vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Clean up llava surgery and remove name substitution hacks

Browse source 
Signed-off-by: Alex-Brooks <Alex.Brooks@ibm.com>
This commit is contained in:
Alex-Brooks 2025-01-22 01:43:49 -07:00
parent 92046a103d
commit cc1c135367
1 changed files with 32 additions and 21 deletions
Download patch file Download diff file Expand all files Collapse all files

53
examples/llava/llava_surgery_v2.py
View file

@ -33,6 +33,33 @@ def save_model(model, file_path, file_type):
else: else:
torch.save(model, file_path) torch.save(model, file_path)
# Helpers to match weight names from specific components or
# determine if a saved shard contains that component
def is_vision_tower(weight_name):
return (
weight_name.startswith("model.vision_tower") or
weight_name.startswith("vit.") or
weight_name.startswith("vision_tower")
)
def is_newline(weight_name):
return (
weight_name.startswith("model.image_newline") or
weight_name.startswith("image_newline")
)
def is_mm_projector(weight_name):
return (
weight_name.startswith("model.mm_projector") or
weight_name.startswith("vision_proj.") or
weight_name.startswith("multi_modal_projector")
)
def newline_criteria(checkpoint):
return any(is_newline(k) for k in checkpoint.keys())
def proj_criteria(checkpoint):
return any(is_mm_projector(k) for k in checkpoint.keys())
# Adapted function to clean vision tower from checkpoint # Adapted function to clean vision tower from checkpoint
def clean_vision_tower_from_checkpoint(checkpoint_path): def clean_vision_tower_from_checkpoint(checkpoint_path):
@ -40,7 +67,7 @@ def clean_vision_tower_from_checkpoint(checkpoint_path):
# file_type = 'pytorch' # file_type = 'pytorch'
model_path = os.path.dirname(checkpoint_path) model_path = os.path.dirname(checkpoint_path)
print(f"Searching for vision tower tensors in {checkpoint_path}") print(f"Searching for vision tower tensors in {checkpoint_path}")
clip_tensors = [k for k, v in checkpoint.items() if (k.startswith("model.vision_tower") or k.startswith("vit.") or k.startswith("vision_tower"))] clip_tensors = [k for k, v in checkpoint.items() if is_vision_tower(k)]
if len(clip_tensors) > 0: if len(clip_tensors) > 0:
print(f"Found {len(clip_tensors)} tensors to extract from {checkpoint_path}") print(f"Found {len(clip_tensors)} tensors to extract from {checkpoint_path}")
@ -84,12 +111,6 @@ def find_relevant_checkpoints(checkpoint_paths, newline_criteria, projector):
return newline_checkpoint_path, projector_checkpoint_path return newline_checkpoint_path, projector_checkpoint_path
def newline_criteria(checkpoint):
return any(k.startswith("model.image_newline") or k.startswith("image_newline") for k in checkpoint.keys())
def proj_criteria(checkpoint):
return any(k.startswith("model.mm_projector") or k.startswith("vision_proj.") or k.startswith("multi_modal_projector") for k in checkpoint.keys())
# Command-line interface setup # Command-line interface setup
ap = argparse.ArgumentParser() ap = argparse.ArgumentParser()
@ -123,14 +144,14 @@ first_checkpoint = None
if newline_checkpoint_path is not None: if newline_checkpoint_path is not None:
print(f"Taking newline from {newline_checkpoint_path}") print(f"Taking newline from {newline_checkpoint_path}")
first_checkpoint, file_type = load_model(newline_checkpoint_path) first_checkpoint, file_type = load_model(newline_checkpoint_path)
first_mm_tensors = [k for k, v in first_checkpoint.items() if k.startswith("model.image_newline") or k.startswith("image_newline")] first_mm_tensors = [k for k, v in first_checkpoint.items() if is_newline(k)]
# Load the checkpoint # Load the checkpoint
mm_tensors = [] mm_tensors = []
last_checkpoint = None last_checkpoint = None
if projector_checkpoint_path is not None: if projector_checkpoint_path is not None:
last_checkpoint, file_type = load_model(projector_checkpoint_path) last_checkpoint, file_type = load_model(projector_checkpoint_path)
mm_tensors = [k for k, v in last_checkpoint.items() if k.startswith("model.mm_projector") or k.startswith("vision_proj.") or k.startswith("multi_modal_projector")] mm_tensors = [k for k, v in last_checkpoint.items() if is_mm_projector(k)]
if len(mm_tensors) == 0: if len(mm_tensors) == 0:
if last_checkpoint is not None: if last_checkpoint is not None:
@ -146,20 +167,10 @@ print(f"Found additional {len(first_mm_tensors)} tensors to extract.")
projector = {} projector = {}
for name in mm_tensors: for name in mm_tensors:
assert last_checkpoint is not None assert last_checkpoint is not None
# HACK - this should probably be in the second script... projector[name] = last_checkpoint[name].float()
new_name = name
if new_name.startswith("multi_modal_projector.linear_1"):
new_name = new_name.replace("multi_modal_projector.linear_1", "mm.0")
elif new_name.startswith("multi_modal_projector.linear_2"):
new_name = new_name.replace("multi_modal_projector.linear_2", "mm.2")
projector[new_name] = last_checkpoint[name].float()
for name in first_mm_tensors: for name in first_mm_tensors:
assert first_checkpoint is not None assert first_checkpoint is not None
# HACK - this should probably be in the second script too... projector[name] = first_checkpoint[name].float()
new_name = name
if new_name == "image_newline":
new_name = "model.image_newline"
projector[new_name] = first_checkpoint[name].float()
if len(projector) > 0: if len(projector) > 0:
save_model(projector, f"{args.model}/llava.projector", 'pytorch') save_model(projector, f"{args.model}/llava.projector", 'pytorch')