Why do I get this error when training the model, running the anaconda, py notebook locally?

[nshivani]

So here is my problem while execution code from the Lecture “Training Process”.

**model = BasicModelRunner(“EleutherAI/pythia-410m”) **
model.load_data_from_jsonlines(“C:\\lamini_docs.jsonl”)
model.train(is_public=True)

When I execute the above, I get the below error:

File ~\AppData\Local\anaconda3\Lib\site-packages\llama\runners\basic_model_runner.py:82, in BasicModelRunner.load_data(self, data, verbose, input_key, output_key)
74 input_output_objects = [
75 [
76 Input(input=d[input_key]),
(…)
79 for d in data
80 ]
81 except KeyError:
—> 82 raise ValueError(“Each object must have input_key={input_key} (and optionally output_key={output_key}). You can pass in different input_key and output_keys”)
83 self.data.extend(input_output_objects)
84 if verbose:

ValueError: Each object must have input_key={input_key} (and optionally output_key={output_key}). You can pass in different input_key and output_keys

[elirod]

Hi @nshivani

If you are running locally you have to download all the files attached on the course platform to get it work.

You can do that by following:

Click on the jupyter logo at the top

Choose the file:

Then, on the new tab, click on “Files” and “Download”

Best regards
elirod

[nshivani]

@elirod

elirod

I have downloaded all the files and I still receive the same error.

[nshivani]

So it eventually worked.
It seems the basic_model_runner.py accepts dataset which are of the form “input” and “output” rather than “question” and “answer”.
I had to replace the “question” and “answer” keys in lamini_docs.jsonl to “input” and “output”. And it started working.
I have no clue if it is the problem with the llamini version I have installed.
Thanks.

[elirod]

Nice work @nshivani

i’m glad to help