Spaces:

Dovakiins
/

qwerrwe

Build error

App Files Files Community

qwerrwe / src /axolotl /utils

100 contributors

History: 411 commits

Simon Hällqvist

Enable or disable bf16 support based on availability (#1116)

0865613 unverified about 1 year ago

samplers
Efficiently get the length of the tokenized docs (#1063) about 1 year ago
__init__.py

0 Bytes

WIP large refactor to make finetune script a little more manageable (#3) almost 2 years ago
bench.py

2.06 kB

misc fixes to add gptq tests (#621) over 1 year ago
callbacks.py

22.2 kB

[WandB] Push axolotl config to top level wandb files (#1014) about 1 year ago
chat_templates.py
1.52 kB

Adds chat templates (#1022) about 1 year ago
collators.py
7.99 kB

streaming multipack for pretraining dataset (#959) about 1 year ago
config.py
18.8 kB

Enable or disable bf16 support based on availability (#1116) about 1 year ago
data.py
33.5 kB

Efficiently get the length of the tokenized docs (#1063) about 1 year ago
dict.py

326 Bytes

fix DefaultDict.__or__ over 1 year ago
distributed.py

7.07 kB

cleanup verbosity a bit over 1 year ago
freeze.py

1.25 kB

Fix Deepspeed loading (#950) about 1 year ago
lora_embeddings.py

377 Bytes

add gptneox embeddings, fix phi2 inputs, also fix the casting (#1083) about 1 year ago
mlflow_.py

511 Bytes

Add: mlflow for experiment tracking (#1059) [skip ci] about 1 year ago
models.py
29.3 kB

keep gate in fp32 for 16 bit loras (#1105) about 1 year ago
schedulers.py

4.75 kB

Cosine learning rate schedule - minimum learning rate (#1062) about 1 year ago
tokenization.py

1.43 kB

add a latest tag for regular axolotl image, cleanup extraneous print statement (#746) over 1 year ago
trainer.py
10.7 kB

additional logging to get maximum token length of a sequence in the dataset (#1066) [skip ci] about 1 year ago
wandb_.py

609 Bytes

Feat(wandb): Refactor to be more flexible (#767) about 1 year ago