alfworld chat

alfworld_runs/alfworld_trial.py

@@ -8,6 +8,7 @@ import openai
 import importlib
 import alfworld
 import alfworld.agents.environment
+from utils import Model, get_chat, get_completion
 from env_history import EnvironmentHistory
 
 from typing import List, Dict, Any, Tuple
@@ -20,24 +21,17 @@ with open(os.path.join(FOLDER, PROMPT_FILE), 'r') as f:
 with open('./challenge_few_shot_examples.txt', 'r') as f:
     challenge_examples = f.read()
 
-def llm(prompt, stop=["\n"]):
+def llm(prompt: str, model: Model, stop: List[str] = ["\n"]):
     try:
         cur_try = 0
         while cur_try < 6:
-            response = openai.Completion.create(
-              model="text-davinci-002",
-              prompt=prompt,
-              temperature=cur_try * 0.2,
-              max_tokens=100,
-              top_p=1,
-              frequency_penalty=0.0,
-              presence_penalty=0.0,
-              stop=stop
-            )
-            text = response["choices"][0]["text"]
+            if model == "text-davinci-003":
+                text = get_completion(prompt=prompt, temperature=cur_try * 0.2)
+            else:
+                text = get_chat(prompt=prompt, model=model, temperature=cur_try * 0.2)
             # dumb way to do this
             if len(text.strip()) >= 5:
-                return response["choices"][0]["text"]
+                return text
             cur_try += 1
         return ""
     except Exception as e:
@@ -51,7 +45,7 @@ def process_ob(ob):
         ob = ob[ob.find('. ')+2:]    
     return ob
 
-def alfworld_run(env, base_prompt, memory: List[str], to_print=True, ob='') -> Tuple[EnvironmentHistory, bool]:
+def alfworld_run(env, base_prompt, memory: List[str], to_print=True, ob='', model: Model = "text-davinci-003") -> Tuple[EnvironmentHistory, bool]:
     if len(memory) > 3:
         env_history = EnvironmentHistory(base_prompt, ob, memory[-3:], [])
     else:
@@ -97,7 +91,8 @@ def run_trial(
         world_log_path: str,
         trial_idx: int,
         env_configs: List[Dict[str, Any]],
-        use_memory: bool
+        use_memory: bool,
+        model: Model,
     ) -> List[Dict[str, Any]]:
     importlib.reload(alfworld)
     importlib.reload(alfworld.agents.environment)
@@ -133,7 +128,7 @@ def run_trial(
         for i, (k, v) in enumerate(PREFIXES.items()):
             if name.startswith(k):
                 base_prompt = 'Interact with a household to solve a task. Here are two examples.\n' + d[f'react_{v}_1'] + d[f'react_{v}_0']
-                final_env_history, is_success = alfworld_run(env, base_prompt, env_config["memory"] if use_memory else [], to_print=True, ob=ob)
+                final_env_history, is_success = alfworld_run(env, base_prompt, env_config["memory"] if use_memory else [], to_print=True, ob=ob, model=model)
 
                 # update env config
                 if is_success:

alfworld_runs/main.py

@@ -16,6 +16,7 @@ def get_args():
     parser.add_argument("--is_resume", action='store_true', help="To resume run")
     parser.add_argument("--resume_dir", type=str, help="If resume, the logging directory", default="")
     parser.add_argument("--start_trial_num", type=int, help="If resume, the start trial num", default=0)
+    parser.add_argument("--model", type=str, help="The model to use. One of `gpt-4`, `gpt-3.5-turbo`, or `text-davinci-003")
 
     args = parser.parse_args()
 
@@ -96,7 +97,7 @@ def main(args) -> None:
             open(trial_env_configs_log_path, 'w').close()
 
         # run trial
-        run_trial(trial_log_path, world_log_path, trial_idx, env_configs, args.use_memory)
+        run_trial(trial_log_path, world_log_path, trial_idx, env_configs, args.use_memory, args.model)
 
         # update memory if needed
         if args.use_memory:

alfworld_runs/run_base.sh

@@ -1,4 +0,0 @@
-python main.py \
-        --num_trials 10 \
-        --num_envs 134 \
-        --run_name "base_run_logs" \

alfworld_runs/run_reflexion.sh

@@ -2,4 +2,5 @@ python main.py \
         --num_trials 10 \
         --num_envs 134 \
         --run_name "reflexion_run_logs" \
-        --use_memory
+        --use_memory \
+        --model "gpt-3.5-turbo"

alfworld_runs/run_simple.sh

@@ -0,0 +1,6 @@
+python main.py \
+        --num_trials 10 \
+        --num_envs 134 \
+        --run_name "base_run_logs_gpt_35_turbo" \
+        --model "gpt-3.5-turbo"
+

alfworld_runs/utils.py

@@ -6,26 +6,39 @@ from tenacity import (
     wait_random_exponential, # type: ignore
 )
 
-from typing import Optional, List, Union
+from typing import Optional, List, Union, Literal
+
+Model = Literal["gpt-4", "gpt-3.5-turbo", "text-davinci-003"]
 
 openai.api_key = os.getenv('OPENAI_API_KEY')
 
 @retry(wait=wait_random_exponential(min=1, max=60), stop=stop_after_attempt(6))
-def get_completion(prompt: Union[str, List[str]], max_tokens: int = 256, stop_strs: Optional[List[str]] = None, is_batched: bool = False) -> Union[str, List[str]]:
-    assert (not is_batched and isinstance(prompt, str)) or (is_batched and isinstance(prompt, list))
+def get_completion(prompt: str, temperature: float = 0.0, max_tokens: int = 256, stop_strs: Optional[List[str]] = None) -> str:
     response = openai.Completion.create(
         model='text-davinci-003',
         prompt=prompt,
-        temperature=0.0,
+        temperature=temperature,
         max_tokens=max_tokens,
         top_p=1,
         frequency_penalty=0.0,
         presence_penalty=0.0,
         stop=stop_strs,
     )
-    if is_batched:
-        res: List[str] = [""] * len(prompt)
-        for choice in response.choices:
-            res[choice.index] = choice.text
-        return res
     return response.choices[0].text
+
+@retry(wait=wait_random_exponential(min=1, max=60), stop=stop_after_attempt(6))
+def get_chat(prompt: str, model: Model, max_tokens: int = 256, stop_strs: Optional[List[str]] = None, is_batched: bool = False) -> str:
+    assert model != "text-davinci-003"
+    messages = [
+        {
+            "role": "user",
+            "content": prompt
+        }
+    ]
+    response = openai.Completion.create(
+        model=model,
+        messages=messages,
+        max_tokens=max_tokens,
+        stop=stop_strs,
+    )
+    return response.choices[0].message.content

programming_runs/test.py

@@ -1,55 +0,0 @@
-# Fails 2
-def minReverseOperations(self, n: int, p: int, banned: List[int], k: int) -> List[int]:
-        from collections import deque
-        banned = set(banned)
-        arr = tuple(0 if i in banned else (1 if i == p else 0) for i in range(n))
-        queue = deque([(arr, p, 0)])  # Add a third element to the tuple to store the number of
-        ans = [-1] * n
-        visited = set()
-
-        while queue:
-            cur_arr, cur_pos, ops = queue.popleft()
-            
-            if cur_pos not in visited:
-                visited.add(cur_pos)
-                ans[cur_pos] = ops
-
-                for i in range(n):
-                    for j in range(i + k, n + 1):
-                        new_arr = cur_arr[:i] + tuple(reversed(cur_arr[i:j])) + cur_arr[j:]
-                        new_pos = new_arr.index(1)
-                        
-                        if new_pos not in banned and (new_arr, new_pos) not in visited:
-                            queue.append((new_arr, new_pos, ops + 1))
-
-        return ans
-
-# Fails 1
-def minReverseOperations(self, n: int, p: int, banned: List[int], k: int) -> List[int]:
-        from collections import deque
-        banned = set(banned)
-        arr = tuple(0 if i in banned else (1 if i == p else 0) for i in range(n))
-        queue = deque([(arr, p, 0)])  # Add a third element to the tuple to store the number of operations
-        ans = [-1] * n
-        visited = set()
-
-        while queue:
-            cur_arr, cur_pos, ops = queue.popleft()
-            
-            if cur_pos not in visited:
-                visited.add(cur_pos)
-                ans[cur_pos] = ops
-
-                for i in range(n):
-                    for j in range(i + k, n + 1):
-                        # Check if the subarray to be reversed contains any banned positions
-                        if any(cur_arr[i:x] for x in range(i, j) if x in banned):
-                            continue
-                        
-                        new_arr = cur_arr[:i] + tuple(reversed(cur_arr[i:j])) + cur_arr[j:]
-                        new_pos = new_arr.index(1)
-                        
-                        if new_pos not in banned and (new_arr, new_pos) not in visited:
-                            queue.append((new_arr, new_pos, ops + 1))
-
-        return ans