github/langchain
1
0
Fork 0
mirror of https://github.com/hwchase17/langchain.git synced 2025-09-03 12:07:36 +00:00
Code Issues Packages Projects Releases Wiki Activity

community[patch] : publishes duration as milliseconds to Fiddler (#19166)

Browse Source 
**Description:** Many LLM steps complete in sub-second duration, which
can lead to non-collection of duration field for Fiddler. This PR
updates duration from seconds to milliseconds.
**Issue:** [INTERNAL] FDL-17568
**Dependencies:** NA
**Twitter handle:** behalder

Co-authored-by: Barun Halder <barun@fiddler.ai>
This commit is contained in:
Barun Amalkumar Halder
2024-03-15 14:04:56 -07:00
committed by GitHub
parent 745d2476a2
commit 34d6f0557d
1 changed files with 5 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
libs/community/langchain_community/callbacks/fiddler_callback.py
View File

@@ -24,7 +24,7 @@ FAILURE = "failure"
# Default values # Default values
DEFAULT_MAX_TOKEN = 65536 DEFAULT_MAX_TOKEN = 65536
DEFAULT_MAX_DURATION = 120 DEFAULT_MAX_DURATION = 120000
# Fiddler specific constants # Fiddler specific constants
PROMPT = "prompt" PROMPT = "prompt"
@@ -154,7 +154,7 @@ class FiddlerCallbackHandler(BaseCallbackHandler):
dataset_info=dataset_info, dataset_info=dataset_info,
dataset_id="train", dataset_id="train",
model_task=self.fdl.ModelTask.LLM, model_task=self.fdl.ModelTask.LLM,
features=[PROMPT, RESPONSE, CONTEXT], features=[PROMPT, CONTEXT, RESPONSE],
target=FEEDBACK, target=FEEDBACK,
metadata_cols=[ metadata_cols=[
RUN_ID, RUN_ID,
@@ -297,12 +297,12 @@ class FiddlerCallbackHandler(BaseCallbackHandler):
) -> Any: ) -> Any:
run_id = kwargs[RUN_ID] run_id = kwargs[RUN_ID]
self.run_id_prompts[run_id] = prompts self.run_id_prompts[run_id] = prompts
self.run_id_starttime[run_id] = int(time.time()) self.run_id_starttime[run_id] = int(time.time() * 1000)
def on_llm_end(self, response: LLMResult, **kwargs: Any) -> None: def on_llm_end(self, response: LLMResult, **kwargs: Any) -> None:
flattened_llmresult = response.flatten() flattened_llmresult = response.flatten()
run_id = kwargs[RUN_ID] run_id = kwargs[RUN_ID]
run_duration = self.run_id_starttime[run_id] - int(time.time()) run_duration = int(time.time() * 1000) - self.run_id_starttime[run_id]
model_name = "" model_name = ""
token_usage_dict = {} token_usage_dict = {}
@@ -329,7 +329,7 @@ class FiddlerCallbackHandler(BaseCallbackHandler):
def on_llm_error(self, error: BaseException, **kwargs: Any) -> None: def on_llm_error(self, error: BaseException, **kwargs: Any) -> None:
run_id = kwargs[RUN_ID] run_id = kwargs[RUN_ID]
duration = int(time.time()) - self.run_id_starttime[run_id] duration = int(time.time() * 1000) - self.run_id_starttime[run_id]
self._publish_events( self._publish_events(
run_id, [""] * len(self.run_id_prompts[run_id]), duration, FAILURE run_id, [""] * len(self.run_id_prompts[run_id]), duration, FAILURE