ManifoldRG NEKO issues - Githubissues

ManifoldRG / NEKO

In Progress Implementation of GATO style Generalist Multimodal model capable of image, text, RL and Robotics tasks

https://discord.gg/brsPnzNd8h

GNU General Public License v3.0

38 stars 9 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

We're not skipping empty text samples in TextTask.sample_batch.

#92 eihli closed 3 weeks ago
1
Test the codebase with timm vit

#91 henryj18 opened 1 month ago
0
Fix perplexity and bring text modality to SOTA possible

#90 bhavul opened 1 month ago
0
Fix divide by 0 error in scheduler

#89 eihli opened 1 month ago
0
Experiment with various Vision Transformers (ViT)

#88 henryj18 opened 1 month ago
0
Experimenting with TextTask training on Penn Treebank.

#87 eihli closed 1 month ago
1
Fix vqa loss

#86 henryj18 opened 2 months ago
0
Fix the deprecation of the usage Accelerator, add code for new way

#85 henryj18 closed 1 month ago
1
Remove token length restriction on vqa eval

#84 eihli opened 3 months ago
0
Possible use of padding instead of max tokens to avoid error when calculating loss.

#83 eihli opened 3 months ago
0
`deterministic=[False|True]` might not be working.

#82 eihli closed 3 months ago
1
Add vqa playground notebook

#81 eihli opened 3 months ago
1
Fix issues with distributed training

#80 eihli closed 2 months ago
1
Issue 73: Add Debug Logging

#79 maupind opened 4 months ago
5
The training loss and perplexity is not decreasing for the VQA task

#78 henryj18 opened 4 months ago
0
Merge add_vqa to master

#77 henryj18 closed 4 months ago
0
Remove old Neko and Forge branding

#76 harshsikka opened 4 months ago
0
Issue 64: Remove build level definition in dependencies in env.yml

#75 pjurczen closed 4 months ago
1
Replace `--eval-mode = ['deterministic', 'stochastic']` with temperature.

#74 eihli opened 5 months ago
1
Add debug logging?

#73 eihli opened 5 months ago
4
test issue

#72 sidhsikka1 closed 5 months ago
2
Error running `eval.py` on checkpoint created using `--pretrained_lm=gpt2`.

#71 eihli opened 5 months ago
0
Changed saved model name from an ID to a timestamp

#70 nomitchell closed 5 months ago
1
[Minor] Update experiment name to include timestamp so that experiments are sorted in places like file systems.

#69 eihli opened 5 months ago
1
Get BabyAI Datasets Working

#68 eihli opened 5 months ago
0
Use a dataclass for train/eval arguments so that we get accurate type hinting and don't tie our runtime to the command-line-intentioned ArgumentParser.

#67 eihli closed 4 months ago
5
Training run with certain non-default values causes evaluation run to fail.

#66 eihli opened 5 months ago
0
Enable eval.py to work on text datasets

#65 SasankYadati closed 5 months ago
2
env.yml file specifies dependencies at build level, so it doesn't work straight away on Windows (and maybe other hosts)

#64 SasankYadati opened 6 months ago
1
Investigate control data

#63 snat-s opened 6 months ago
0
Pick models to profile for the datasets

#62 harshsikka opened 6 months ago
0
Profiling existing models

#61 harshsikka opened 6 months ago
0
Outline Language & Vision Modality next steps

#60 harshsikka opened 6 months ago
2
Outline Control Modality Next Steps across Data, Training

#59 harshsikka opened 6 months ago
3
v0 Dataset and Benchmark Specification

#58 harshsikka opened 6 months ago
4
Updated Resource Analysis

#57 harshsikka opened 6 months ago
0
Dataset Reconcilliation

#56 harshsikka opened 6 months ago
1
Fix version dependency issue #54

#55 eihli closed 5 months ago
1
Dependency version compatibility issue.

#54 eihli closed 4 months ago
1
Add Namespace to argparse for type hints

#36 eihli closed 7 months ago
0
Update TaskType references to use enum, not str

#35 eihli closed 5 months ago
2
Minor cleanups as I review the VQA PR.

#34 eihli closed 4 months ago
1
Wandb tracking issue for distributed runs launched with accelerate.

#33 eihli opened 7 months ago
0
fix broken link to DeepMind's GATO publication

#32 eihli closed 7 months ago
1
Evaluation of datasets used for flamingo

#31 BobakBagheri closed 5 months ago
1
Add vqa

#30 henryj18 closed 4 months ago
1
Add image caption

#29 henryj18 closed 6 months ago
1
text_prop default value needs to be set to 0, and that should also apply to caption_prop

#28 henryj18 closed 8 months ago
1
Add code to cut sequence that is longer than context_len to fit in

#27 henryj18 opened 8 months ago
0
Update README.md

#26 harshsikka closed 8 months ago
0