Sachin Raja

The requirements file for the deformable DETR project uses jax 0.3.17 and flax 0.5.3. However, I cannot go back in the version beyond 0.4.26 for jax and 0.8.4 for flax...

https://pytorch.org/docs/stable/generated/torch.roll.html

[BUG] ValueError: [scatter] Cannot calculate VJP with respect to indices.

6

**Describe the bug** I understand that HungarianMatching algorithm requires linear_sum_assignment from scipy, which needs cost matrix to be evaluated. Hence, I cannot compile my train step function. However, if I...

Would it be possible to have MLX equivalents for jax.lax.scan and jax.lax.while_loop? Matching algorithms like Hungarian matching for object detection (DETR) need boolean arrays to be evaluated to correctly compute...

Thank you for this amazing work! Here is an Apple MLX port of DINO DETR: https://github.com/sachinraja13/DINO_DETR_MLX Results on the COCO 2017 validation dataset are within 2 basis points AP of...

Does MLX-VLM support fine-tuning of vision tower layers?

7

Hi @Blaizzy , thank you for your amazing contribution to the MLX community. Quick question please: Does MLX-VLM support fine-tuning of vision layers? From what I see in [this piece...

Error in FineTuning deepseek-vl-7b-chat-8bit

2

This is the command I'm using: ``` python -m mlx_vlm.lora --dataset ~/Datasets/BusinessVQA/fintabnet/val/vqa_dataset.hf --model-path ~/.cache/lm-studio/models/mlx-community/deepseek-vl-7b-chat-8bit --epochs 2 --batch-size 4 --learning-rate 5e-5 ``` Here is the console output: ``` INFO:__main__:Loading model from...

Thanks for this amazing work! Quick Question : Is Fine-tuning of Llava-Next supported using LORA/QLORA? Also, if I have sufficiently large task specific instruction fine-tuning dataset, can I Finetune all...