[Refactor] Remove LLM features for release

ghstack-source-id: f10f7a57e9140f5bf2544dbaeb5329891b151911
Pull Request resolved: #2912

Author: vmoens

.github/unittest/linux_libs/scripts_llm/install.sh

@@ -27,8 +27,6 @@ python -c "import transformers, datasets"
 
 pytest test/test_rlhf.py --instafail -v --durations 200 --capture no --error-for-skips
 
-pytest test/test_actors.py test/test_collector.py -k llm --instafail -v --durations 200 --capture no --error-for-skips --runslow
-
 pytest examples/rlhf/train_rlhf.py \
   sys.device=cuda:0 sys.ref_device=cuda:0 \
   model.name_or_path=gpt2 train.max_epochs=2 \

.github/unittest/linux_libs/scripts_llm/post_process.sh

@@ -521,7 +521,6 @@ TorchRL offers a series of custom built-in environments.
     ChessEnv
     PendulumEnv
     TicTacToeEnv
-    LLMEnv
     LLMHashingEnv