Merge pull request #35 from stanford-oval/auto-sync-2024-04-25-23-24-14

auto-sync-2024-04-25-23-24-14

README.md

@@ -236,3 +236,4 @@ Please cite our paper if you use this code or part of it in your work:
 }
 ```
 
+

src/lm.py

@@ -12,11 +12,6 @@
 from dsp.modules.hf_client import send_hfvllm_request_v00, send_hftgi_request_v01_wrapped
 from transformers import AutoTokenizer
 
-try:
-    import anthropic
-except ImportError:
-    pass
-
 
 class OpenAIModel(dspy.OpenAI):
     """A wrapper class for dspy.OpenAI."""
@@ -120,7 +115,7 @@ def __init__(
     ):
         super().__init__(model)
         try:
-            from anthropic import Anthropic
+            from anthropic import Anthropic, RateLimitError
         except ImportError as err:
             raise ImportError("Claude requires `pip install anthropic`.") from err
 
@@ -193,7 +188,7 @@ def basic_request(self, prompt: str, **kwargs):
 
     @backoff.on_exception(
         backoff.expo,
-        (anthropic.RateLimitError,),
+        (RateLimitError,),
         max_time=1000,
         max_tries=8,
         on_backoff=backoff_hdlr,
@@ -224,11 +219,8 @@ def __call__(self, prompt, only_completed=True, return_sorted=False, **kwargs):
         for _ in range(n):
             response = self.request(prompt, **kwargs)
             self.log_usage(response)
-            # This is the original behavior in dspy/dsp/modules/anthropic.py.
-            # Comment it out because it can cause "IndexError: list index out of range" silently
-            # which is not transparent to developers.
-            # if only_completed and response.stop_reason == "max_tokens":
-            #     continue
+            if only_completed and response.stop_reason == "max_tokens":
+                continue
             completions = [c.text for c in response.content]
         return completions