i am facing the degraded or low accuracy from the API (model: Llama-4-Maverick-17B-128E-Instruct).
scenarios and platform:
when used in google collab in sequential manner, like images goes to api one after another for pre-processing super high accuracy.
When used in local development in linux or windows env, with parallel tasks or asyncio
where one image goes to api and earlier image go to next step.
accuracy is very low.
i did check all the package versions, internet speet etc, nothing worked.
is it because of parallel task management ?
language: Python.
1 Like
Hi @bharatkumar ,
Thank you for reaching out to us. We acknowledge your concern and have noted it for review. Our team will look into this thoroughly, and we will get back to you with an update.
Regards,
Shivani Moze
Thanks for quick response.
Hi Bharat Kumar,
We reviewed your query and tested a few cases on our side, but did not encounter any issues with accuracy. Could you please clarify what you mean by “accuracy” in your case and share more details about the issue you faced? If possible, providing a code snippet along with sample images and the corresponding output where you observed the inaccuracy would be very helpful. This will help us understand your scenario more clearly and guide you better.
If you’re not comfortable posting these details publicly in the community, you’re welcome to DM me directly.
Best regards,
Tanish Sakate
Hi Tanish,
Thanks for the clarity, let me replicate the issue with some test images, then will send you the results with test images and code snippet.
for us accuracy is: Extracting all visible text on the images without missing and wrongly interpreting.
Thank you