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

Fix flash-attn + qlora not working with llama models #336

Merged
merged 4 commits into from
Aug 3, 2023
Merged

Fix flash-attn + qlora not working with llama models #336

Show file tree
Hide file tree
Changes from 3 commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
248bf90
ensure flash-attn fixes happen in both adapter/lora modes, and use to…
tmm1 Aug 2, 2023
312a9fa
move flash-attn monkey patch alongside the others
tmm1 Aug 3, 2023
78b9efb
scope flash-attn+qlora fix correctly, scope to llama, add comment
tmm1 Aug 3, 2023
2eda9e0
fix typo
tmm1 Aug 3, 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
0 src/axolotl/flash_attn.py → ...tl/monkeypatch/llama_attn_hijack_flash.py
View file
Open in desktop
File renamed without changes.
22 changes: 13 additions & 9 deletions src/axolotl/utils/models.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -92,7 +92,9 @@ def load_model(

if cfg.is_llama_derived_model and cfg.flash_attention:
if cfg.device not in ["mps", "cpu"] and not cfg.inference:
from axolotl.flash_attn import replace_llama_attn_with_flash_attn
from axolotl.monkeypatch.llama_attn_hijack_flash import (
replace_llama_attn_with_flash_attn,
)

LOG.info("patching with flash attention")
replace_llama_attn_with_flash_attn()
Expand Down Expand Up @@ -331,6 +333,16 @@ def load_model(
model, use_gradient_checkpointing=cfg.gradient_checkpointing
)

# LlamaRMSNorm layers are in fp32 after kit call, so we need to
# convert them back to fp16/bf16 for flash-attn compatibility.
if cfg.flash_attention and cfg.is_llama_derived_model:
for name, module in model.named_modules():
if "norm" in name:
module.to(torch_dtype)
if "lm_head" in name or "embed_tokens" in name:
if hasattr(module, "weight"):
module.to(torch_dtype)

model, lora_config = load_adapter(model, cfg, adapter)

if cfg.ddp and not load_in_8bit:
Expand Down Expand Up @@ -407,14 +419,6 @@ def load_llama_adapter(model, cfg):
else:
model = get_peft_model(model, peft_config)

if cfg.flash_attention:
for name, module in model.named_modules():
if "norm" in name:
module.to(torch.float16)
if "lm_head" in name or "embed_tokens" in name:
if hasattr(module, "weight"):
module.to(torch.float16)

model.print_trainable_parameters()

return model, peft_config
Expand Down