@@ -270,9 +270,9 @@ def create_optimizer(
270270
271271 if param_group_fn :
272272 # run custom fn to generate param groups from nn.Module
273- parameters = param_group_fn (model_or_params )
273+ params = param_group_fn (model_or_params )
274274 elif layer_decay is not None :
275- parameters = param_groups_layer_decay (
275+ params = param_groups_layer_decay (
276276 model_or_params ,
277277 weight_decay = weight_decay ,
278278 layer_decay = layer_decay ,
@@ -281,17 +281,17 @@ def create_optimizer(
281281 )
282282 weight_decay = 0.
283283 elif weight_decay and weight_decay_exclude_1d :
284- parameters = param_groups_weight_decay (
284+ params = param_groups_weight_decay (
285285 model_or_params ,
286286 weight_decay = weight_decay ,
287287 no_weight_decay_list = no_weight_decay ,
288288 )
289289 weight_decay = 0.
290290 else :
291- parameters = model_or_params .parameters ()
291+ params = model_or_params .parameters ()
292292 else :
293293 # pass parameters / parameter groups through to optimizer
294- parameters = model_or_params
294+ params = model_or_params
295295
296296 # Parse optimizer name
297297 opt_split = opt .lower ().split ('_' )
@@ -330,7 +330,7 @@ def create_optimizer(
330330
331331 # Create optimizer
332332 opt_class = self .get_optimizer_class (opt_info , bind_defaults = False )
333- optimizer = opt_class (parameters , ** opt_args )
333+ optimizer = opt_class (params , ** opt_args )
334334
335335 # Apply Lookahead if requested
336336 if use_lookahead :
@@ -685,12 +685,14 @@ def _register_bnb_optimizers(registry: OptimizerRegistry) -> None:
685685 'bnblion' ,
686686 'bitsandbytes.optim.Lion' ,
687687 description = 'bitsandbytes Lion' ,
688+ has_eps = False ,
688689 has_betas = True
689690 ),
690691 OptimInfo (
691692 'bnblion8bit' ,
692693 'bitsandbytes.optim.Lion8bit' ,
693694 description = 'bitsandbytes 8-bit Lion with dynamic quantization' ,
695+ has_eps = False ,
694696 has_betas = True
695697 ),
696698 OptimInfo (
0 commit comments