Marcel Ambo Ndowah comments

Results 12 comments of


                                            Marcel Ambo Ndowah

DTensor issues when running Llama4ForConditionalGeneration with tensor parallel.

Hi @czkkkkkk, Are you already working on this? If not, I'd be happy to take it.

Add rounding error check to _maybe_log_save_evaluate

Sure, I'll proceed with this right away

Add rounding error check to _maybe_log_save_evaluate

Certainly, I will proceed with this.

Add rounding error check to _maybe_log_save_evaluate

I tried the logging with wandb and it pointed out an issue logs["loss"] = str(round(log_eval), 4) instead of logs["loss"] = str(round(log_eval, 4)). After this fix the logging seems to work...

Add rounding error check to _maybe_log_save_evaluate

I'm not sure how to verify if the values are correctly displayed on the wandb table/website. Could you provide any guidance on how to confirm this, please?

Add rounding error check to _maybe_log_save_evaluate

![Screen Shot 2025-06-17 at 11 48 02 PM](https://github.com/user-attachments/assets/eb3857b5-5105-4aa9-87f1-2871a9798899)

Add rounding error check to _maybe_log_save_evaluate

Hi @SunMarc, please let me know if any modification is needed on this PR.

Add rounding error check to _maybe_log_save_evaluate

Hi @SunMarc, just wanted to kindly follow up on the PR to see if everything looks good now or if any further changes are needed. Thanks!

`ConditionalDetrImageProcessor` still accepts the deprecated parameter `max_size`

Hi @Rahul7-77, do you still plan to work on this? If not, I’d be happy to pick it up.

`ConditionalDetrImageProcessor` still accepts the deprecated parameter `max_size`

Hi @qubvel, I've looked into this issue and attempted to filter models using the following code: from huggingface_hub import list_models from huggingface_hub import ModelFilter models = list_models(filter="conditional_detr")) for model in...