allenai OLMo issues - Githubissues

allenai / OLMo

Modeling, training, eval, and inference code for OLMo

https://allenai.org/olmo

Apache License 2.0

4.19k stars 390 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

What did OLMo 1B converge to?

#642 sidereior opened 1 day ago
0
Resuming training on unsharded checkpoint

#641 lecifire opened 1 day ago
0
Multi node training

#640 shahizat opened 2 days ago
0
MoE

#639 Muennighoff opened 5 days ago
0
WIP: Scaling laws pipeline

#638 AkshitaB opened 6 days ago
0
muP implementation

#637 AkshitaB opened 6 days ago
0
Add option to skip optim steps for 0 grad params

#636 epwalsh opened 1 week ago
0
Unit tests

#635 AkshitaB opened 1 week ago
0
Fix Z-loss calculation

#634 epwalsh closed 1 week ago
4
Tokenizer with relative path import fails when using olmo as pip library

#633 viking-sudo-rm opened 1 week ago
0
How the 1B and 7B model are initialized?

#632 sanyalsunny111 opened 1 week ago
0
Make hf_olmo support AutoModelForCausalLM

#631 2015aroras closed 1 week ago
0
Amberish 7B hero run

#629 epwalsh opened 2 weeks ago
0
Olmo tiny scripts

#628 ananyahjha93 closed 1 week ago
0
Key 'https://olmo_checkpoints' not in 'TrainConfig'

#627 jeqcho opened 2 weeks ago
1
Inspect training data improvements

#626 2015aroras closed 2 weeks ago
0
What is the true MLP ratio for OLMo 7B?

#625 jeqcho closed 2 weeks ago
2
Make olmo-core checkpointer more robust on weka

#624 epwalsh closed 2 weeks ago
0
HF dataset loading optimizations

#623 2015aroras closed 3 weeks ago
0
Cant use LORA

#622 bdytx5 opened 3 weeks ago
6
Config for Amberish experiments at 1B

#621 drschwenk opened 3 weeks ago
0
Running Amber experiments at 7B

#620 epwalsh opened 3 weeks ago
0
Add most OLMo 1.7-7B checkpoints

#619 2015aroras closed 3 weeks ago
0
Normal baselines

#618 AkshitaB opened 3 weeks ago
0
added git ref to the config keys

#617 drschwenk opened 3 weeks ago
0
Chameleon stability experiments

#616 AkshitaB opened 3 weeks ago
0
Officially add OLMo-core as a dependency

#615 epwalsh closed 3 weeks ago
0
Make include_instance_metadata a kwarg of build_train_dataloader

#614 2015aroras closed 3 weeks ago
0
Make include_instance_metadata a kwarg of build_train_dataloader

#613 2015aroras closed 3 weeks ago
0
adding DDP to the codebase

#612 ananyahjha93 closed 3 weeks ago
3
Read and use tokenizer identifier from config

#611 2015aroras closed 3 weeks ago
0
[HF Converter] Get tokenizer path from config as default

#610 2015aroras closed 3 weeks ago
0
Finetuning config file

#609 joellliu opened 3 weeks ago
2
How many tokens were trained for 7B model.

#608 mathfinder opened 3 weeks ago
1
Rewrite initialization

#607 AkshitaB closed 3 weeks ago
2
now accepts wandb project and entity as options

#606 drschwenk closed 1 month ago
2
Add option to record step size metrics from AdamW

#605 epwalsh opened 1 month ago
0
Adds a tool that diffs two wandb runs

#604 dirkgr closed 3 weeks ago
1
Unshard without passing checkpointer type

#603 2015aroras closed 1 month ago
1
fixed host-device sync at each clipping step

#602 ananyahjha93 closed 1 month ago
0
Fixes clipping

#601 ananyahjha93 closed 1 month ago
0
Remove usages of Auto* methods in hf_olmo tests

#600 2015aroras closed 1 month ago
0
Merging the train-olmo-large branch

#599 dirkgr closed 1 month ago
0
is_causal=attention_bias is None

#598 nkkbr opened 1 month ago
1
Default eos_token_id in `scripts/prepare-tulu-data.py`

#597 y0mingzhang closed 1 month ago
1
why is the total_grad_norm increasing across training?

#596 ryanyxw opened 1 month ago
5
Expose memmap_dtype in the data configuration

#595 leon-g-xu closed 4 weeks ago
2
Expose memmap dtype in data config

#594 leon-g-xu closed 4 weeks ago
2
Inspect training data without data indices

#593 2015aroras closed 1 month ago
0
training directly from object storage?

#592 joellliu closed 3 weeks ago
2