Ecosyste.ms: Issues
An open API service for providing issue and pull request metadata for open source projects.
GitHub / mertyg/vision-language-models-are-bows issues and pull requests
#40 - Performance Recall@1 COCO (table 6)
Issue -
State: open - Opened by FiorenzoParascandolo1 4 months ago
#39 - Does NegCLIP now support mulit-gpu and grad_accumulation?
Issue -
State: open - Opened by Vicent0205 7 months ago
#38 - A question on results from Figure 2 and bag-or-wordness
Issue -
State: open - Opened by iburenko 10 months ago
#37 - Why does `visual_genome_relation.json` still contain symmetric relations?
Issue -
State: closed - Opened by lcxrocks 11 months ago
- 4 comments
#36 - Question regarding numbers in Figure 1
Issue -
State: closed - Opened by YunYunY 12 months ago
- 1 comment
#35 - Evaluation bug when using GELU vs QuickGELU -- changes the results for some benchmarks
Issue -
State: open - Opened by bryant1410 about 1 year ago
- 1 comment
#34 - Could you share the fine-tuned fiber checkpoint with EqSim constraints?
Issue -
State: closed - Opened by ytaek-oh about 1 year ago
- 1 comment
#33 - Questions on evaluation results
Issue -
State: open - Opened by ytaek-oh about 1 year ago
- 1 comment
#32 - about the performance of originial CLIP
Issue -
State: closed - Opened by hiker-lw about 1 year ago
- 16 comments
#31 - How you run expreriments with batch size 1024 on a Single RTX-2080Ti
Issue -
State: closed - Opened by wujianP over 1 year ago
- 1 comment
#30 - I cannot run on RTX 3060 with batch-size=256!
Issue -
State: closed - Opened by shuguang99 over 1 year ago
- 2 comments
#29 - I can't reproduce Table 6
Issue -
State: closed - Opened by shuguang99 over 1 year ago
- 2 comments
#28 - parameter file problem
Issue -
State: closed - Opened by haoshuai714 over 1 year ago
- 2 comments
#27 - train negCLIP result problem
Issue -
State: closed - Opened by haoshuai714 over 1 year ago
- 10 comments
#26 - Similarity scores for NegCLIP are pretty similar
Issue -
State: closed - Opened by kochsebastian over 1 year ago
#25 - Model weights of regular COCO finetuning.
Issue -
State: open - Opened by wildphoton over 1 year ago
#24 - Flava image preprocessing
Issue -
State: closed - Opened by DianeBouchacourt almost 2 years ago
- 8 comments
#23 - slow evaluation for xvlm
Issue -
State: closed - Opened by lezhang7 almost 2 years ago
- 1 comment
#22 - Projections W_i and W_t
Issue -
State: closed - Opened by DianeBouchacourt almost 2 years ago
- 7 comments
#21 - Where to find the training data of NegCLIP?
Issue -
State: closed - Opened by yu-wyatt-wu almost 2 years ago
- 2 comments
#20 - dataset size of flickr and coco order datasets
Issue -
State: closed - Opened by HarmanDotpy almost 2 years ago
- 4 comments
#19 - matrix size for contrastive learning in model training
Issue -
State: closed - Opened by Lycus99 almost 2 years ago
- 5 comments
#18 - Table 6, COCO and Flickr Image/Text R@1 results
Issue -
State: closed - Opened by HarmanDotpy almost 2 years ago
- 8 comments
#17 - Calling model.eval() when computing scores otherwise non-deterministic results (torch._no_grad_() is not enough)
Issue -
State: closed - Opened by DianeBouchacourt almost 2 years ago
- 1 comment
#16 - Requirements (e.g. torch versions)
Issue -
State: closed - Opened by DianeBouchacourt almost 2 years ago
- 4 comments
#15 - Questions on BLIP score computation
Issue -
State: closed - Opened by DianeBouchacourt almost 2 years ago
- 3 comments
#14 - Fix a bit typo README
Pull Request -
State: open - Opened by guspan-tanadi almost 2 years ago
#13 - Code for generating negatives for training NegCLIP
Issue -
State: closed - Opened by HarmanDotpy almost 2 years ago
- 2 comments
#12 - fast
Pull Request -
State: closed - Opened by vinid almost 2 years ago
#11 - eval coco order and flickr order
Issue -
State: closed - Opened by lezhang7 almost 2 years ago
- 3 comments
#10 - Concrete benchmark results of attributes understanding
Issue -
State: closed - Opened by Yangyi-Chen almost 2 years ago
- 5 comments
#9 - mismatching results on compositional task
Issue -
State: closed - Opened by lezhang7 almost 2 years ago
- 25 comments
#8 - question about VG-Relation categories
Issue -
State: closed - Opened by hiker-lw almost 2 years ago
- 4 comments
#7 - Models are not in eval() mode.
Issue -
State: closed - Opened by linzhiqiu almost 2 years ago
- 1 comment
#6 - Will the CoCo-order and Flickr-order dataset be released?
Issue -
State: closed - Opened by linzhiqiu almost 2 years ago
- 2 comments
#5 - why concat df to all_df?
Issue -
State: closed - Opened by lezhang7 almost 2 years ago
- 1 comment
#4 - Exact hyperparameters for NegCLIP training. and question about imagenet accuracy reported in the paper
Issue -
State: closed - Opened by HarmanDotpy almost 2 years ago
- 20 comments
#3 - Can you release the dataset during the finetuning of NegCLIP? Thanks!
Issue -
State: closed - Opened by mu-cai almost 2 years ago
- 5 comments
#2 - When can you provide code and dataset?
Issue -
State: closed - Opened by BigHyf almost 2 years ago
- 5 comments
#1 - Thanks for your great work! When are you planning to share the code?
Issue -
State: closed - Opened by mu-cai about 2 years ago
- 6 comments