Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

update README w deepspeed info #605

Merged
merged 4 commits into from
Sep 22, 2023
Merged

update README w deepspeed info #605

Show file tree
Hide file tree
Changes from 2 commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
52be465
update README w deepspeed info
winglian Sep 19, 2023
9bdb423
use linear warmup for deepspeed
winglian Sep 19, 2023
2bb1f9d
update requirements to include deepspeed and update readme
winglian Sep 21, 2023
e7d8c46
don't need redundant deepspeed install instructions
winglian Sep 22, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
18 changes: 18 additions & 0 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -31,6 +31,7 @@ Features:
- [How to Use Custom Pretokenized Dataset](#how-to-use-your-custom-pretokenized-dataset)
- [Config](#config)
- [Train](#train)
- [Training w/ Deepspeed](#training-with-deepspeed)
- [Inference](#inference)
- [Merge LORA to Base](#merge-lora-to-base)
- [Common Errors](#common-errors-)
Expand Down Expand Up @@ -88,6 +89,8 @@ cd axolotl
pip3 install packaging
pip3 install -e .[flash-attn]
pip3 install -U git+https://github.com/huggingface/peft.git
# optionally install deepspeed
winglian marked this conversation as resolved.
Show resolved Hide resolved
pip3 install deepspeed

# finetune lora
accelerate launch -m axolotl.cli.train examples/openllama-3b/lora.yml
Expand Down Expand Up @@ -732,6 +735,21 @@ wandb_run_id:
wandb_log_model:
```

### Training with Deepspeed
winglian marked this conversation as resolved.
Show resolved Hide resolved

Deepspeed is an optimization suite for multi-gpu systems allowing you to train much larger models than you
might typically be able to fit into your GPU's VRAM. More information about the various optimization types
for deepspeed is available at https://huggingface.co/docs/accelerate/main/en/usage_guides/deepspeed#what-is-integrated

We provide several default deepspeed JSON configurations for ZeRO stage 1, 2, and 3.

```shell
# install deepspeed if you haven't already
pip3 install deepspeed
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This pip install is likely redundant if it's included in main instructions.


accelerate launch -m axolotl.cli.train examples/llama-2/config.py --deepspeed deepspeed/zero1.json
```

### Inference

Pass the appropriate flag to the train command:
Expand Down
76 changes: 39 additions & 37 deletions deepspeed/zero1.json
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,39 +1,41 @@
{
"zero_optimization": {
"stage": 1,
"overlap_comm": true
},
"bf16": {
"enabled": "auto"
},
"fp16": {
"enabled": "auto",
"auto_cast": false,
"loss_scale": 0,
"initial_scale_power": 32,
"loss_scale_window": 1000,
"hysteresis": 2,
"min_loss_scale": 1
},
"optimizer": {
"type": "AdamW",
"params": {
"lr": "auto",
"betas": "auto",
"eps": "auto",
"weight_decay": "auto"
}
},
"scheduler": {
"type": "WarmupDecayLR",
"params": {
"warmup_min_lr": "auto",
"warmup_max_lr": "auto",
"warmup_num_steps": "auto",
"total_num_steps": "auto"
}
},
"train_batch_size": "auto",
"train_micro_batch_size_per_gpu": "auto",
"wall_clock_breakdown": false
"zero_optimization": {
"stage": 1,
"overlap_comm": true
},
"bf16": {
"enabled": "auto"
},
"fp16": {
"enabled": "auto",
"auto_cast": false,
"loss_scale": 0,
"initial_scale_power": 32,
"loss_scale_window": 1000,
"hysteresis": 2,
"min_loss_scale": 1
},
"optimizer": {
"type": "AdamW",
"params": {
"lr": "auto",
"betas": "auto",
"eps": "auto",
"weight_decay": "auto"
}
},
"scheduler": {
"type": "WarmupDecayLR",
"params": {
"warmup_min_lr": "auto",
"warmup_max_lr": "auto",
"warmup_num_steps": "auto",
"warmup_type": "linear",
"total_num_steps": "auto"
}
},
"gradient_accumulation_steps": "auto",
"train_batch_size": "auto",
"train_micro_batch_size_per_gpu": "auto",
"wall_clock_breakdown": false
}
82 changes: 42 additions & 40 deletions deepspeed/zero2.json
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,43 +1,45 @@
{
"zero_optimization": {
"stage": 2,
"offload_optimizer": {
"device": "cpu"
},
"contiguous_gradients": true,
"overlap_comm": true
"zero_optimization": {
"stage": 2,
"offload_optimizer": {
"device": "cpu"
},
"bf16": {
"enabled": "auto"
},
"fp16": {
"enabled": "auto",
"auto_cast": false,
"loss_scale": 0,
"initial_scale_power": 32,
"loss_scale_window": 1000,
"hysteresis": 2,
"min_loss_scale": 1
},
"optimizer": {
"type": "AdamW",
"params": {
"lr": "auto",
"betas": "auto",
"eps": "auto",
"weight_decay": "auto"
}
},
"scheduler": {
"type": "WarmupDecayLR",
"params": {
"warmup_min_lr": "auto",
"warmup_max_lr": "auto",
"warmup_num_steps": "auto",
"total_num_steps": "auto"
}
},
"train_batch_size": "auto",
"train_micro_batch_size_per_gpu": "auto",
"wall_clock_breakdown": false
"contiguous_gradients": true,
"overlap_comm": true
},
"bf16": {
"enabled": "auto"
},
"fp16": {
"enabled": "auto",
"auto_cast": false,
"loss_scale": 0,
"initial_scale_power": 32,
"loss_scale_window": 1000,
"hysteresis": 2,
"min_loss_scale": 1
},
"optimizer": {
"type": "AdamW",
"params": {
"lr": "auto",
"betas": "auto",
"eps": "auto",
"weight_decay": "auto"
}
},
"scheduler": {
"type": "WarmupDecayLR",
"params": {
"warmup_min_lr": "auto",
"warmup_max_lr": "auto",
"warmup_num_steps": "auto",
"warmup_type": "linear",
"total_num_steps": "auto"
}
},
"gradient_accumulation_steps": "auto",
"train_batch_size": "auto",
"train_micro_batch_size_per_gpu": "auto",
"wall_clock_breakdown": false
}
4 changes: 3 additions & 1 deletion deepspeed/zero3.json
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -45,9 +45,11 @@
"params": {
"warmup_min_lr": "auto",
"warmup_max_lr": "auto",
"warmup_num_steps": "auto"
"warmup_num_steps": "auto",
"warmup_type": "linear"
}
},
"gradient_accumulation_steps": "auto",
"train_batch_size": "auto",
"train_micro_batch_size_per_gpu": "auto",
"wall_clock_breakdown": false
Expand Down
1 change: 1 addition & 0 deletions examples/phi/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,6 +3,7 @@
Due to some nuances with the phi code, please use deepspeed when training phi for full finetune.

```shell
# You may need to install deepspeed with `pip3 install deepspeed`
accelerate launch -m axolotl.cli.train examples/phi/phi-ft.yml --deepspeed deepspeed/zero1.json

# OR
Expand Down