-
**Describe the bug**
MiniCPM-V-2 lora微调后推理报错: AssertionError: Current sentence length exceeds the model max_length: 4096
![image](https://github.com/user-attachments/assets/1b38f68e-92a5-4620-ae95-d…
-
RT
-
Hi, first of all, thank you for the amazing work! This is more of a curiosity question, will this repository also extend to ColPali/ColQwen as well?
-
I'm trying to run the VLM example that uses SGLang library - https://modal.com/docs/examples/sgl_vlm -
modal run 06_gpu_and_ml/llm-serving/sgl_vlm.py from modal-examples.
I get the following error …
-
### Issue Description
when running server llava1.6-mistral - unable to load model
nexa server llava1.6-vicuna
objc[12203]: Class GGMLMetalClass is implemented in both /Applications/Nexa.app/Content…
-
```shell
09/10 [07:43:21] INFO | >> [*] Loading from local path `/code/Basemodel/ML-Mamba` …
-
Hi,
I love your project. Could you please provide some benchmark (accuracy, f1, etc) about the latest pretrained model in the release?
That will be a great help because I also want to train some…
-
Hi,
Is EXTERNAL keyword supported by asn1c ?
The compiler does not generates error when see EXTERNAL but in generated code there is not corresponding file to handle that (EXTERNAL.h) and c compiler …
-
对minicpm-V2.6模型,我设置了--sft_type full用来进行全量微调,相比于--sft_type lora不经显存使用增加很多,而且速度上也变慢了很多,但是在生成的sft_args.json里面,--sft_type full是没问题的,但是仍然存在"target_modules": "^(llm|resampler)(?!.*(lm_head|output|emb|wte|sh…
-
step 0: install nexa-sdk on M3 Macbook16
step 1: nexa run -st nanoLLaVA:model-fp16
step 2: upload an image and ask 'please describe the image.'
step 2: report error in cli:
```
Traceback (most re…