shikras shikra issues - Githubissues

shikras / shikra

Other

734 stars 46 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

python mllm/demo/client.py , get answer : noreferrer

#69 ovjust opened 2 weeks ago
0
Error in loading 8bit

#68 zhoustan opened 4 weeks ago
0
anyone can add versions in requirements.txt?

#67 ovjust opened 2 months ago
0
Fixed a serious bug: input_embeds overwrited by image features

#66 lixit closed 2 months ago
0
anyone run success? come and chat

#65 ovjust opened 3 months ago
0
which version of llama should download? can you give me a download url?

#64 ovjust opened 3 months ago
1
Pull Request

#63 Ommos92 closed 6 months ago
2
How to implement cross-modal referencing?

#62 hzdzkjdxyjs opened 6 months ago
2
Can the LLama model here be replaced with GPT-4?

#61 PlutoXN opened 6 months ago
0
Custom Dataset Creation

#60 imr555 opened 7 months ago
0
train on a single dataset

#59 CYF2000127 opened 8 months ago
1
When I run the example in the demo, I get the error.

#58 qinbaigao opened 9 months ago
0
ValueError: Trying to set a tensor of shape torch.Size([32003, 4096]) in "weight" (which has shape torch.Size([32000, 4096])), this look incorrect.

#57 qinbaigao closed 9 months ago
2
Requirements for the demo

#56 roy651 opened 9 months ago
1
When I was creating a demo using "Python mllm/demo/webdemo. py -- model_path/path/to/shikra/ckpt", the following error occurred.

#55 yubo97 opened 10 months ago
4
TypeError: Image.__init__() got an unexpected keyword argument 'source'

#54 WANGSHAOXIA11 opened 10 months ago
2
question about detection

#53 an1018 opened 10 months ago
0
How to make the model accurately output the dimensions of bbox and point

#52 ImmortalSdm opened 11 months ago
0
training detail

#51 yeonju7kim opened 11 months ago
1
Wrong output when the inference stage

#50 Yiveen opened 11 months ago
1
Inconsistent performance on REC task

#49 ZhanYang-nwpu opened 11 months ago
4
Could you share your prompt or code to generate QA data from GPT4

#48 double-fire-0 opened 1 year ago
0
training on 8 V100 is too slow, shikra_pretrain_final19_stage2 nearly 800h。 Does anyone have a similar situation?

#47 Anymake opened 1 year ago
0
I have collected the download addresses for all the training data and posted them here for others to download conveniently.

#46 Anymake opened 1 year ago
4
About accelerate config

#45 jun0wanan opened 1 year ago
0
Inconsistent performance on MMBench

#44 scenarios opened 1 year ago
1
it seems like you use llava model, I want to know when training model, do you add the position information like " the cat is at [0.2, 0.2, 0.5, 0.5], or without any position information in training?

#43 wjfwjfwjf opened 1 year ago
0
Question about coordinate numerical representation

#42 gray311 opened 1 year ago
1
An NCCL RuntimeError occurred when saving the model

#41 Lanxin1011 opened 1 year ago
0
Shikra-RD

#40 yuntaodu opened 1 year ago
0
question: How to get output for a single image

#39 yuntaodu closed 1 year ago
1
TypeError: cfg should be a dict, ConfigDict or Config, but got <class 'NoneType'>

#38 hangzeli08 opened 1 year ago
0
Question about the training parameter setting at stage1 and stage2

#37 Lanxin1011 opened 1 year ago
3
Evaluation on PointQA, VQAv2, OK-VQA and Captioning

#36 ShramanPramanick opened 1 year ago
3
Question about the training init weight

#35 Lanxin1011 closed 1 year ago
1
Web demo output is weird

#34 1049451037 opened 1 year ago
7
Online demo not working

#33 1049451037 closed 1 year ago
1
Could you provide more information about the instruction to GPT4 when generating Shikra-RD "cot_with_ans" data?

#32 Lanxin1011 opened 1 year ago
2
Could you please share the prompt for GPT4?

#31 aixiaodewugege opened 1 year ago
0
May I ask for advice on process_conv_multimage in single_image_convsation.py?What does it for? Can it handle multiple images?

#30 hangzeli08 closed 1 year ago
1
RuntimeError: Internal: unk is not define

#29 Lanxin1011 closed 1 year ago
1
gqa_scene_graph_index.json

#28 Morizhaoyang closed 1 year ago
1
why can't I get the right answer？

#27 wowhahad opened 1 year ago
3
May I ask if it supports multi image input, that is, multiple images and one text

#26 hangzeli08 closed 1 year ago
4
what's the image preprocess method?

#25 double-fire-0 closed 1 year ago
2
Doubts about Training Commands: Inconsistency between the Ratio in the Two-Stage Training

#24 Dongshengjiang opened 1 year ago
2
What's the definition of shikra model's generation function and how to download the transformers version provided in requirements.txt?

#23 BellXP opened 1 year ago
1
infer problem: ModuleNotFoundError: No module named 'petrel_client'

#22 SeeeeShiwei closed 1 year ago
1
Cuda memory requirement

#21 wanghao-cst opened 1 year ago
11
How is the toy shikra trained in Table 2?

#20 qumengxue closed 1 year ago
2