mlfoundations/open_clip

Issues

• The result is random

  #862 opened 14 hours ago by qfmy
  2

• Evaluation of Trained SigLIP Checkpoints

  #824 opened 14 hours ago by work4cs
  1

• How to inference trained checkpoint on my own dataset?

  #882 opened 14 hours ago by guoyanan1g
  0

• CPU memory leak when training with CsvDataset-like dataset

  #849 opened 2 months ago by estherxue
  10

• CoCa RoBERTa Attention Map Size Issue

  #864 opened a month ago by sandeepmukh
  1

• Incorrect license appears on PyPI

  #798 opened 24 days ago by alhirzel
  0

• Long context CLIP

  #857 opened 24 days ago by nicolas-dufour
  3

• Issue with inference values with 'hf-hub:laion/CLIP-convnext_xxlarge-laion2B-s34B-b82K-augreg'

  #865 opened 24 days ago by dbugger29
  2

• Training speed slow

  #861 opened 24 days ago by lezhang7
  1

• [1,512] data

  #875 opened 24 days ago by lwtgithublwt
  1

• Model coca_ViT-B-32 not found; available models = ['RN50', 'RN50-quickgelu', 'RN101', 'RN101-quickgelu', 'RN50x4', 'RN50x16', 'RN50x64', 'ViT-B-32', 'ViT-B-32-quickgelu', 'ViT-B-16', 'ViT-L-14', 'ViT-L-14-336']

  #814 opened 24 days ago by narminGhaffari
  1

• Loss isn't converging

  #821 opened 24 days ago by zhaowenZhou
  1

• How to initialize the encoder to pretrain CLIP？

  #859 opened 24 days ago by shihuai
  1

• Will fine-tuning affect the ability of the original search?

  #826 opened 24 days ago by Daydaylight
  1

• How to extract 1024 width patch embeddings and CLS embedding

  #844 opened 2 months ago by alvaro-stylesage
  1

• A NotImplementedError occurred when I imported the pre-trained weights of RN50.

  #846 opened 24 days ago by FudaMingdaXsc
  0

• How much GPU memory can SigLIP save compared with CLIP?

  #825 opened 24 days ago by Pluto-Jin
  0

• unlocking layer groups in open_clip/src/open_clip /timm_model.py for fine-tuning

  #854 opened 24 days ago by mikelee-dev
  1

• What is the max value for unlocked_groups in TimmModel.lock()? I am trying to fine-tune the entire vision tower

  #866 opened 24 days ago by mikelee-dev
  1

• [1,512] Data conversion

  #868 opened 24 days ago by lwtgithublwt
  0

• Handling Negative Pairs in Fine-Tuning of CLIP Models

  #869 opened 24 days ago by doramasma
  1

• RuntimeError: Boolean value of Tensor with more than one value is ambiguous

  #847 opened 2 months ago by wujohns
  8

• Load hybrid-clip in open_clip

  #863 opened a month ago by code-ishwar
  0

• how to get hidden_state from every layers of ViT of openclip vision encoder?

  #827 opened 3 months ago by jzssz
  2

• S3 Broken Pipe Error

  #845 opened 2 months ago by samuelstevens
  10

• datasets for pretrained models

  #851 opened 2 months ago by mikelee-dev
  3

• On mismatch of model input, does not give correct list of valid ones

  #833 opened 2 months ago by ppbrown
  1

• Triton error in int8-support

  #835 opened 2 months ago by aleablu
  1

• Logging for cuda usage could be a little confused when specifying gpu-id

  #829 opened 2 months ago by Vikho
  1

• ```coca_base``` configuration does not have any ```pretrained``` arguments

  #837 opened 2 months ago by shyammarjit
  1

• How to load Clip-VIT-bigG-14 in local which has two .bin files and can not find any information how to load, thanks!!!!!!!

  #841 opened 2 months ago by hanggun
  2

• coca model get dimension mismatch error when dimension is 77

  #840 opened 3 months ago by Yufang-Liu
  2

• Question: How/where to evaluate GeoDE and Dollar Street?

  #839 opened 3 months ago by gregor-ge
  1

• Is there someone who can share implemented script for --lock-text-unlocked-layers for "Native Clip Model"?

  #830 opened 3 months ago by Vikho
  0

• Example of launcher script with Horovod

  #823 opened 3 months ago by guillaumeguy
  1

• oscillations in the loss

  #822 opened 3 months ago by nicolas-dufour
  2

• ValueError: Batch dimension of `input_ids` should be 0, but is 6.

  #816 opened 3 months ago by ShenZheng2000
  1

• How to reproduce the VQA results of CoCa?

  #819 opened 3 months ago by LixDemon
  0

• How does the logit_scale vary while training , i noitced that in my case it starts from the 14.28(1/0.07) and then just goes down and towards the end of the training it reaches 1

  #815 opened 4 months ago by Akshay1-6180
  2

• Model weights missing ViT-SO400M-14-SigLIP-384 (webli pretrained)

  #809 opened 4 months ago by anshudaur
  2

• What is impact on the image embedding result when the alpha channel is dropped by the converter

  #801 opened 4 months ago by githubusersel
  0

• demo code will cause memory leak

  #803 opened 4 months ago by ltebean
  1

• About width of the VIT-G model

  #802 opened 4 months ago by OliverHuang1220
  1

• Is there any particular reason why bias term is kept as False in the projection layers

  #807 opened 4 months ago by Akshay1-6180
  2

• Model Download fails on Mac, works on Linux

  #810 opened 4 months ago by thomasf1
  2

• A beginner's guide to fine-tuning the CLIP model for your downstream tasks using OpenClip

  #806 opened 4 months ago by ShuxunoO
  2

• Finetuning CLIP VIT32 on COCO captions

  #811 opened 4 months ago by MLRadfys
  0

• Q-LORA for BiomedCLIP

  #808 opened 4 months ago by Abhiram-kandiyana
  0

• Issue with --lock-text Parameter in Bash Script

  #805 opened 4 months ago by Lylinnnnn
  3

• Option to load just text or image models

  #799 opened 4 months ago by rbrtcs1
  1