deprecate TSModelForCausalLM (huggingface#1173)

* deprecate TSModelForCausalLM

* add prepare_jit_inputs

* remove workflow

Author: echarlaix

.github/workflows/test_generation.yml

@@ -18,8 +18,6 @@ limitations under the License.
 
 Based on the script [`run_generation.py`](https://github.com/huggingface/transformers/blob/main/examples/pytorch/text-generation/run_generation.py).
 
-The original generation task only supported the PyTorch eager model. By calling the `TSModelForCausalLM` class, we can now support a TorchScript model for generation tasks.
-
 This example also allows us to apply different quantization approaches (such as dynamic, static, The example applies post-training static quantization on a gptj model).
 
 Example usage: