Spaces:
Running
Unit 2 test
So I keep failing the final test of Unit 2 because my code apparently doesn't follow EXACTLY what the "reference" is and I keep getting told "These deviations mean the student's solution does not fully meet the requirements and criteria outlined in the challenge", when the challenge itself only gives one direction and a few comments about what to add and where. A couple of examples is the First question:
Create a Basic Code Agent with Web Search Capability
I kept failing it because I apparently wasn't putting in the "correct" model id.
Second question: Make an Manager Agent that calls a tool calling agent to search the web:
Info
❌ Incorrect! The student's solution has several issues that prevent it from being functionally equivalent to the reference solution: 1. Web Agent Tools: The student adds only DuckDuckGOSearchTool, whereas the reference includes both DuckDuckGoSearchTool() and visit_webpage. This is incomplete. 2. Manager Agent Tools: The student incorrectly includes web_agent under tools instead of managed_agents. 3. Model Usage: The student uses HfApiModel() instead of model as specified in the reference. This could lead to errors if model is a specific predefined model elsewhere. 4. Managed Agents: The manager agent should have web_agent under managed_agents, not tools. 5. Max Steps: The student sets max_steps=3 for the web agent, whereas the reference sets it to 10. 6. Authorized Imports: The student did not include additional_authorized_imports=["time", "numpy", "pandas"] in the CodeAgent. These deviations mean the student's solution does not fully meet the requirements and criteria outlined in the challenge. The student's solution has several issues that prevent it from being functionally equivalent to the reference solution: 1. Web Agent Tools: The student adds only DuckDuckGOSearchTool, whereas the reference includes both DuckDuckGoSearchTool() and visit_webpage. This is incomplete. 2. Manager Agent Tools: The student incorrectly includes web_agent under tools instead of managed_agents. 3. Model Usage: The student uses HfApiModel() instead of model as specified in the reference. This could lead to errors if model is a specific predefined model elsewhere. 4. Managed Agents: The manager agent should have web_agent under managed_agents, not tools. 5. Max Steps: The student sets max_steps=3 for the web agent, whereas the reference sets it to 10. 6. Authorized Imports: The student did not include additional_authorized_imports=["time", "numpy", "pandas"] in the CodeAgent. These deviations mean the student's solution does not fully meet the requirements and criteria outlined in the challenge.
NOWHERE did the challenge say to use additional authorized imports or to create a tool called visit_webpage
I get this isn't graded, but I feel like this quiz was poorly executed since unless you do EXACTLY what the "reference" code is, you fail.
I had similar problems, I also think this quiz was poorly executed.
i have same feelings, in the test there is not many instructions on what to write in the code,
you just get the information, make a web search agent/tool
and when you do so you get the information to use the specific model? i mean how can i know what model to use if its not specified in the task? same with other questions
in one of the tasks i got
The student's solution does not meet all the assessment criteria. Specifically, it does not set the step limit appropriately.
it uses specific tools (DuckDuckGoSearchTool and HfApiModel) which may not match the intended custom_tool and model from the reference solution
so i had to make the custom_tool and model as it is in the reference solution
the test is really nce and it tests your practical skills but the comparision to the reference model is kind of not working now ;/