mlfoundations/open_clip

Issues

• ❓ [Question] Can't reproduce imagenet results of RN50 model trained on `pixparse/cc3m-wds`

  #930 opened 2 months ago by clownrat6
  1

• RuntimeError: Mask shape should match input shape

  #954 opened 2 months ago by MINXIANWEN
  1

• Intermediate checkpoints with ResNet backbone

  #1009 opened 2 months ago by ndrsn0208
  1

• [AssertionError] when trying to inference model

  #1011 opened 2 months ago by bright-arparwut
  3

• Training stuck on first epoch

  #1016 opened 2 months ago by alexisdrakopoulos
  12

• error when use try to use int8 operations with OpenCLIP.

  #1012 opened 2 months ago by xiaohoua
  1

• AttributeError: module 'triton.language' has no attribute 'libdevice', when using method "convert_int8_model_to_inference_mode()"

  #1010 opened 3 months ago by bright-arparwut
  1

• Dataset's mean and std

  #1007 opened 3 months ago by xiaohoua
  1

• Cannot train again on pretrained checkpoint due to change in default `weights_only=True`

  #998 opened 3 months ago by ishaaq
  9

• ZERO-SHOT Classification On SUN397

  #1005 opened 3 months ago by xiaohoua
  0

• how to update the position_embedding from embedding to rope position？

  #997 opened 3 months ago by Johnson-yue
  1

• Where can I get the LAION-80M dataset?

  #1003 opened 3 months ago by XxFChen
  0

• is it a bug while computing contrastive loss?

  #1002 opened 3 months ago by gaofei
  1

• ViT-L-14-336 fine-tuning failed

  #999 opened 4 months ago by zhaozhipeng1997
  1

• Unable to load model in offline mode using local files

  #968 opened 4 months ago by yarden4998
  5

• Fine tune arguments to learn new knowledge without forgetting previous

  #973 opened 4 months ago by SutirthaChakraborty
  1

• I wonder how to extract patch/local features in 768 dim of CoCa for downstream tasks? Should I use the attn_pool (for caption) to get (256,768)?

  #991 opened 4 months ago by Arsiuuu
  0

• could you please give me some advice on how to read directory or a lot of tar, thanks!!!

  #992 opened 4 months ago by leo23ui
  1

• Clarification on using --train-num-samples (lower value) without --dataset-resampled

  #993 opened 4 months ago by fadamsyah
  6

• Hello, I divided yfcc15m into 16 parts, but dataset download too many files, compression failure , could you please give me some advice on how to solve this!! thanks

  #988 opened 4 months ago by leo23ui
  0

• Fine-tune ViT Models on Higher Resolution Images

  #985 opened 4 months ago by D0miH
  1

• Issue with all_gather

  #980 opened 4 months ago by scopello
  3

• The model and pretraining parameters do not match.

  #983 opened 4 months ago by q664171689
  2

• Inference speed

  #981 opened 4 months ago by tppqt
  1

• SigLIP attention mask

  #984 opened 4 months ago by chs20
  1

• SigLip memory consumption increases as we scale number of GPUs

  #942 opened 5 months ago by khalidsaifullaah
  5

• RuntimeError: The shape of the 2D attn_mask is torch.Size([77, 77]), but should be (4, 4)

  #937 opened 6 months ago by JaspinXu
  2

• Error loading ViT-L-14-quickgelu (metaclip_fullcc) model with version v2.27.0+

  #966 opened 5 months ago by aivarasbaranauskas
  12

• segmentation fault

  #931 opened 5 months ago by vadim0x60
  2

• Fix torch.load weights_only FutureWarning

  #928 opened 5 months ago by johnbradley
  2

• Conflict between webdataset implementation and OpenCLIP

  #935 opened 5 months ago by wanghao14
  8

• Does open_clip support add_tokens?

  #961 opened 5 months ago by HyelinNAM
  2

• How to subdivide the same category, for example, how to distinguish Persian cats, coffee cats, and jingle cats, which are all cat categories?

  #962 opened 5 months ago by watertianyi
  2

• Apply scale loss when performing accum_freq

  #957 opened 5 months ago by AshStuff
  1

• Couldnot find MobileCLIP-S0

  #958 opened 5 months ago by SutirthaChakraborty
  1

• Question about fine-tuning

  #941 opened 5 months ago by jimmyparadm
  1

• when I finetune CLIP_ViT_L_14 model , Logit Scale is decrease from 100.0 to 95. and keep going , is right?

  #948 opened 5 months ago by Johnson-yue
  0

• Separately Optimizing CLIP Image and Text Encoders with Different Loss Functions

  #946 opened 5 months ago by omrisuissabrown
  0

• Questions regarding the implementation of SigLIP loss

  #944 opened 5 months ago by binwang777
  1

• Does the “ViT-L/14” support conversion to TensorRT parameter files?

  #947 opened 5 months ago by gyd-a
  0

• Fine Tune for emotion

  #945 opened 5 months ago by SutirthaChakraborty
  0

• Question about the SigLipTokenizer

  #940 opened 6 months ago by LuFan31
  2

• Question about the SigLipTokenizer

  #939 opened 6 months ago by LuFan31
  0

• Is it possible to load only the visual encoder or the text encoder?

  #938 opened 6 months ago by zhangyx1998
  2

• Load model is Error

  #934 opened 6 months ago by ROC-Star
  2

• Inconsistent performance on pretrained checkpoints of the same architecture from different sources

  #936 opened 6 months ago by bdevnani3
  1

• Any Plan for direct support of parquet dataset?

  #933 opened 6 months ago by cxxgtxy
  2

• training details for convnext_large_d.laion2B-s26B-b102K-augreg

  #932 opened 6 months ago by mumu-mu
  1

• How to fine tune open clip?

  #922 opened 7 months ago by capricixhk
  0

• How to persist int8 model?

  #921 opened 7 months ago by EdenChen233
  0