Spaces:

Dovakiins
/

qwerrwe

Build error

App Files Files Community

qwerrwe / src /axolotl /utils

100 contributors

History: 253 commits

winglian's picture

workaround so training doesn't hang when packed dataloader batches aren't even (#461)

c69faee unverified about 1 year ago

__init__.py

0 Bytes

WIP large refactor to make finetune script a little more manageable (#3) over 1 year ago
bench.py

1.24 kB

Fix crash when running without CUDA about 1 year ago
callbacks.py

2.75 kB

improve GPU logging to break out pytorch cache and system mem about 1 year ago
collators.py

5.39 kB

Attention mask and position id fixes for packing (#285) about 1 year ago
config.py
7.72 kB

recast loralayer, norm, lmhead + embed token weights per original qlora (#393) about 1 year ago
data.py
24 kB

support user defined prompters, pretokenized datasets in config, local parquet, local arrow files (#348) about 1 year ago
dataloader.py

9.86 kB

workaround so training doesn't hang when packed dataloader batches aren't even (#461) about 1 year ago
dict.py

326 Bytes

fix DefaultDict.__or__ about 1 year ago
distributed.py

1.31 kB

use context manager to run things on rank0 before others (#397) about 1 year ago
models.py
19.4 kB

recast loralayer, norm, lmhead + embed token weights per original qlora (#393) about 1 year ago
schedulers.py

3.58 kB

add axolotl trainer and quadratic warmup over 1 year ago
tokenization.py

1.37 kB

Adding logging enhancement about 1 year ago
trainer.py
24.1 kB

always drop samples that are too long (#452) about 1 year ago
wandb.py

858 Bytes

Add wandb_entity to wandb options, update example configs, update README (#361) about 1 year ago