mlfoundations / open_flamingo

An open-source framework for training large multimodal models.

mlfoundations/open_flamingo Issues

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?
Updated 2 months ago
Unable to reproduce the evaluation results on textvqa
Closed 2 months ago
How to try my own fine-tuning experiments
Updated 3 months ago1
Implementation
Updated 3 months ago
how to get synthetic data mentioned in the paper
Closed 3 months ago
Allow `PerceiverResampler` to take in attention masks
Closed 4 months ago
TypeError: Flamingo.__init__() got an unexpected keyword argument 'cache_dir'
Updated 4 months ago1
Failure to use https://huggingface.co/anas-awadalla/mpt-7b model downloading issue?
Updated 4 months ago
[FEATURE REQUEST] Is it possible to train a 3B openflamingo on 4 V100 on one node with deepspeed by simplifying pre-training data used？
Updated 4 months ago
Open Flamingo Perplexity Calculation
Updated 4 months ago3
[BUG] ViT-L-14 downloaded in Demo compared to Huggingface Demo Card don't seem to be the same
Updated 5 months ago
Worse performance compared to vanilla cross attention
Closed 6 months ago2
Understanding MP policy for parameters
Updated 6 months ago
Unable to run on GPU with error: "Cannot copy out of meta tensor; no data!"
Updated 7 months ago2
[BUG] an error in get_dataset_size about args.shards
Updated 7 months ago
How to set hyperparameters?
Updated 7 months ago
Import Error
Closed 7 months ago5
'media_token_id' AttributeError
Updated 8 months ago
GPU not distributed
Closed 9 months ago
Will it be possible to train the whole model using the 7B LLM
Closed 9 months ago9
Is context images really needed during few-shot inference?
Updated 9 months ago
Example for other pretrained language models
Closed 9 months ago1
Some questions about OpenFlamingo V1
Closed 10 months ago3
Regarding Our Paper Citing the open-flamingo Repository at NeurIPS 2023
Closed 10 months ago
About the random selection during pre-training
Updated 10 months ago6
Does Flamingo support NLVR^2 and its paired-image input setting?
Updated 10 months ago3
[BUG] Train Utils Does Not Set `label=-100` for `eos_token_id`
Closed 10 months ago1
I'm having some problems initializing the open-Flamingo model
Closed 10 months ago1
Evaluation on ImageNet1k seems very slow.
Closed 10 months ago2
Repeated answers in output.
Updated 10 months ago1
Question around Generated Output
Closed a year ago9
Mismatch input type and weight type when training with precision fp16
Updated 10 months ago5
[BUG] Generating text is timing out
Updated 10 months ago1
Support for Onnx format?
Updated 10 months ago1
fsdp Error report
Updated 10 months ago3
Always reported OOM when train the model with LLama2
Closed a year ago2
Question about next version
Closed a year ago2
How to finetune all paramaters, including vision_encoder and original decoder layers?
Closed a year ago2
[Dependencies] Does `sentencepiece` need to be pinned to `0.1.98`?
Closed a year ago1
About old blocks
Updated a year ago
Questions about compatibility between model version 1 and version 2
Closed a year ago4
How to set up multigpus?
Closed a year ago4
[BUG] You are using config.init_device='cpu', but you can also use config.init_device="meta" with Composer + FSDP for fast initialization.
Closed a year ago1
Use GPU instead of CPU
Closed a year ago3
Multi-GPU training crashes with "Cannot copy out of meta tensor; no data"
Updated a year ago
Questions about mask from MaskedCrossAttention.forward method
Closed a year ago
[FEATURE REQUEST] Reduce pip dependencies to inference-time requirements.
Closed a year ago1
Can open_flamingo handle image classification tasks？
Closed a year ago1
dataset_flickr30k_coco_style.json
Closed a year ago1
labels in MMC4 FORWARD PASS
Closed a year ago3