Using the real model name instead of hard code "model" (#231)

* Using the real model name instead of "model"

* Fix CLI multiple line backslash

* NeuronDecoderConfig from ExportConfig to NeuronConfig

* Revert to ExportConfig

README.md

@@ -98,7 +98,7 @@ trainer = Trainer(
 You can compile and export your 🤗 Transformers models to a serialized format before inference on Neuron devices:
 
 ```bash
-optimum-cli export neuron 
+optimum-cli export neuron \
   --model distilbert-base-uncased-finetuned-sst-2-english \
   --batch_size 1 \
   --sequence_length 32 \

optimum/exporters/neuron/__main__.py

@@ -205,8 +205,9 @@ def main_export(
         neuron_config = neuron_config_constructor(model.config, dynamic_batch_size=dynamic_batch_size, **input_shapes)
         if atol is None:
             atol = neuron_config.ATOL_FOR_VALIDATION
-        output_model_names = {"model": "model.neuron"}
-        models_and_neuron_configs = {"model": (model, neuron_config)}
+        model_name = model.name_or_path.split("/")[-1]
+        output_model_names = {model_name: "model.neuron"}
+        models_and_neuron_configs = {model_name: (model, neuron_config)}
         maybe_save_preprocessors(model, output.parent)
 
     if is_stable_diffusion: