first version that works well when evaluating ppo and ddpt on semantic level

convlab/policy/vector/vector_base.py

convlab/policy/vector/vector_binary.py

convlab/policy/vector/vector_nodes.py

convlab/policy/vector/vector_uncertainty.py

convlab/policy/vtrace_DPT/create_descriptions.py

...ab/policy/vtrace_DPT/transformer_model/action_embedder.py

...ace_DPT/transformer_model/action_embeddings_multiwoz21.pt

...icy/vtrace_DPT/transformer_model/action_embeddings_sgd.pt

...ransformer_model/embedded_descriptions_base_multiwoz21.pt

...e_DPT/transformer_model/embedded_descriptions_base_sgd.pt

...olicy/vtrace_DPT/transformer_model/small_action_dict.json

...e_DPT/transformer_model/small_action_dict_multiwoz21.json

...y/vtrace_DPT/transformer_model/small_action_dict_sgd.json

convlab/util/multiwoz/lexicalize.py