show layer class names
This commit is contained in:
21
train.py
21
train.py
@@ -36,6 +36,27 @@ def train(config_path):
|
|||||||
print(f"\n[INFO] Loading {config['base_model']}...")
|
print(f"\n[INFO] Loading {config['base_model']}...")
|
||||||
errors = []
|
errors = []
|
||||||
|
|
||||||
|
|
||||||
|
model = AutoModelForCausalLM.from_pretrained(
|
||||||
|
"YOUR_MODEL_PATH",
|
||||||
|
torch_dtype=torch.bfloat16,
|
||||||
|
trust_remote_code=True,
|
||||||
|
)
|
||||||
|
print(type(model))
|
||||||
|
|
||||||
|
classes = set()
|
||||||
|
|
||||||
|
for _, module in model.named_modules():
|
||||||
|
classes.add(module.__class__.__name__)
|
||||||
|
|
||||||
|
print("\n=== Transformer classes ===")
|
||||||
|
for cls in sorted(classes):
|
||||||
|
if "Block" in cls or "Layer" in cls or "Decoder" in cls:
|
||||||
|
print(cls)
|
||||||
|
|
||||||
|
raise SystemExit
|
||||||
|
|
||||||
|
|
||||||
# ------------------------------------------------------------------
|
# ------------------------------------------------------------------
|
||||||
# Strategy 1: QLoRA with FSDP (preferred)
|
# Strategy 1: QLoRA with FSDP (preferred)
|
||||||
# ------------------------------------------------------------------
|
# ------------------------------------------------------------------
|
||||||
|
|||||||
Reference in New Issue
Block a user