additional fixes for docker and saving compressed

This commit is contained in:
Wing Lian
2025-04-28 13:06:59 -04:00
parent 5246aebc04
commit fd5c985038
3 changed files with 17 additions and 16 deletions

View File

@@ -33,9 +33,9 @@ RUN if [ "$NIGHTLY_BUILD" = "true" ] ; then \
RUN pip install packaging==23.2 setuptools==75.8.0
RUN if [ "$AXOLOTL_EXTRAS" != "" ] ; then \
pip install --no-build-isolation -e .[deepspeed,flash-attn,ring-flash-attn,optimizers,ray,$AXOLOTL_EXTRAS] $AXOLOTL_ARGS; \
pip install --no-build-isolation -e .[deepspeed,flash-attn,ring-flash-attn,optimizers,ray,llmcompressor,$AXOLOTL_EXTRAS] $AXOLOTL_ARGS; \
else \
pip install --no-build-isolation -e .[deepspeed,flash-attn,ring-flash-attn,optimizers,ray] $AXOLOTL_ARGS; \
pip install --no-build-isolation -e .[deepspeed,flash-attn,ring-flash-attn,optimizers,ray,llmcompressor] $AXOLOTL_ARGS; \
fi
RUN python scripts/unsloth_install.py | sh

View File

@@ -20,9 +20,9 @@ WORKDIR /workspace/axolotl
# If AXOLOTL_EXTRAS is set, append it in brackets
RUN if [ "$AXOLOTL_EXTRAS" != "" ] ; then \
pip install --no-build-isolation -e .[deepspeed,flash-attn,ring-flash-attn,optimizers,ray,$AXOLOTL_EXTRAS] $AXOLOTL_ARGS; \
pip install --no-build-isolation -e .[deepspeed,flash-attn,ring-flash-attn,optimizers,ray,llmcompressor,$AXOLOTL_EXTRAS] $AXOLOTL_ARGS; \
else \
pip install --no-build-isolation -e .[deepspeed,flash-attn,ring-flash-attn,optimizers,ray] $AXOLOTL_ARGS; \
pip install --no-build-isolation -e .[deepspeed,flash-attn,ring-flash-attn,optimizers,ray,llmcompressor] $AXOLOTL_ARGS; \
fi
RUN python scripts/unsloth_install.py | sh

View File

@@ -288,7 +288,19 @@ def save_trained_model(
os.remove(os.path.join(cfg.output_dir, "model.safetensors"))
except FileNotFoundError:
pass
elif hasattr(cfg, "llmcompressor") and cfg.llmcompressor:
elif cfg.local_rank == 0:
if cfg.flash_optimum and BetterTransformer:
model = BetterTransformer.reverse(model)
if cfg.rl and cfg.adapter and not cfg.rl_adapter_ref_model:
trainer.model.save_pretrained(
cfg.output_dir, safe_serialization=safe_serialization
)
model.save_pretrained(cfg.output_dir, safe_serialization=safe_serialization)
if hasattr(cfg, "llmcompressor") and cfg.llmcompressor:
# TODO: add integration support so this can be implemented completely within the plugin
from axolotl.integrations.llm_compressor.utils import (
save_compressed_model,
)
@@ -301,17 +313,6 @@ def save_trained_model(
save_compressed=cfg.llmcompressor.save_compressed,
)
elif cfg.local_rank == 0:
if cfg.flash_optimum and BetterTransformer:
model = BetterTransformer.reverse(model)
if cfg.rl and cfg.adapter and not cfg.rl_adapter_ref_model:
trainer.model.save_pretrained(
cfg.output_dir, safe_serialization=safe_serialization
)
model.save_pretrained(cfg.output_dir, safe_serialization=safe_serialization)
def create_model_card(cfg: DictDefault, trainer: Trainer):
"""