mlfoundations/open_flamingo issues and pull requests

#306 - problem when load model, image_processor, tokenizer using `create_model_and_transforms()`

Issue - State: open - Opened by MrNquyen about 2 months ago

#305 - [FEATURE REQUEST] Enable Video Training

Issue - State: open - Opened by simplaj 3 months ago - 2 comments

#304 - The format of dataset of mmc4 and laion? So I can train on other dataset.

Issue - State: open - Opened by hhy150 4 months ago

#303 - [BUG] Cannot Initialize Any Models in transformers==4.41.2

Issue - State: open - Opened by JiahuiKChen 5 months ago - 1 comment
Labels: bug

#302 - The bug in getting attention weight

Issue - State: open - Opened by SjokerLily 5 months ago

#301 - what's the meaning of media in input to PerceiverResampler ? Why give time embedding to different media ? Shouldn't given time embedding to different frames?

Issue - State: open - Opened by Yang-bug-star 5 months ago

#300 - Unable to reproduce the evaluation results on textvqa

Issue - State: closed - Opened by woshidengweimo 6 months ago
Labels: bug

#299 - added gqa as eval dataset

Pull Request - State: open - Opened by olo126 6 months ago

#298 - Added GQA as eval dataset

Pull Request - State: closed - Opened by olo126 6 months ago

#297 - Implementation

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

#296 - how to get synthetic data mentioned in the paper

Issue - State: closed - Opened by lijinginfo 7 months ago

#295 - Allow `PerceiverResampler` to take in attention masks

Issue - State: closed - Opened by azshue 7 months ago

#294 - Failure to use https://huggingface.co/anas-awadalla/mpt-7b model downloading issue?

Issue - State: open - Opened by MRAWAY77 8 months ago - 1 comment
Labels: bug

#293 - [FEATURE REQUEST] Is it possible to train a 3B openflamingo on 4 V100 on one node with deepspeed by simplifying pre-training data used？

Issue - State: open - Opened by Yang-bug-star 8 months ago

#292 - How to try my own fine-tuning experiments

Issue - State: open - Opened by minghehe-nobug 8 months ago - 1 comment

#291 - [BUG] ViT-L-14 downloaded in Demo compared to Huggingface Demo Card don't seem to be the same

Issue - State: open - Opened by rphlstck 9 months ago
Labels: bug

#290 - TypeError: Flamingo.init() got an unexpected keyword argument 'cache_dir'

Issue - State: open - Opened by Mei0211 9 months ago - 1 comment
Labels: bug

#289 - Open Flamingo Perplexity Calculation

Issue - State: open - Opened by mustafaadogan 9 months ago - 3 comments

#288 - Understanding MP policy for parameters

Issue - State: open - Opened by rphlstck 10 months ago
Labels: bug

#287 - [BUG] an error in get_dataset_size about args.shards

Issue - State: open - Opened by ChrisZhangyu 10 months ago
Labels: bug

#286 - Worse performance compared to vanilla cross attention

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

#285 - How to set hyperparameters?

Issue - State: open - Opened by hzhz2020 11 months ago

#284 - Import Error

Issue - State: closed - Opened by Hambaobao 11 months ago - 5 comments
Labels: bug

#283 - train_utils media token fix

Pull Request - State: closed - Opened by olo126 12 months ago

#282 - media token fix

Pull Request - State: closed - Opened by olo126 12 months ago

#281 - Update requirements.txt

Pull Request - State: closed - Opened by sumeetghumare4 12 months ago - 2 comments

#280 - 'media_token_id' AttributeError

Issue - State: open - Opened by olo126 about 1 year ago
Labels: bug

#279 - GPU not distributed

Issue - State: closed - Opened by rafael-alencar29 about 1 year ago

#278 - Is context images really needed during few-shot inference?

Issue - State: open - Opened by zhang9302002 about 1 year ago

#277 - Unable to run on GPU with error: "Cannot copy out of meta tensor; no data!"

Issue - State: open - Opened by shomikj about 1 year ago - 2 comments
Labels: bug

#276 - Example for other pretrained language models

Issue - State: closed - Opened by shomikj about 1 year ago - 1 comment

#275 - fix for none past_key_values, getting supported tasks

Pull Request - State: closed - Opened by hannahyklee about 1 year ago

#274 - Will it be possible to train the whole model using the 7B LLM

Issue - State: closed - Opened by ElegantLin about 1 year ago - 9 comments

#273 - Add pre-commit hooks

Pull Request - State: closed - Opened by isaac-chung about 1 year ago - 3 comments

#272 - Create extra samples with surplus images

Pull Request - State: open - Opened by isaac-chung about 1 year ago - 11 comments

#271 - Regarding Our Paper Citing the open-flamingo Repository at NeurIPS 2023

Issue - State: closed - Opened by yongliang-wu about 1 year ago

#270 - Remove incorrect eos token label mask

Pull Request - State: closed - Opened by anas-awadalla about 1 year ago

#269 - [BUG] Train Utils Does Not Set `label=-100` for `eos_token_id`

Issue - State: closed - Opened by loris2222 about 1 year ago - 1 comment
Labels: bug

#268 - I'm having some problems initializing the open-Flamingo model

Issue - State: closed - Opened by JohnneyQin about 1 year ago - 1 comment
Labels: bug

#267 - Evaluation on ImageNet1k seems very slow.

Issue - State: closed - Opened by weivision about 1 year ago - 2 comments

#266 - Merge wilds mllm

Pull Request - State: open - Opened by i-gao about 1 year ago - 4 comments

#265 - Repeated answers in output.

Issue - State: open - Opened by essamsleiman about 1 year ago - 1 comment

#264 - Language stream changes

Pull Request - State: closed - Opened by anas-awadalla about 1 year ago - 1 comment

#263 - [BUG] Generating text is timing out

Issue - State: open - Opened by sejalgupta about 1 year ago - 1 comment
Labels: bug

#262 - add z-loss

Pull Request - State: closed - Opened by anas-awadalla about 1 year ago

#261 - Major refactor to support new architectures

Pull Request - State: open - Opened by i-gao about 1 year ago - 5 comments

#260 - Mismatch input type and weight type when training with precision fp16

Issue - State: open - Opened by hungvo304ml about 1 year ago - 7 comments
Labels: bug

#259 - Always reported OOM when train the model with LLama2

Issue - State: closed - Opened by RyanYip-Kat about 1 year ago - 2 comments
Labels: bug

#258 - Question about next version

Issue - State: closed - Opened by gray311 about 1 year ago - 2 comments

#257 - Does Flamingo support NLVR^2 and its paired-image input setting?

Issue - State: open - Opened by chenxshuo about 1 year ago - 3 comments

#256 - Support for Onnx format?

Issue - State: open - Opened by mhjort about 1 year ago - 1 comment

#255 - Deepspeed inference

Pull Request - State: closed - Opened by i-gao about 1 year ago

#254 - Deepspeed + embeddings fix

Pull Request - State: closed - Opened by anas-awadalla about 1 year ago

#253 - How to finetune all paramaters, including vision_encoder and original decoder layers?

Issue - State: closed - Opened by liyongqi67 about 1 year ago - 2 comments

#252 - Some questions about OpenFlamingo V1

Issue - State: closed - Opened by SjokerLily about 1 year ago - 3 comments

#251 - Question around Generated Output

Issue - State: closed - Opened by agneet42 about 1 year ago - 9 comments

#250 - Fix device_id in mmc4 pass

Pull Request - State: closed - Opened by anas-awadalla about 1 year ago

#249 - Unpin sentencepiece version

Pull Request - State: closed - Opened by anas-awadalla about 1 year ago

#248 - [Dependencies] Does `sentencepiece` need to be pinned to `0.1.98`?

Issue - State: closed - Opened by siddk about 1 year ago - 1 comment

#247 - About old blocks

Issue - State: open - Opened by liyongqi67 about 1 year ago

#246 - Questions about compatibility between model version 1 and version 2

Issue - State: closed - Opened by gray311 about 1 year ago - 4 comments

#245 - Add Support for Specifying Custom `cache_dir`

Pull Request - State: closed - Opened by siddk about 1 year ago - 1 comment

#244 - [BUG] You are using config.init_device='cpu', but you can also use config.init_device="meta" with Composer + FSDP for fast initialization.

Issue - State: closed - Opened by DarkAlchy about 1 year ago - 2 comments
Labels: bug

#243 - Multi-GPU training crashes with "Cannot copy out of meta tensor; no data"

Issue - State: open - Opened by rohan-mehta about 1 year ago
Labels: bug

#242 - Questions about mask from MaskedCrossAttention.forward method

Issue - State: closed - Opened by Seerkfang about 1 year ago

#241 - update setup and requirements

Pull Request - State: closed - Opened by loftusa over 1 year ago - 3 comments

#240 - dataset_flickr30k_coco_style.json

Issue - State: closed - Opened by liyongqi67 over 1 year ago - 1 comment

#240 - dataset_flickr30k_coco_style.json

Issue - State: closed - Opened by liyongqi67 over 1 year ago - 1 comment

#239 - Update README.md

Pull Request - State: closed - Opened by eltociear over 1 year ago - 1 comment

#238 - one-to-one image-sentence assignment for MMC4

Pull Request - State: closed - Opened by yassouali over 1 year ago

#237 - Can open_flamingo handle image classification tasks？

Issue - State: closed - Opened by RayL0221 over 1 year ago - 1 comment

#237 - Can open_flamingo handle image classification tasks？

Issue - State: closed - Opened by RayL0221 over 1 year ago - 1 comment

#236 - Pass **kwargs in "generate" function instead of passing each argument by name

Pull Request - State: closed - Opened by DavidMChan over 1 year ago - 3 comments

#236 - Pass **kwargs in "generate" function instead of passing each argument by name

Pull Request - State: closed - Opened by DavidMChan over 1 year ago - 3 comments

#235 - Update README.md evaluation numbers

Pull Request - State: closed - Opened by anas-awadalla over 1 year ago

#234 - How to set up multigpus?

Issue - State: closed - Opened by Vaxxacavier over 1 year ago - 4 comments

#233 - labels in MMC4 FORWARD PASS

Issue - State: closed - Opened by liyongqi67 over 1 year ago - 3 comments

#232 - fsdp Error report

Issue - State: open - Opened by liyongqi67 over 1 year ago - 3 comments

#231 - About the random selection during pre-training

Issue - State: open - Opened by vateye over 1 year ago - 6 comments

#230 - [FEATURE REQUEST] Reduce pip dependencies to inference-time requirements.

Issue - State: closed - Opened by DavidMChan over 1 year ago - 1 comment

#230 - [FEATURE REQUEST] Reduce pip dependencies to inference-time requirements.

Issue - State: closed - Opened by DavidMChan over 1 year ago - 1 comment

#229 - Update requirements.txt

Pull Request - State: closed - Opened by ElegantLin over 1 year ago - 1 comment

#229 - Update requirements.txt

Pull Request - State: closed - Opened by ElegantLin over 1 year ago - 1 comment

#228 - Use GPU instead of CPU

Issue - State: closed - Opened by gemek123 over 1 year ago - 4 comments

#227 - The context_text of zero-shot and the performance of one-shot.

Issue - State: closed - Opened by yongliang-wu over 1 year ago - 3 comments

#226 - Training loss curve on MMC4 dataset?

Issue - State: open - Opened by tonylins over 1 year ago - 8 comments

#226 - Training loss curve on MMC4 dataset?

Issue - State: open - Opened by tonylins over 1 year ago - 8 comments

#221 - Are there some format limitation of the training batches?

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

#221 - Are there some format limitation of the training batches?

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

#220 - Evaluation updates: add RICES + prompt ensembling

Pull Request - State: closed - Opened by i-gao over 1 year ago

#220 - Evaluation updates: add RICES + prompt ensembling

Pull Request - State: closed - Opened by i-gao over 1 year ago

#219 - Image-Sentence matching for MMC4 Dataset

Issue - State: closed - Opened by yassouali over 1 year ago - 4 comments

#217 - Does the OpenFlamingo based on MPT-7B have an Apache 2.0 license?

Issue - State: closed - Opened by alpayariyak over 1 year ago - 1 comment

#214 - fix atten from Long to Bool

Pull Request - State: closed - Opened by ElegantLin over 1 year ago - 7 comments

#214 - fix atten from Long to Bool

Pull Request - State: closed - Opened by ElegantLin over 1 year ago - 7 comments

#213 - Pin versions for language encoders (i.e. anas-awadalla/mpt-1b-redpajama-200b)

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

#205 - [FEATURE REQUEST] Support for OpenLLaMA

Issue - State: closed - Opened by alfredplpl over 1 year ago - 1 comment
Labels: wontfix

GitHub / mlfoundations/open_flamingo issues and pull requests