Merge branch 'main' into bug/fix-scenario-agent-name

.gitignore

@@ -5,6 +5,13 @@ __pycache__
 dist
 .venv
 
+# Byte-compiled / optimized / DLL files
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
 # Log
 *.log
 *.log.*
@@ -13,6 +20,7 @@ llm_ft/checkpoints/*
 llm_ft/*_checkpoints/*
 !**/dummy_conversation.json
 !llm_ft/deepspeed_config_s2.json
+!llm_rl/data/*.json
 
 # Editor
 .idea
@@ -33,4 +41,161 @@ tests/state_of_the_union.txt
 
 # Build
 build
-!dummy_file
+!dummy_file
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+#poetry.lock
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#pdm.lock
+#   pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it
+#   in version control.
+#   https://pdm.fming.dev/#use-with-ide
+.pdm.toml
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# PyCharm
+#  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
+#  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
+#  and can be added to the global gitignore or merged into this file.  For a more nuclear
+#  option (not recommended) you can uncomment the following to ignore the entire idea folder.
+#.idea/
+
+./llm_rl/preprocess/GPT4-4_Redis_Easy_No_Slide
+
+llm_rl/*cache/

README.md

@@ -7,4 +7,4 @@ We split our overall framework into multiple parts
 2. Together AI Finetuning --> Input the train and test data / Output model checkpoint
 3. LLM Finetuning --> Input the train and test data / Output model checkpoint
 4. LLM Deplyment --> Input LLM Finetuned model checkpoint / Output Deployable OpenAI type API
-5. Eval --> Input model checkpoint / Output evaluation scores
+5. Eval --> Input model checkpoint / Output evaluation scores

data_process/fastchat_data_preprocess.py

@@ -0,0 +1,20 @@
+import json
+import os
+
+sotopia_data_dir = "/Users/pamela/Documents/capstone/sotopia-ft-data/ft-data-gpt4-gpt4-easy-2-side-partial"
+
+ft_data_list = []
+count = 0
+for file in os.listdir(sotopia_data_dir):
+    with open(os.path.join(sotopia_data_dir, file), 'r') as f:
+        file_dict = json.load(f)
+        fastchat_dict = {"id": f"identity_{count}", "conversations": []}
+        fastchat_dict["conversations"].append(
+            {"from": "human", "value": file_dict["prompt"]})
+        fastchat_dict["conversations"].append(
+            {"from": "gpt", "value": file_dict["result"]})
+        ft_data_list.append(fastchat_dict)
+        count += 1
+
+with open("fastchat-ft-gp4-gpt4-easy-2-side-partial.json", "w") as f:
+    f.write(json.dumps(ft_data_list, indent=4))

eval/llm_eval.py

@@ -0,0 +1,75 @@
+import json
+from tqdm import tqdm
+
+from langchain.llms import OpenAI
+from langchain.output_parsers import PydanticOutputParser
+from langchain.prompts import PromptTemplate
+from langchain.pydantic_v1 import BaseModel, Field
+
+
+class QuantitativeEval(BaseModel):
+    agent1_name: str = Field(description="Agent 1's name")
+    agent1_gain: int = Field(description="Agent 1's gain/loss")
+    agent2_name: str = Field(description="Agent 2's name")
+    agent2_gain: int = Field(description="Agent 2's gain/loss")
+
+
+def get_model_parser(model_name='text-davinci-003') -> (PromptTemplate, PydanticOutputParser):
+    model = OpenAI(model_name=model_name, temperature=0.0)
+    parser = PydanticOutputParser(pydantic_object=QuantitativeEval)
+
+    prompt_text = (
+        "Try to understand the following situation and answer the question in the end. "
+        "\n Situation: {situation}"
+        "\n Question: {question}"
+        "\n Please represent loss as negative values. {format_instructions}\n "
+    )
+
+    prompt = PromptTemplate(
+        template=prompt_text,
+        input_variables=["situation", "question"],
+        partial_variables={"format_instructions": parser.get_format_instructions()}
+    )
+
+    prompt_and_model = prompt | model
+
+    return prompt_and_model, parser
+
+
+def evaluate(environment_episode_map, environment_question_map, model_name='text-davinci-003'):
+    results = {}
+    model, response_parser = get_model_parser(model_name=model_name)
+
+    for environment_id, episodes in tqdm(environment_episode_map.items()):
+        results_for_env = []
+
+        for episode in episodes:
+            situation = episode["messages_and_rewards"]
+            question = environment_question_map.get(environment_id)
+
+            if question:
+                model_response = model.invoke({"situation": situation, "question": question})
+                parsed_output = response_parser.parse(model_response)
+                episode["output"] = parsed_output.dict()
+
+            results_for_env.append(episode)
+
+        results[environment_id] = results_for_env
+
+    return results
+
+
+def main():
+    with open("human_readable_eps_by_env.json", "r") as f:
+        env_eps_map = json.load(f)
+
+    with open("env_specific_eval.json", "r") as f:
+        env_question_map = json.load(f)
+
+    res = evaluate(env_eps_map, env_question_map)
+
+    with open("env_specific_eval_with_output.json", "w") as f:
+        json.dump(res, f)
+
+if __name__ == "__main__":
+    main()

eval/pull_data.py

@@ -0,0 +1,25 @@
+from sotopia.database.logs import EpisodeLog
+from sotopia.database.persistent_profile import EnvironmentProfile
+from sotopia.database.persistent_profile import AgentProfile
+import json
+
+TAG = "ft-llama-2-13b-chat_baseline_ruiyi_1010_7"  # Baseline tag
+
+HARD_ENVS = ["01H7VFHNV13MHN97GAH73E3KM8", "01H7VFHN5WVC5HKKVBHZBA553R", "01H7VFHN9W0WAFZCBT09PKJJNK", "01H7VFHPDZVVCDZR3AARA547CY", "01H7VFHPQQQY6H4DNC6NBQ8XTG", "01H7VFHN7WJK7VWVRZZTQ6DX9T", "01H7VFHPS5WJW2694R1MNC8JFY",
+             "01H7VFHNN7XTR99319DS8KZCQM", "01H7VFHQ11NAMZS4A2RDGDB01V", "01H7VFHPSWGDGEYRP63H2DJKV0", "01H7VFHNF4G18PC9JHGRC8A1R6", "01H7VFHNNYH3W0VRWVY178K2TK", "01H7VFHP8AN5643B0NR0NP00VE", "01H7VFHN7A1ZX5KSMT2YN9RXC4"]
+
+envs = []
+eps_by_env = dict()
+human_readable_eps_by_env = dict()
+
+for env_profile_id in HARD_ENVS:
+    eps = list(EpisodeLog.find(EpisodeLog.tag == TAG,
+                            EpisodeLog.environment == env_profile_id))
+    eps_by_env[env_profile_id] = eps
+    human_readable_eps_by_env[env_profile_id] = []
+    for ep in eps:
+        agent_profiles, messages_and_rewards = ep.render_for_humans()
+        human_readable_eps_by_env[env_profile_id].append({"env_pk": env_profile_id, "ep_pk": ep.pk, "agents": ep.agents, "messages_and_rewards": "\n".join(messages_and_rewards)})
+
+with open("human_readable_eps_by_env.json", "w") as f:
+    json.dump(human_readable_eps_by_env, f)