Return all scores parameter doesn't work with model deployed on Inf1

Hello there,

I’ve deployed a text classification model (based on xlm-roberta) on ml.inf1.2xlarge instance following @philschmid’s blog article.

My problem is that when I call the model and put the return_all_scores=True parameter in the payload, I only have the class with the highest confidence score. I don’t know whether the problem comes from the custom predict_fn() or DLC I’m using or something else, if someone could help me resolve the issue it’d be much appreciated :blush:

  • Custom inference script
import os
from transformers import AutoConfig, AutoTokenizer
import torch
import torch.neuron

# To use one neuron core per worker
os.environ["NEURON_RT_NUM_CORES"] = "1"

# saved weights name

def model_fn(model_dir):
    # load tokenizer and neuron model from model_dir
    tokenizer = AutoTokenizer.from_pretrained(model_dir)
    model = torch.jit.load(os.path.join(model_dir, AWS_NEURON_TRACED_WEIGHTS_NAME))
    model_config = AutoConfig.from_pretrained(model_dir)

    return model, tokenizer, model_config

def predict_fn(data, model_tokenizer_model_config):
    # destruct model, tokenizer and model config
    model, tokenizer, model_config = model_tokenizer_model_config

    # create embeddings for inputs
    inputs = data.pop("inputs", data)
    embeddings = tokenizer(
    # convert to tuple for neuron model
    neuron_inputs = tuple(embeddings.values())

    # run prediciton
    with torch.no_grad():
        predictions = model(*neuron_inputs)[0]
        scores = torch.nn.Softmax(dim=1)(predictions)

    # return dictonary, which will be json serializable
    return [{"label": model_config.id2label[item.argmax().item()], "score": item.max().item()} for item in scores]
  • DLC container: