Merge pull request #336 from tmm1/flash-attn
Fix flash-attn + qlora not working with llama models
This commit is contained in:
@@ -92,7 +92,9 @@ def load_model(
|
|||||||
|
|
||||||
if cfg.is_llama_derived_model and cfg.flash_attention:
|
if cfg.is_llama_derived_model and cfg.flash_attention:
|
||||||
if cfg.device not in ["mps", "cpu"] and not cfg.inference:
|
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")
|
LOG.info("patching with flash attention")
|
||||||
replace_llama_attn_with_flash_attn()
|
replace_llama_attn_with_flash_attn()
|
||||||
@@ -331,6 +333,16 @@ def load_model(
|
|||||||
model, use_gradient_checkpointing=cfg.gradient_checkpointing
|
model, use_gradient_checkpointing=cfg.gradient_checkpointing
|
||||||
)
|
)
|
||||||
|
|
||||||
|
# LlamaRMSNorm layers are in fp32 after kbit_training, 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)
|
model, lora_config = load_adapter(model, cfg, adapter)
|
||||||
|
|
||||||
if cfg.ddp and not load_in_8bit:
|
if cfg.ddp and not load_in_8bit:
|
||||||
@@ -407,14 +419,6 @@ def load_llama_adapter(model, cfg):
|
|||||||
else:
|
else:
|
||||||
model = get_peft_model(model, peft_config)
|
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()
|
model.print_trainable_parameters()
|
||||||
|
|
||||||
return model, peft_config
|
return model, peft_config
|
||||||
|
|||||||
Reference in New Issue
Block a user