Xformers is not installed correctly.

[david-waterworth]

System Info

• transformers version: 4.30.2
• Platform: Linux-5.15.0-76-generic-x86_64-with-glibc2.35
• Python version: 3.10.6
• Huggingface_hub version: 0.16.4
• Safetensors version: 0.3.1
• PyTorch version (GPU?): 2.0.1+cu117 (True)
• Tensorflow version (GPU?): not installed (NA)
• Flax version (CPU?/GPU?/TPU?): not installed (NA)
• Jax version: not installed
• JaxLib version: not installed
• Using GPU in script?: yes
• Using distributed or parallel set-up in script?: no

Who can help?

No response

Information

• The official example scripts
• My own modified scripts

Tasks

• An officially supported task in the examples folder (such as GLUE/SQuAD, ...)
• My own task or dataset (give details below)

Reproduction

from transformers import pipeline
pipe = pipeline("text-classification", model="roberta-base", device=0)

Edit: I know this model isn't trained for the "text-classification" task, I get the same problem with a private model I fine tuned.

Results in the message

...
Xformers is not installed correctly. If you want to use memory_efficient_attention to accelerate training use the following command to install Xformers
pip install xformers.

But I'm using torch==2.0.1 and memory-efficient-attention states "If you have PyTorch 2.0 installed, you shouldn’t use xFormers!"

The message is confusing - I have torch 2.0 installed and pipeline is for inference. This message doesn't occur if I use AutoModelForSequenceClassification.from_pretrained

Expected behavior

The documentation or the warning message are inconsistent.

[sgugger]

It looks like the pipeline is back to importing every model (this message comes from trying to access an unrelated model). I'll have a look later this week. You can ignore that warning in the meantime, it's irrelevant.

[sgugger]

Should be fixed by the PR linked above.

[engageintellect]

same issue for me on this basic example:

import argparse
from transformers import pipeline

# Create the parser
parser = argparse.ArgumentParser(description="Perform sentiment analysis")

# Add an argument
parser.add_argument('Text', type=str, help="the text to analyze")

# Parse the argument
args = parser.parse_args()

# Load the classifier
classifier = pipeline("sentiment-analysis", model="distilbert-base-uncased-finetuned-sst-2-english")

# Perform sentiment analysis
res = classifier(args.Text)

# Print the result
print(res)

Reinstalled transformers, using v.4.31.0

[sgugger]

The fix is not in v4.31.0, you will need to use a source install.