MCPcopy
hub / github.com/deepspeedai/DeepSpeed / set_optimizer_flags

Function set_optimizer_flags

deepspeed/__init__.py:71–77  ·  view source on GitHub ↗
(config_class, model)

Source from the content-addressed store, hash-verified

69
70
71def set_optimizer_flags(config_class, model):
72 if config_class.optimizer_name == MUON_OPTIMIZER:
73 for name, p in model.named_parameters():
74 if p.ndim >= 2 and not any(keyword in name.lower() for keyword in ("embed", "lm_head")):
75 setattr(p, "use_muon", True)
76 else:
77 setattr(p, "use_muon", False)
78
79
80def initialize(args=None,

Callers 2

initializeFunction · 0.85

Calls

no outgoing calls

Tested by

no test coverage detected

Used in the wild real call sites across dependent graphs

searching dependent graphs…