Fix parametrization-based weight norm (#33275)

* refactor weight_norm + propose uniformed solution to reconcile meta load_state_dict with classic loading

* make style

* fix sew

* fix sew and sew_d tests

src/transformers/modeling_utils.py

@@ -818,7 +818,6 @@ def _move_model_to_meta(model, loaded_state_dict_keys, start_prefix):
 def _load_state_dict_into_meta_model(
     model,
     state_dict,
-    loaded_state_dict_keys,  # left for now but could be removed, see below
     start_prefix,
     expected_keys,
     device_map=None,
@@ -847,8 +846,6 @@ def _load_state_dict_into_meta_model(
     # - deepspeed zero 3 support
     # - need to copy metadata if any - see _load_state_dict_into_model
     # - handling error_msgs - mimicking the error handling in module._load_from_state_dict()
-    # - Is there a situation where some keys aren't in `loaded_state_dict_keys` and in which case
-    #   they won't get loaded.
 
     error_msgs = []
 
@@ -868,6 +865,18 @@ def _load_state_dict_into_meta_model(
             # We add only the first key as an example
             new_key = key.replace("beta", "bias")
             renamed_beta[key] = new_key if not renamed_beta else renamed_beta
+
+        # To reproduce `_load_state_dict_into_model` behaviour, we need to manually rename parametrized weigth norm, if necessary.
+        if hasattr(nn.utils.parametrizations, "weight_norm"):
+            if "weight_g" in key:
+                new_key = key.replace("weight_g", "parametrizations.weight.original0")
+            if "weight_v" in key:
+                new_key = key.replace("weight_v", "parametrizations.weight.original1")
+        else:
+            if "parametrizations.weight.original0" in key:
+                new_key = key.replace("parametrizations.weight.original0", "weight_g")
+            if "parametrizations.weight.original1" in key:
+                new_key = key.replace("parametrizations.weight.original1", "weight_v")
         if new_key:
             old_keys.append(key)
             new_keys.append(new_key)
@@ -884,8 +893,7 @@ def _load_state_dict_into_meta_model(
     is_torch_e4m3fn_available = hasattr(torch, "float8_e4m3fn")
 
     for param_name, param in state_dict.items():
-        # First part of the test is always true as load_state_dict_keys always contains state_dict keys.
+        if param_name not in expected_keys:
-        if param_name not in loaded_state_dict_keys or param_name not in expected_keys:
             continue
 
         if param_name.startswith(start_prefix):
@@ -4132,6 +4140,18 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
                 return key.replace("beta", "bias")
             if "gamma" in key:
                 return key.replace("gamma", "weight")
+
+            # to avoid logging parametrized weight norm renaming
+            if hasattr(nn.utils.parametrizations, "weight_norm"):
+                if "weight_g" in key:
+                    return key.replace("weight_g", "parametrizations.weight.original0")
+                if "weight_v" in key:
+                    return key.replace("weight_v", "parametrizations.weight.original1")
+            else:
+                if "parametrizations.weight.original0" in key:
+                    return key.replace("parametrizations.weight.original0", "weight_g")
+                if "parametrizations.weight.original1" in key:
+                    return key.replace("parametrizations.weight.original1", "weight_v")
             return key
 
         original_loaded_keys = loaded_keys
@@ -4376,7 +4396,6 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
                 error_msgs, offload_index, state_dict_index = _load_state_dict_into_meta_model(
                     model_to_load,
                     state_dict,
-                    loaded_keys,
                     start_prefix,
                     expected_keys,
                     device_map=device_map,
@@ -4453,7 +4472,6 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
                         new_error_msgs, offload_index, state_dict_index = _load_state_dict_into_meta_model(
                             model_to_load,
                             state_dict,
-                            loaded_keys,
                             start_prefix,
                             expected_keys,
                             device_map=device_map,
@@ -4609,7 +4627,6 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
         error_msgs = _load_state_dict_into_meta_model(
             model,
             state_dict,
-            loaded_state_dict_keys,
             start_prefix,
             expected_keys=expected_keys,
             hf_quantizer=hf_quantizer,

src/transformers/models/dac/modeling_dac.py

@@ -494,33 +494,37 @@ class DacPreTrainedModel(PreTrainedModel):
             nn.init.constant_(module.bias, 0)
 
     def apply_weight_norm(self):
-        for layer in self.quantizer.quantizers:
+        weight_norm = nn.utils.weight_norm
-            nn.utils.weight_norm(layer.in_proj)
+        if hasattr(nn.utils.parametrizations, "weight_norm"):
-            nn.utils.weight_norm(layer.out_proj)
+            weight_norm = nn.utils.parametrizations.weight_norm
 
-        nn.utils.weight_norm(self.encoder.conv1)
+        for layer in self.quantizer.quantizers:
-        nn.utils.weight_norm(self.encoder.conv2)
+            weight_norm(layer.in_proj)
+            weight_norm(layer.out_proj)
+
+        weight_norm(self.encoder.conv1)
+        weight_norm(self.encoder.conv2)
 
         for layer in self.encoder.block:
-            nn.utils.weight_norm(layer.conv1)
+            weight_norm(layer.conv1)
-            nn.utils.weight_norm(layer.res_unit1.conv1)
+            weight_norm(layer.res_unit1.conv1)
-            nn.utils.weight_norm(layer.res_unit1.conv2)
+            weight_norm(layer.res_unit1.conv2)
-            nn.utils.weight_norm(layer.res_unit2.conv1)
+            weight_norm(layer.res_unit2.conv1)
-            nn.utils.weight_norm(layer.res_unit2.conv2)
+            weight_norm(layer.res_unit2.conv2)
-            nn.utils.weight_norm(layer.res_unit3.conv1)
+            weight_norm(layer.res_unit3.conv1)
-            nn.utils.weight_norm(layer.res_unit3.conv2)
+            weight_norm(layer.res_unit3.conv2)
 
-        nn.utils.weight_norm(self.decoder.conv1)
+        weight_norm(self.decoder.conv1)
-        nn.utils.weight_norm(self.decoder.conv2)
+        weight_norm(self.decoder.conv2)
 
         for layer in self.decoder.block:
-            nn.utils.weight_norm(layer.conv_t1)
+            weight_norm(layer.conv_t1)
-            nn.utils.weight_norm(layer.res_unit1.conv1)
+            weight_norm(layer.res_unit1.conv1)
-            nn.utils.weight_norm(layer.res_unit1.conv2)
+            weight_norm(layer.res_unit1.conv2)
-            nn.utils.weight_norm(layer.res_unit2.conv1)
+            weight_norm(layer.res_unit2.conv1)
-            nn.utils.weight_norm(layer.res_unit2.conv2)
+            weight_norm(layer.res_unit2.conv2)
-            nn.utils.weight_norm(layer.res_unit3.conv1)
+            weight_norm(layer.res_unit3.conv1)
-            nn.utils.weight_norm(layer.res_unit3.conv2)
+            weight_norm(layer.res_unit3.conv2)
 
     def remove_weight_norm(self):
         for layer in self.quantizer.quantizers:

src/transformers/models/encodec/modeling_encodec.py

@@ -103,8 +103,12 @@ class EncodecConv1d(nn.Module):
             )
 
         self.conv = nn.Conv1d(in_channels, out_channels, kernel_size, stride, dilation=dilation)
+        weight_norm = nn.utils.weight_norm
+        if hasattr(nn.utils.parametrizations, "weight_norm"):
+            weight_norm = nn.utils.parametrizations.weight_norm
+
         if self.norm_type == "weight_norm":
-            self.conv = nn.utils.weight_norm(self.conv)
+            self.conv = weight_norm(self.conv)
         elif self.norm_type == "time_group_norm":
             self.norm = nn.GroupNorm(1, out_channels)
 
@@ -186,8 +190,13 @@ class EncodecConvTranspose1d(nn.Module):
             )
 
         self.conv = nn.ConvTranspose1d(in_channels, out_channels, kernel_size, stride)
+
+        weight_norm = nn.utils.weight_norm
+        if hasattr(nn.utils.parametrizations, "weight_norm"):
+            weight_norm = nn.utils.parametrizations.weight_norm
+
         if config.norm_type == "weight_norm":
-            self.conv = nn.utils.weight_norm(self.conv)
+            self.conv = weight_norm(self.conv)
         elif config.norm_type == "time_group_norm":
             self.norm = nn.GroupNorm(1, out_channels)
 

src/transformers/models/fastspeech2_conformer/modeling_fastspeech2_conformer.py

@@ -1416,10 +1416,14 @@ class HifiGanResidualBlock(nn.Module):
         return (kernel_size * dilation - dilation) // 2
 
     def apply_weight_norm(self):
+        weight_norm = nn.utils.weight_norm
+        if hasattr(nn.utils.parametrizations, "weight_norm"):
+            weight_norm = nn.utils.parametrizations.weight_norm
+
         for layer in self.convs1:
-            nn.utils.weight_norm(layer)
+            weight_norm(layer)
         for layer in self.convs2:
-            nn.utils.weight_norm(layer)
+            weight_norm(layer)
 
     def remove_weight_norm(self):
         for layer in self.convs1:
@@ -1493,12 +1497,16 @@ class FastSpeech2ConformerHifiGan(PreTrainedModel):
                 module.bias.data.zero_()
 
     def apply_weight_norm(self):
-        nn.utils.weight_norm(self.conv_pre)
+        weight_norm = nn.utils.weight_norm
+        if hasattr(nn.utils.parametrizations, "weight_norm"):
+            weight_norm = nn.utils.parametrizations.weight_norm
+
+        weight_norm(self.conv_pre)
         for layer in self.upsampler:
-            nn.utils.weight_norm(layer)
+            weight_norm(layer)
         for layer in self.resblocks:
             layer.apply_weight_norm()
-        nn.utils.weight_norm(self.conv_post)
+        weight_norm(self.conv_post)
 
     def remove_weight_norm(self):
         nn.utils.remove_weight_norm(self.conv_pre)

src/transformers/models/seamless_m4t/modeling_seamless_m4t.py

@@ -2361,10 +2361,14 @@ class HifiGanResidualBlock(nn.Module):
         return (kernel_size * dilation - dilation) // 2
 
     def apply_weight_norm(self):
+        weight_norm = nn.utils.weight_norm
+        if hasattr(nn.utils.parametrizations, "weight_norm"):
+            weight_norm = nn.utils.parametrizations.weight_norm
+
         for layer in self.convs1:
-            nn.utils.weight_norm(layer)
+            weight_norm(layer)
         for layer in self.convs2:
-            nn.utils.weight_norm(layer)
+            weight_norm(layer)
 
     def remove_weight_norm(self):
         for layer in self.convs1:
@@ -2633,12 +2637,16 @@ class SeamlessM4TCodeHifiGan(PreTrainedModel):
                 module.weight.data[module.padding_idx].zero_()
 
     def apply_weight_norm(self):
-        nn.utils.weight_norm(self.hifi_gan.conv_pre)
+        weight_norm = nn.utils.weight_norm
+        if hasattr(nn.utils.parametrizations, "weight_norm"):
+            weight_norm = nn.utils.parametrizations.weight_norm
+
+        weight_norm(self.hifi_gan.conv_pre)
         for layer in self.hifi_gan.upsampler:
-            nn.utils.weight_norm(layer)
+            weight_norm(layer)
         for layer in self.hifi_gan.resblocks:
             layer.apply_weight_norm()
-        nn.utils.weight_norm(self.hifi_gan.conv_post)
+        weight_norm(self.hifi_gan.conv_post)
 
     def remove_weight_norm(self):
         nn.utils.remove_weight_norm(self.hifi_gan.conv_pre)

src/transformers/models/seamless_m4t_v2/modeling_seamless_m4t_v2.py

@@ -2608,10 +2608,14 @@ class HifiGanResidualBlock(nn.Module):
         return (kernel_size * dilation - dilation) // 2
 
     def apply_weight_norm(self):
+        weight_norm = nn.utils.weight_norm
+        if hasattr(nn.utils.parametrizations, "weight_norm"):
+            weight_norm = nn.utils.parametrizations.weight_norm
+
         for layer in self.convs1:
-            nn.utils.weight_norm(layer)
+            weight_norm(layer)
         for layer in self.convs2:
-            nn.utils.weight_norm(layer)
+            weight_norm(layer)
 
     def remove_weight_norm(self):
         for layer in self.convs1:
@@ -2889,12 +2893,16 @@ class SeamlessM4Tv2CodeHifiGan(PreTrainedModel):
 
     # Copied from transformers.models.seamless_m4t.modeling_seamless_m4t.SeamlessM4TCodeHifiGan.apply_weight_norm
     def apply_weight_norm(self):
-        nn.utils.weight_norm(self.hifi_gan.conv_pre)
+        weight_norm = nn.utils.weight_norm
+        if hasattr(nn.utils.parametrizations, "weight_norm"):
+            weight_norm = nn.utils.parametrizations.weight_norm
+
+        weight_norm(self.hifi_gan.conv_pre)
         for layer in self.hifi_gan.upsampler:
-            nn.utils.weight_norm(layer)
+            weight_norm(layer)
         for layer in self.hifi_gan.resblocks:
             layer.apply_weight_norm()
-        nn.utils.weight_norm(self.hifi_gan.conv_post)
+        weight_norm(self.hifi_gan.conv_post)
 
     # Copied from transformers.models.seamless_m4t.modeling_seamless_m4t.SeamlessM4TCodeHifiGan.remove_weight_norm
     def remove_weight_norm(self):

src/transformers/models/sew/modeling_sew.py

@@ -274,11 +274,15 @@ class SEWPositionalConvEmbedding(nn.Module):
             stride=config.squeeze_factor,
         )
 
+        weight_norm = nn.utils.weight_norm
+        if hasattr(nn.utils.parametrizations, "weight_norm"):
+            weight_norm = nn.utils.parametrizations.weight_norm
+
         if is_deepspeed_zero3_enabled():
             import deepspeed
 
             with deepspeed.zero.GatheredParameters(self.conv.weight, modifier_rank=0):
-                self.conv = nn.utils.weight_norm(self.conv, name="weight", dim=2)
+                self.conv = weight_norm(self.conv, name="weight", dim=2)
             if hasattr(self.conv, "parametrizations"):
                 weight_g = self.conv.parametrizations.weight.original0
                 weight_v = self.conv.parametrizations.weight.original1
@@ -288,7 +292,7 @@ class SEWPositionalConvEmbedding(nn.Module):
             deepspeed.zero.register_external_parameter(self, weight_v)
             deepspeed.zero.register_external_parameter(self, weight_g)
         else:
-            self.conv = nn.utils.weight_norm(self.conv, name="weight", dim=2)
+            self.conv = weight_norm(self.conv, name="weight", dim=2)
 
         self.padding = SEWSamePadLayer(config.num_conv_pos_embeddings)
         self.activation = ACT2FN[config.feat_extract_activation]

src/transformers/models/sew_d/modeling_sew_d.py

@@ -349,11 +349,15 @@ class SEWDPositionalConvEmbedding(nn.Module):
             stride=config.squeeze_factor,
         )
 
+        weight_norm = nn.utils.weight_norm
+        if hasattr(nn.utils.parametrizations, "weight_norm"):
+            weight_norm = nn.utils.parametrizations.weight_norm
+
         if is_deepspeed_zero3_enabled():
             import deepspeed
 
             with deepspeed.zero.GatheredParameters(self.conv.weight, modifier_rank=0):
-                self.conv = nn.utils.weight_norm(self.conv, name="weight", dim=2)
+                self.conv = weight_norm(self.conv, name="weight", dim=2)
             if hasattr(self.conv, "parametrizations"):
                 weight_g = self.conv.parametrizations.weight.original0
                 weight_v = self.conv.parametrizations.weight.original1
@@ -363,7 +367,7 @@ class SEWDPositionalConvEmbedding(nn.Module):
             deepspeed.zero.register_external_parameter(self, weight_v)
             deepspeed.zero.register_external_parameter(self, weight_g)
         else:
-            self.conv = nn.utils.weight_norm(self.conv, name="weight", dim=2)
+            self.conv = weight_norm(self.conv, name="weight", dim=2)
 
         self.padding = SEWDSamePadLayer(config.num_conv_pos_embeddings)
         self.activation = ACT2FN[config.feat_extract_activation]

src/transformers/models/speecht5/modeling_speecht5.py

@@ -3234,10 +3234,14 @@ class HifiGanResidualBlock(nn.Module):
         return (kernel_size * dilation - dilation) // 2
 
     def apply_weight_norm(self):
+        weight_norm = nn.utils.weight_norm
+        if hasattr(nn.utils.parametrizations, "weight_norm"):
+            weight_norm = nn.utils.parametrizations.weight_norm
+
         for layer in self.convs1:
-            nn.utils.weight_norm(layer)
+            weight_norm(layer)
         for layer in self.convs2:
-            nn.utils.weight_norm(layer)
+            weight_norm(layer)
 
     def remove_weight_norm(self):
         for layer in self.convs1:
@@ -3310,12 +3314,16 @@ class SpeechT5HifiGan(PreTrainedModel):
                 module.bias.data.zero_()
 
     def apply_weight_norm(self):
-        nn.utils.weight_norm(self.conv_pre)
+        weight_norm = nn.utils.weight_norm
+        if hasattr(nn.utils.parametrizations, "weight_norm"):
+            weight_norm = nn.utils.parametrizations.weight_norm
+
+        weight_norm(self.conv_pre)
         for layer in self.upsampler:
-            nn.utils.weight_norm(layer)
+            weight_norm(layer)
         for layer in self.resblocks:
             layer.apply_weight_norm()
-        nn.utils.weight_norm(self.conv_post)
+        weight_norm(self.conv_post)
 
     def remove_weight_norm(self):
         nn.utils.remove_weight_norm(self.conv_pre)

src/transformers/models/univnet/modeling_univnet.py

@@ -87,8 +87,12 @@ class UnivNetKernelPredictorResidualBlock(nn.Module):
         return hidden_states + residual
 
     def apply_weight_norm(self):
-        nn.utils.weight_norm(self.conv1)
+        weight_norm = nn.utils.weight_norm
-        nn.utils.weight_norm(self.conv2)
+        if hasattr(nn.utils.parametrizations, "weight_norm"):
+            weight_norm = nn.utils.parametrizations.weight_norm
+
+        weight_norm(self.conv1)
+        weight_norm(self.conv2)
 
     def remove_weight_norm(self):
         nn.utils.remove_weight_norm(self.conv1)
@@ -197,11 +201,15 @@ class UnivNetKernelPredictor(nn.Module):
         return kernels, biases
 
     def apply_weight_norm(self):
-        nn.utils.weight_norm(self.input_conv)
+        weight_norm = nn.utils.weight_norm
+        if hasattr(nn.utils.parametrizations, "weight_norm"):
+            weight_norm = nn.utils.parametrizations.weight_norm
+
+        weight_norm(self.input_conv)
         for layer in self.resblocks:
             layer.apply_weight_norm()
-        nn.utils.weight_norm(self.kernel_conv)
+        weight_norm(self.kernel_conv)
-        nn.utils.weight_norm(self.bias_conv)
+        weight_norm(self.bias_conv)
 
     def remove_weight_norm(self):
         nn.utils.remove_weight_norm(self.input_conv)
@@ -328,7 +336,11 @@ class UnivNetLvcResidualBlock(nn.Module):
         return output_hidden_states
 
     def apply_weight_norm(self):
-        nn.utils.weight_norm(self.conv)
+        weight_norm = nn.utils.weight_norm
+        if hasattr(nn.utils.parametrizations, "weight_norm"):
+            weight_norm = nn.utils.parametrizations.weight_norm
+
+        weight_norm(self.conv)
 
     def remove_weight_norm(self):
         nn.utils.remove_weight_norm(self.conv)
@@ -398,7 +410,11 @@ class UnivNetLvcBlock(nn.Module):
         return hidden_states
 
     def apply_weight_norm(self):
-        nn.utils.weight_norm(self.convt_pre)
+        weight_norm = nn.utils.weight_norm
+        if hasattr(nn.utils.parametrizations, "weight_norm"):
+            weight_norm = nn.utils.parametrizations.weight_norm
+
+        weight_norm(self.convt_pre)
         self.kernel_predictor.apply_weight_norm()
         for layer in self.resblocks:
             layer.apply_weight_norm()
@@ -619,10 +635,14 @@ class UnivNetModel(PreTrainedModel):
                 module.bias.data.zero_()
 
     def apply_weight_norm(self):
-        nn.utils.weight_norm(self.conv_pre)
+        weight_norm = nn.utils.weight_norm
+        if hasattr(nn.utils.parametrizations, "weight_norm"):
+            weight_norm = nn.utils.parametrizations.weight_norm
+
+        weight_norm(self.conv_pre)
         for layer in self.resblocks:
             layer.apply_weight_norm()
-        nn.utils.weight_norm(self.conv_post)
+        weight_norm(self.conv_post)
 
     def remove_weight_norm(self):
         nn.utils.remove_weight_norm(self.conv_pre)

src/transformers/models/vits/modeling_vits.py

@@ -461,10 +461,14 @@ class HifiGanResidualBlock(nn.Module):
         return (kernel_size * dilation - dilation) // 2
 
     def apply_weight_norm(self):
+        weight_norm = nn.utils.weight_norm
+        if hasattr(nn.utils.parametrizations, "weight_norm"):
+            weight_norm = nn.utils.parametrizations.weight_norm
+
         for layer in self.convs1:
-            nn.utils.weight_norm(layer)
+            weight_norm(layer)
         for layer in self.convs2:
-            nn.utils.weight_norm(layer)
+            weight_norm(layer)
 
     def remove_weight_norm(self):
         for layer in self.convs1:
@@ -521,8 +525,12 @@ class VitsHifiGan(nn.Module):
             self.cond = nn.Conv1d(config.speaker_embedding_size, config.upsample_initial_channel, 1)
 
     def apply_weight_norm(self):
+        weight_norm = nn.utils.weight_norm
+        if hasattr(nn.utils.parametrizations, "weight_norm"):
+            weight_norm = nn.utils.parametrizations.weight_norm
+
         for layer in self.upsampler:
-            nn.utils.weight_norm(layer)
+            weight_norm(layer)
         for layer in self.resblocks:
             layer.apply_weight_norm()
 

tests/models/sew/test_modeling_sew.py

@@ -420,6 +420,7 @@ class SEWModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase):
             model = model_class(config=configs_no_init)
             for name, param in model.named_parameters():
                 uniform_init_parms = [
+                    "conv.parametrizations.weight",
                     "conv.weight",
                     "masked_spec_embed",
                     "quantizer.weight_proj.weight",

tests/models/sew_d/test_modeling_sew_d.py

@@ -422,6 +422,7 @@ class SEWDModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase):
             model = model_class(config=configs_no_init)
             for name, param in model.named_parameters():
                 uniform_init_parms = [
+                    "conv.parametrizations.weight",
                     "conv.weight",
                     "masked_spec_embed",
                     "quantizer.weight_proj.weight",