-
I encountered an issue while fine-tuning the TinyChart-3B-768 model using the TinyChartData dataset. The initial loss is unexpectedly high, reaching 7.6. Additionally, when using the original DeepSpee…
-
### System Info
In current implementation of VLMs, the "_supports_sdpa" attribute checks and activates SDPA attention only for the language model. For example in [Llava](https://github.com/huggingf…
-
### 0.5b response is norm but 7b wrong
the same image,where i chage the code is` pretrained = "/home/shihongyu/MMLM_models/lmms-lab/llava-onevision-qwen2-7b-ov"
model_name = "llava_qwen"
device = "…
-
Hello, I would like to fine-tune or train TinyChart to improve its summarization skills. I have the impression that it doesn't capture all the data during summarization, whereas it does during data co…
-
Would love to see: [Bunny-Llama-3-8B-V](https://huggingface.co/BAAI/Bunny-Llama-3-8B-V) included in the Ollama models.
> Bunny is a family of lightweight but powerful multimodal models. It offers m…
-
Hi! I would like to ask few questions regarding the visual encoder part.
1. How does the SpatialBot model load the SigLip pre-trained model? I have downloaded the `siglip-so400m-patch14-384` model…
-
What should I specify as the `model_type` in the JSON file?
from transformers import AutoModel
model = AutoModel.from_pretrained("zxhezexin/openlrm-obj-base-1.1")
ValueError: Unrecogniz…
-
Hi @dusty-nv , amazing work. I am the creator of Nous Obsidian 3B and found out that NanoLLM supports it, so thank you very much!
I just dropped nanoLLaVA which is based on Qwen1.5-0.5B and SigLIP-…
-
Hi, I tried to train with siglip loss on a large dataset and I found that during training (not evaluation), CPU memory usage kept increasing. The program was finally killed by the system. The data loa…
-
报下面的错误
```bash
RuntimeError: weight should have at least three dimensions
Traceback (most recent call last):
File "/mnt/bn/arnold-ghh-test/mlx/users/guihonghao/playground/ghh_swift/swift/example…