1])

megatron/core/transformer/attention.py

@@ -9,6 +9,7 @@
 
 from megatron.core import tensor_parallel
 from megatron.core.inference.contexts import BaseInferenceContext
+from megatron.core.jit import jit_fuser
 from megatron.core.models.common.embeddings.rope_utils import (
     apply_rotary_pos_emb,
     apply_rotary_pos_emb_with_cos_sin,
@@ -958,7 +959,7 @@ def forward(
         # Output gate
         if gate is not None:
             nvtx_range_push(suffix="output_gate")
-            core_attn_out = self._torch_compiled_output_gate(core_attn_out, gate)
+            core_attn_out = self._apply_output_gate(core_attn_out, gate)
             nvtx_range_pop(suffix="output_gate")
 
         # =================
@@ -978,8 +979,8 @@ def forward(
 
         return output, bias
 
-    @torch.compile
-    def _torch_compiled_output_gate(self, x, gate):
+    @jit_fuser
+    def _apply_output_gate(self, x, gate):
         x_dtype = x.dtype
         gate = gate.contiguous()
         gate = gate.view(*x.shape)

megatron/core/transformer/transformer_block.py

@@ -801,6 +801,12 @@ def sharded_state_dict(
         elif isinstance(self.config.moe_layer_freq, list):
             non_homogeneous_layers = True
 
+        if isinstance(self.config.linear_attention_freq, int):
+            if self.config.linear_attention_freq > 1:
+                non_homogeneous_layers = True
+        elif isinstance(self.config.linear_attention_freq, list):
+            non_homogeneous_layers = True
+
         if self.config.heterogeneous_block_specs:
             non_homogeneous_layers = True
 

megatron/training/arguments.py

@@ -2349,6 +2349,8 @@ def _add_training_args(parser):
                        help='The submodules to offload its input. Choices: "attn_norm", "core_attn", "attn_proj", "mlp_norm", "expert_fc1", "moe_act".')
     group.add_argument('--min-offloaded-tensor-size', type=int, default=1024*1024,
                        help='The minimum size of the tensor to be offloaded.')
+    group.add_argument('--disable-jit-fuser', action='store_true',
+                       help='Disable the JIT fuser.')
     return parser
 
 

megatron/training/global_vars.py

@@ -9,6 +9,7 @@
 from megatron.core import Timers
 from megatron.core.config import set_experimental_flag
 from megatron.core.energy_monitor import EnergyMonitor
+from megatron.core.jit import disable_jit_fuser
 from megatron.core.num_microbatches_calculator import init_num_microbatches_calculator, unset_num_microbatches_calculator
 from megatron.training import dist_signal_handler
 from megatron.training.tokenizer import build_tokenizer
@@ -111,6 +112,9 @@ def set_global_variables(args, build_tokenizer=True):
     if args.exit_signal_handler:
         _set_signal_handler()
 
+    if args.disable_jit_fuser:
+        disable_jit_fuser()
+
 
 def unset_global_variables():
     """Unset global vars.