phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-06-28 19:04:19 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

feat: [new open benchmark] BFCL_v3 (#1578)

Browse source 
# What does this PR do?
create a new dataset BFCL_v3 from
https://gorilla.cs.berkeley.edu/blogs/13_bfcl_v3_multi_turn.html

overall each question asks the model to perform a task described in
natural language, and additionally a set of available functions and
their schema are given for the model to choose from. the model is
required to write the function call form including function name and
parameters , to achieve the stated purpose. the results are validated
against provided ground truth, to make sure that the generated function
call and the ground truth function call are syntactically and
semantically equivalent, by checking their AST .



## Test Plan

start server by 

```
llama stack run ./llama_stack/templates/ollama/run.yaml
```

then send traffic
```
 llama-stack-client eval run-benchmark "bfcl"  --model-id   meta-llama/Llama-3.2-3B-Instruct    --output-dir /tmp/gpqa    --num-examples   2
```




[//]: # (## Documentation)
This commit is contained in:
yyymeta 2025-03-14 12:50:49 -07:00 • committed by GitHub
parent 78d4872c0c
commit a626b7bce3
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
15 changed files with 1546 additions and 9 deletions
Download patch file Download diff file Expand all files Collapse all files

7
llama_stack/providers/inline/eval/meta_reference/eval.py
View file

@ -12,7 +12,7 @@ from llama_stack.apis.agents import Agents, StepType
from llama_stack.apis.benchmarks import Benchmark
from llama_stack.apis.datasetio import DatasetIO
from llama_stack.apis.datasets import Datasets
from llama_stack.apis.inference import Inference, UserMessage
from llama_stack.apis.inference import Inference, SystemMessage, UserMessage
from llama_stack.apis.scoring import Scoring
from llama_stack.distribution.datatypes import Api
from llama_stack.providers.datatypes import BenchmarksProtocolPrivate
@ -118,7 +118,7 @@ class MetaReferenceEvalImpl(
for i, x in tqdm(enumerate(input_rows)):
assert ColumnName.chat_completion_input.value in x, "Invalid input row"
input_messages = json.loads(x[ColumnName.chat_completion_input.value])
input_messages = [UserMessage(**x) for x in input_messages]
input_messages = [UserMessage(**x) for x in input_messages if x["role"] == "user"]
# NOTE: only single-turn agent generation is supported. Create a new session for each input row
session_create_response = await self.agents_api.create_agent_session(agent_id, f"session-{i}")
@ -168,10 +168,11 @@ class MetaReferenceEvalImpl(
generations.append({ColumnName.generated_answer.value: response.completion_message.content})
elif ColumnName.chat_completion_input.value in x:
chat_completion_input_json = json.loads(x[ColumnName.chat_completion_input.value])
input_messages = [UserMessage(**x) for x in chat_completion_input_json]
input_messages = [UserMessage(**x) for x in chat_completion_input_json if x["role"] == "user"]
messages = []
if candidate.system_message:
messages.append(candidate.system_message)
messages += [SystemMessage(**x) for x in chat_completion_input_json if x["role"] == "system"]
messages += input_messages
response = await self.inference_api.chat_completion(
model_id=candidate.model,