facebookresearch/TimeSformer issues and pull requests

#137 - training from scratch, loss decreased very slow

Issue - State: open - Opened by chrisx599 6 months ago - 1 comment

#136 - There is a typo in the file `timesformer/models/vit.py`.

Issue - State: open - Opened by pyqjw 7 months ago

#135 - Prediction per frame

Issue - State: open - Opened by DanaOsama 9 months ago

#134 - Allow for resume finetuning from a checkpoint

Pull Request - State: open - Opened by feemthan 11 months ago - 2 comments
Labels: CLA Signed

#133 - Finetuning code can be improved

Issue - State: closed - Opened by feemthan 11 months ago - 2 comments

#132 - Ok

Issue - State: open - Opened by Pshk1310136 about 1 year ago

#131 - How to reason about the video with the demo

Issue - State: open - Opened by xqc-qc over 1 year ago - 1 comment

#130 - What can I do if I can't find the train.csv file?

Issue - State: open - Opened by xqc-qc over 1 year ago

#129 - Accuracy of each category

Issue - State: open - Opened by subin2202 over 1 year ago - 3 comments

#128 - Train on AVA dataset

Issue - State: open - Opened by chenqqqqqqq over 1 year ago

#127 - Update resnet_helper.py

Pull Request - State: open - Opened by bdytx5 over 1 year ago
Labels: CLA Signed

#126 - Update vit_utils.py

Pull Request - State: open - Opened by bdytx5 over 1 year ago - 1 comment
Labels: CLA Signed

#125 - In "divided_space_time", Why is the class token not used during temporal attention?

Issue - State: open - Opened by Backdrop9019 almost 2 years ago - 1 comment

#124 - Just got crash while running demo while working in jupyter notebook

Issue - State: open - Opened by XinyueZ almost 2 years ago

#123 - 503 Service Temporarily Unavailable, Something-Something V2

Issue - State: open - Opened by XinyueZ almost 2 years ago

#122 - Video preprocessing to apply for trained model

Issue - State: open - Opened by sakethattaluri about 2 years ago - 1 comment

#121 - How to apply this model for regression

Issue - State: open - Opened by Taimoor-R about 2 years ago - 1 comment

#120 - During training, why is optimizer step called after 32 iterations?

Issue - State: open - Opened by karthik1145 over 2 years ago

#119 - How to train driving dataset on timesformer?

Issue - State: open - Opened by pgupta119 over 2 years ago - 1 comment

#118 - Why is the output 400 instead of 768？

Issue - State: open - Opened by lamsongianm over 2 years ago

#117 - Question about train time

Issue - State: open - Opened by shiyi-z over 2 years ago - 1 comment

#116 - How to load pretrained SlowFast Model?

Issue - State: open - Opened by HashmatShadab over 2 years ago

#115 - Training at epoch 1 end up with CUDA error and Assertion `t >= 0 && t < n_classes` failed

Issue - State: open - Opened by kct22aws over 2 years ago - 2 comments

#114 - Kinetic dataset labels

Issue - State: open - Opened by kct22aws over 2 years ago - 2 comments

#113 - Kinetics dataset preparation

Issue - State: open - Opened by kct22aws over 2 years ago - 3 comments

#112 - How to train only the head？

Issue - State: open - Opened by lws6885 over 2 years ago

#111 - RuntimeError: Sizes of tensors must match except in dimension 0. Got 192 and 12 (The offending index is 0)

Issue - State: open - Opened by 906070493 almost 3 years ago

#110 - Training issue

Issue - State: open - Opened by RebelliousTT almost 3 years ago - 1 comment

#109 - problem about default_cfgs

Issue - State: closed - Opened by cser-liu almost 3 years ago

#108 - What exactly does cfg.TASK mean？

Issue - State: open - Opened by mudexuyan almost 3 years ago - 2 comments

#107 - How to set "find_unused_parameters=True"

Issue - State: closed - Opened by ppriyank almost 3 years ago - 1 comment

#106 - Attention Rollout for TimeSformer

Issue - State: open - Opened by yiyixuxu almost 3 years ago - 1 comment

#105 - what is the video_id id the train val csv file

Issue - State: open - Opened by yihui8776 almost 3 years ago

#104 - Training on custom dataset

Issue - State: open - Opened by Jeba-create almost 3 years ago

#103 - is multi-view testing not actually implemented for ssv2?

Issue - State: open - Opened by yiyixuxu almost 3 years ago

#102 - Inference code reg

Issue - State: open - Opened by Jeba-create almost 3 years ago - 2 comments

#101 - Pre-trained models on ImageNet1k and 21k

Issue - State: closed - Opened by wp2333 almost 3 years ago

#100 - RotaryEmbedding

Issue - State: open - Opened by xieanheng almost 3 years ago - 1 comment

#99 - Modified AVA dataset problem

Issue - State: open - Opened by MitchellWei almost 3 years ago - 2 comments

#98 - batchnorm and visualization

Issue - State: open - Opened by HC-2016 almost 3 years ago - 1 comment

#97 - question

Issue - State: closed - Opened by hbrachemi almost 3 years ago

#96 - ImageNet-1K Models

Issue - State: open - Opened by owaisCS almost 3 years ago - 1 comment

#95 - CUDA error: CUBLAS_STATUS_EXECUTION_FAILED

Issue - State: closed - Opened by ardofski almost 3 years ago - 1 comment

#94 - Has anyone managed to reproduce the TimeSformer (the basic model) results on K400?

Issue - State: open - Opened by kami93 almost 3 years ago - 2 comments

#93 - Difference between Space and Joint Space-Time

Issue - State: open - Opened by himidev about 3 years ago - 2 comments

#92 - Add benchmark code for TimeSformer

Pull Request - State: closed - Opened by HydrogenSulfate about 3 years ago - 1 comment

#91 - How should i test this model

Issue - State: open - Opened by happyBroJk about 3 years ago

#90 - That is a suggestion.

Issue - State: open - Opened by happyBroJk about 3 years ago - 1 comment

#89 - influence of depth?

Issue - State: open - Opened by fanbooo about 3 years ago - 1 comment

#88 - Fine-tuning TimeSformer on a smaller dataset

Issue - State: closed - Opened by mtv21 about 3 years ago - 3 comments

#87 - K600 labels?

Issue - State: closed - Opened by cnut1648 about 3 years ago - 2 comments

#86 - input clips

Issue - State: open - Opened by W4ngH4o over 3 years ago - 2 comments

#85 - How to get Video level features rather than clip features

Issue - State: open - Opened by mayankand over 3 years ago - 1 comment

#84 - Pre-trained model output prediction different from testing output prediction

Issue - State: closed - Opened by chrisdt1998 over 3 years ago - 1 comment

#83 - Question about run_net.py

Issue - State: open - Opened by wssxjtu over 3 years ago - 3 comments

#82 - checkpoints of joint attention model

Issue - State: open - Opened by wangjk666 over 3 years ago - 1 comment

#81 - Training Time Required

Issue - State: closed - Opened by kevaldoshi17 over 3 years ago - 4 comments

#80 - Training on custom dataset in Something-Something V2 format

Issue - State: open - Opened by parthplc over 3 years ago - 3 comments

#79 - Dev benchmark

Pull Request - State: closed - Opened by HydrogenSulfate over 3 years ago - 1 comment

#78 - questions about the FLOPs

Issue - State: open - Opened by wangjk666 over 3 years ago - 1 comment

#77 - No module named 'timm'

Issue - State: closed - Opened by 87nohigher over 3 years ago - 2 comments

#76 - Spatial and temporal attention calculation

Issue - State: closed - Opened by y8han over 3 years ago - 3 comments

#75 - DATA.PATH_TO_DATA_DIR contains all data related files or only mp4s?

Issue - State: closed - Opened by parthplc over 3 years ago - 1 comment

#74 - questions about the implementation of the "divided_space_time" attn

Issue - State: open - Opened by HC-2016 over 3 years ago - 1 comment

#73 - How to use output features

Issue - State: open - Opened by Victor-Gump over 3 years ago - 1 comment

#72 - Installation "Please Share your env.yml"

Issue - State: open - Opened by menatallh over 3 years ago - 1 comment

#71 - How to normalize video frames?

Issue - State: open - Opened by augustasm over 3 years ago - 1 comment

#70 - Label

Issue - State: open - Opened by AItechnology over 3 years ago - 2 comments

#69 - Add num_frames parameter when calling load_pretrained

Pull Request - State: closed - Opened by sathvikask0 over 3 years ago - 1 comment

#68 - problem of load_pretrained function

Issue - State: open - Opened by Christinepan881 over 3 years ago - 1 comment

#67 - HowTo100M labels file

Issue - State: open - Opened by parthplc over 3 years ago - 2 comments

#66 - Initialization of time_embed in VisionTransformer

Issue - State: closed - Opened by kenshohara over 3 years ago - 2 comments

#65 - visualization of the attention

Issue - State: open - Opened by Holmes-Alan over 3 years ago - 1 comment

#64 - Can someone share the something-something v2?

Issue - State: closed - Opened by kami93 over 3 years ago - 2 comments

#63 - ImportError: cannot import name '_LinearWithBias' from 'torch.nn.modules.linear'

Issue - State: open - Opened by weiwenying over 3 years ago - 7 comments

#62 - Diving-48 dataset config file

Issue - State: closed - Opened by eladb3 over 3 years ago - 2 comments

#61 - Run test_net.py ERROR!

Issue - State: closed - Opened by wsh-nie over 3 years ago - 2 comments

#60 - rectangular size input on transfer learning

Issue - State: open - Opened by chris-tkinter over 3 years ago - 1 comment

#59 - Quesions about the learning rate for the 4GPUS.

Issue - State: open - Opened by justimyhxu over 3 years ago - 3 comments

#58 - Visualization of activation patterns

Issue - State: open - Opened by shinjinishimoto over 3 years ago - 1 comment

#57 - Regarding attention visualisation using Attention Rollout scheme.

Issue - State: closed - Opened by victor-psiori over 3 years ago - 16 comments

#56 - Issue loading kinetics data

Issue - State: closed - Opened by kahnchana over 3 years ago - 2 comments

#55 - Question about Optimizer SGD, why?

Issue - State: closed - Opened by Joeycho over 3 years ago - 1 comment

#54 - easy to over-fit

Issue - State: open - Opened by Note-Liu over 3 years ago

#53 - My inference too slow

Issue - State: closed - Opened by Note-Liu over 3 years ago - 2 comments

#52 - PyTorch Version Issue

Issue - State: open - Opened by JustinYuu over 3 years ago - 4 comments

#51 - How flexible TimeSformer model is for tiny dataset(1 or 2 training videos)

Issue - State: closed - Opened by Joeycho over 3 years ago - 3 comments

#50 - If the test video is too short

Issue - State: closed - Opened by Note-Liu over 3 years ago - 2 comments

#49 - how to use with test demo?

Issue - State: closed - Opened by thunderlzy over 3 years ago - 3 comments

#48 - CUDA version, which version was tested for TimeSformer?

Issue - State: open - Opened by Joeycho over 3 years ago - 3 comments

#47 - Update code for pytorch 1.9.0

Pull Request - State: open - Opened by wanchichen over 3 years ago - 10 comments
Labels: CLA Signed

#46 - Train on different Dataset(HVU)

Issue - State: closed - Opened by Joeycho over 3 years ago - 2 comments

#45 - How to output the pred and label of each test video

Issue - State: open - Opened by vvsg over 3 years ago - 2 comments

#44 - pretrained weights of TimeSformerHR-K600 only 8frames?

Issue - State: open - Opened by ygfrancois over 3 years ago - 5 comments

#43 - Writing probabilites rather than labels with TEST.SAVE_RESULTS_PATH?

Issue - State: closed - Opened by balaganeshmohan over 3 years ago - 2 comments

#42 - Can I train Kinetics600 with four GPUs(V100)?

Issue - State: closed - Opened by Note-Liu over 3 years ago - 2 comments

#37 - Inconsistent result on k400

Issue - State: open - Opened by wenyuqing over 3 years ago - 4 comments

#35 - Error When Training the Default TimeSformer

Issue - State: open - Opened by shahd-seddik over 3 years ago - 2 comments

#31 - How to extract only features of the pretrained TimeSformer model given a video. Thanks in advance..

Issue - State: open - Opened by Wajahat0 almost 4 years ago - 2 comments

#28 - Failed to meta load video idx 46523 from fmSfLXRLHug; trial 0

Issue - State: closed - Opened by wenyuqing almost 4 years ago - 4 comments

GitHub / facebookresearch/TimeSformer issues and pull requests