v4.55.4

The cherry-picked commit does not match the changes nor the PR
This reverts commit e75d67ec39.

examples/pytorch/audio-classification/run_audio_classification.py

@@ -15,7 +15,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "datasets[audio]>=1.14.0",
 #     "evaluate",
 #     "librosa",

examples/pytorch/contrastive-image-text/run_clip.py

@@ -15,7 +15,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "torch>=1.5.0",
 #     "torchvision>=0.6.0",
 #     "datasets>=1.8.0",

examples/pytorch/image-classification/run_image_classification.py

@@ -14,7 +14,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "accelerate>=0.12.0",
 #     "torch>=1.5.0",
 #     "torchvision>=0.6.0",

examples/pytorch/image-classification/run_image_classification_no_trainer.py

@@ -14,7 +14,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "accelerate>=0.12.0",
 #     "torch>=1.5.0",
 #     "torchvision>=0.6.0",

examples/pytorch/image-pretraining/run_mae.py

@@ -14,7 +14,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "torch>=1.5.0",
 #     "torchvision>=0.6.0",
 #     "datasets>=1.8.0",

examples/pytorch/image-pretraining/run_mim.py

@@ -14,7 +14,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "torch>=1.5.0",
 #     "torchvision>=0.6.0",
 #     "datasets>=1.8.0",

examples/pytorch/image-pretraining/run_mim_no_trainer.py

@@ -14,7 +14,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "torch>=1.5.0",
 #     "torchvision>=0.6.0",
 #     "datasets>=1.8.0",

examples/pytorch/instance-segmentation/run_instance_segmentation.py

@@ -14,7 +14,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "albumentations >= 1.4.16",
 #     "timm",
 #     "datasets",

examples/pytorch/instance-segmentation/run_instance_segmentation_no_trainer.py

@@ -14,7 +14,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "albumentations >= 1.4.16",
 #     "timm",
 #     "datasets",

examples/pytorch/language-modeling/run_clm.py

@@ -15,7 +15,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "albumentations >= 1.4.16",
 #     "accelerate >= 0.12.0",
 #     "torch >= 1.3",

examples/pytorch/language-modeling/run_clm_no_trainer.py

@@ -15,7 +15,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "albumentations >= 1.4.16",
 #     "accelerate >= 0.12.0",
 #     "torch >= 1.3",

examples/pytorch/language-modeling/run_fim.py

@@ -15,7 +15,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "albumentations >= 1.4.16",
 #     "accelerate >= 0.12.0",
 #     "torch >= 1.3",

examples/pytorch/language-modeling/run_fim_no_trainer.py

@@ -15,7 +15,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "albumentations >= 1.4.16",
 #     "accelerate >= 0.12.0",
 #     "torch >= 1.3",

examples/pytorch/language-modeling/run_mlm.py

@@ -15,7 +15,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "albumentations >= 1.4.16",
 #     "accelerate >= 0.12.0",
 #     "torch >= 1.3",

examples/pytorch/language-modeling/run_mlm_no_trainer.py

@@ -15,7 +15,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "albumentations >= 1.4.16",
 #     "accelerate >= 0.12.0",
 #     "torch >= 1.3",

examples/pytorch/language-modeling/run_plm.py

@@ -15,7 +15,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "albumentations >= 1.4.16",
 #     "accelerate >= 0.12.0",
 #     "torch >= 1.3",

examples/pytorch/multiple-choice/run_swag.py

@@ -15,7 +15,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "accelerate >= 0.12.0",
 #     "sentencepiece != 0.1.92",
 #     "protobuf",

examples/pytorch/multiple-choice/run_swag_no_trainer.py

@@ -15,7 +15,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "accelerate >= 0.12.0",
 #     "sentencepiece != 0.1.92",
 #     "protobuf",

examples/pytorch/object-detection/run_object_detection.py

@@ -14,7 +14,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "albumentations >= 1.4.16",
 #     "timm",
 #     "datasets>=4.0",

examples/pytorch/object-detection/run_object_detection_no_trainer.py

@@ -14,7 +14,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "albumentations >= 1.4.16",
 #     "timm",
 #     "datasets>=4.0",

examples/pytorch/semantic-segmentation/run_semantic_segmentation.py

@@ -14,7 +14,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "datasets >= 2.0.0",
 #     "torch >= 1.3",
 #     "accelerate",

examples/pytorch/semantic-segmentation/run_semantic_segmentation_no_trainer.py

@@ -14,7 +14,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "datasets >= 2.0.0",
 #     "torch >= 1.3",
 #     "accelerate",

examples/pytorch/speech-pretraining/run_wav2vec2_pretraining_no_trainer.py

@@ -14,7 +14,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "datasets[audio] >= 1.12.0",
 #     "torch >= 1.5",
 #     "torchaudio",

examples/pytorch/speech-recognition/run_speech_recognition_ctc.py

@@ -15,7 +15,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "datasets[audio] >= 1.18.0",
 #     "torch >= 1.5",
 #     "torchaudio",

examples/pytorch/speech-recognition/run_speech_recognition_ctc_adapter.py

@@ -15,7 +15,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "datasets[audio] >= 1.18.0",
 #     "torch >= 1.5",
 #     "torchaudio",

examples/pytorch/speech-recognition/run_speech_recognition_seq2seq.py

@@ -15,7 +15,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "datasets[audio] >= 1.18.0",
 #     "torch >= 1.5",
 #     "torchaudio",

examples/pytorch/summarization/run_summarization.py

@@ -15,7 +15,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "accelerate >= 0.12.0",
 #     "datasets >= 1.8.0",
 #     "sentencepiece != 0.1.92",

examples/pytorch/summarization/run_summarization_no_trainer.py

@@ -15,7 +15,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "accelerate >= 0.12.0",
 #     "datasets >= 1.8.0",
 #     "sentencepiece != 0.1.92",

examples/pytorch/text-classification/run_classification.py

@@ -15,7 +15,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "accelerate >= 0.12.0",
 #     "datasets >= 1.8.0",
 #     "sentencepiece != 0.1.92",

examples/pytorch/text-classification/run_glue.py

@@ -15,7 +15,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "accelerate >= 0.12.0",
 #     "datasets >= 1.8.0",
 #     "sentencepiece != 0.1.92",

examples/pytorch/text-classification/run_glue_no_trainer.py

@@ -14,7 +14,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "accelerate >= 0.12.0",
 #     "datasets >= 1.8.0",
 #     "sentencepiece != 0.1.92",

examples/pytorch/text-classification/run_xnli.py

@@ -16,7 +16,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "accelerate >= 0.12.0",
 #     "datasets >= 1.8.0",
 #     "sentencepiece != 0.1.92",

examples/pytorch/text-generation/run_generation.py

@@ -16,7 +16,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "accelerate >= 0.21.0",
 #     "sentencepiece != 0.1.92",
 #     "protobuf",

examples/pytorch/text-generation/run_generation_contrastive_search.py

@@ -15,7 +15,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "accelerate >= 0.21.0",
 #     "sentencepiece != 0.1.92",
 #     "protobuf",

examples/pytorch/token-classification/run_ner.py

@@ -15,7 +15,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "accelerate >= 0.12.0",
 #     "seqeval",
 #     "datasets >= 1.8.0",

examples/pytorch/token-classification/run_ner_no_trainer.py

@@ -15,7 +15,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "accelerate >= 0.12.0",
 #     "seqeval",
 #     "datasets >= 1.8.0",

examples/pytorch/translation/run_translation.py

@@ -15,7 +15,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "accelerate >= 0.12.0",
 #     "datasets >= 1.8.0",
 #     "sentencepiece != 0.1.92",

examples/pytorch/translation/run_translation_no_trainer.py

@@ -15,7 +15,7 @@
 
 # /// script
 # dependencies = [
-#     "transformers==4.55.3",
+#     "transformers==4.55.4",
 #     "accelerate >= 0.12.0",
 #     "datasets >= 1.8.0",
 #     "sentencepiece != 0.1.92",

setup.py

@@ -463,7 +463,7 @@ install_requires = [
 
 setup(
     name="transformers",
-    version="4.55.3",  # expected format is one of x.y.z.dev0, or x.y.z.rc1 or x.y.z (no to dashes, yes to dots)
+    version="4.55.4",  # expected format is one of x.y.z.dev0, or x.y.z.rc1 or x.y.z (no to dashes, yes to dots)
     author="The Hugging Face team (past and future) with the help of all our contributors (https://github.com/huggingface/transformers/graphs/contributors)",
     author_email="transformers@huggingface.co",
     description="State-of-the-art Machine Learning for JAX, PyTorch and TensorFlow",

src/transformers/__init__.py

@@ -18,7 +18,7 @@
 # to defer the actual importing for when the objects are requested. This way `import transformers` provides the names
 # in the namespace without actually importing anything (and especially none of the backends).
 
-__version__ = "4.55.3"
+__version__ = "4.55.4"
 
 from pathlib import Path
 from typing import TYPE_CHECKING

src/transformers/integrations/mxfp4.py

@@ -172,7 +172,7 @@ class Mxfp4GptOssExperts(nn.Module):
             torch.zeros(self.num_experts, self.hidden_size, dtype=torch.float32), requires_grad=False
         )
         self.alpha = 1.702
-
+        self.limit = getattr(config, "swiglu_limit", 7.0)
         self.gate_up_proj_precision_config = None
         self.down_proj_precision_config = None
 
@@ -185,7 +185,7 @@ class Mxfp4GptOssExperts(nn.Module):
         swiglu_fn = triton_kernels_hub.swiglu.swiglu_fn
 
         with torch.cuda.device(hidden_states.device):
-            act = FusedActivation(FnSpecs("swiglu", swiglu_fn, ("alpha", "limit")), (self.alpha, None), 2)
+            act = FusedActivation(FnSpecs("swiglu", swiglu_fn, ("alpha", "limit")), (self.alpha, self.limit), 2)
 
             intermediate_cache1 = matmul_ogs(
                 hidden_states,

src/transformers/quantizers/quantizer_mxfp4.py

@@ -61,16 +61,7 @@ class Mxfp4HfQuantizer(HfQuantizer):
             return
 
         if not torch.cuda.is_available():
-            raise RuntimeError("Using MXFP4 quantized models requires a GPU")
-
-        if not is_accelerate_available():
-            raise ImportError("Using mxfp4 requires Accelerate: `pip install accelerate`")
-
-        compute_capability = torch.cuda.get_device_capability()
-        major, minor = compute_capability
-
-        if not is_triton_available("3.4.0") or not is_triton_kernels_availalble():
-            if self.pre_quantized and not self.quantization_config.dequantize:
+            if self.pre_quantized:
                 logger.warning_once(
                     "Using MXFP4 quantized models requires a GPU, we will default to dequantizing the model to bf16"
                 )