issues
search
delphi-suite
/
delphi
small language models training made easy
Apache License 2.0
9
stars
1
forks
source link
issues
Newest
Newest
Most commented
Recently updated
Oldest
Least commented
Least recently updated
configs should be prioritized in argument order (later overrides former)
#102
jaidhyani
closed
6 months ago
0
Simplify run training
#101
jaidhyani
closed
6 months ago
4
setup + requirements deps in pyproject.toml; transformers to 4.39.2
#100
jaidhyani
closed
6 months ago
2
Add "checkpoint_mode" kwarg to plotting
#99
siwei-li
closed
6 months ago
2
final evaluation notebook
#98
menamerai
closed
4 months ago
1
Final evaluation notebook
#97
menamerai
closed
6 months ago
2
tokenize dataset with custom dataset
#96
joshuawe
closed
5 months ago
0
Train rework 2
#95
jaidhyani
closed
6 months ago
0
use custom HF namespaces
#94
joshuawe
closed
6 months ago
0
tokenize dataset script with custom huggingface namespaces
#93
joshuawe
closed
6 months ago
1
Added token selector
#92
menamerai
closed
6 months ago
0
Train rework 2
#91
jaidhyani
closed
6 months ago
2
flatten + simplify training script arguments
#90
jaidhyani
closed
6 months ago
0
Train rework
#89
jaidhyani
closed
6 months ago
1
move any non-trivial logic related to scripts into the main library and test it
#88
jaidhyani
closed
6 months ago
0
make eval plot for model checkpoints
#87
siwei-li
closed
6 months ago
0
support arbitrary tokenized dataset
#86
jettjaniak
closed
6 months ago
4
simplify batching logic
#85
jaidhyani
closed
6 months ago
0
huggingface upload logic
#84
jaidhyani
closed
6 months ago
0
resume_from_dir should be separate argument
#83
jaidhyani
closed
6 months ago
0
Assorted deletions/simplicatications
#82
jaidhyani
closed
6 months ago
2
Checkpoint logic refactoring
#81
jaidhyani
closed
6 months ago
0
Only latest optimizer
#80
jaidhyani
closed
6 months ago
0
test and validate gradient accumulation logic (+ fix if necessary)
#79
jaidhyani
closed
6 months ago
4
isolate and test actual training parts of train_step in their own function
#78
jaidhyani
closed
6 months ago
0
checkpoint parameterization
#77
jaidhyani
closed
6 months ago
0
training step refactor and testing
#76
jaidhyani
closed
6 months ago
0
Simplify model configuration to only use transformers library directly
#75
jaidhyani
closed
6 months ago
0
Make training more legible
#74
jaidhyani
closed
6 months ago
1
tools to skip the actual training/eval steps when debugging
#73
jaidhyani
closed
6 months ago
0
fix broken output_dir and run_name config
#72
jaidhyani
closed
6 months ago
0
Update transformers to 4.39
#71
jaidhyani
closed
6 months ago
0
Support for all transformers library CausalLM models
#70
jaidhyani
closed
6 months ago
0
50 evals research notebook (sampling method)
#69
menamerai
closed
6 months ago
1
53 get new non spaCy token labels
#68
transcendingvictor
closed
6 months ago
0
dataclass fields have metadata! Let's use it for documentation
#67
jaidhyani
closed
6 months ago
1
Update README to reflect how to use Delphi
#66
jaidhyani
closed
6 months ago
2
Starting to update README
#65
jaidhyani
closed
6 months ago
0
Default max_position_embeddings should be 512, not 513
#64
jaidhyani
closed
6 months ago
0
Outline Delphi Project
#63
joshuawe
closed
6 months ago
0
training script improvements
#62
jettjaniak
closed
6 months ago
0
Bug in "delphi-suite/stories-tokenizer"
#61
menamerai
closed
5 months ago
1
Deprecate llama2c
#60
jaidhyani
closed
6 months ago
0
add wandb to requirements.txt
#59
jaidhyani
closed
6 months ago
0
Add requirements to setup.py
#58
jaidhyani
closed
6 months ago
2
update beartype to 0.18
#57
jettjaniak
closed
6 months ago
5
Model loss difference notebook
#56
menamerai
closed
6 months ago
1
Add function to tokenize text stories and split into batches
#55
siwei-li
closed
6 months ago
4
Training_script_refactor
#54
jaidhyani
closed
6 months ago
1
get new non-spacy token labels
#53
transcendingvictor
closed
6 months ago
5
Previous
Next