File size: 2,725 Bytes
ce847d4 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 |
"""Deep-dive into model_11 and model_22 graph structure to understand OneOCRFeatureExtract."""
import onnx
import numpy as np
from pathlib import Path
models_dir = Path("oneocr_extracted/onnx_models")
for idx in [11, 22]:
matches = list(models_dir.glob(f"model_{idx:02d}_*"))
model = onnx.load(str(matches[0]))
print(f"\n{'='*70}")
print(f"FULL GRAPH: model_{idx:02d}")
print(f"{'='*70}")
# All initializers (weights)
print(f"\n Initializers ({len(model.graph.initializer)}):")
for init in model.graph.initializer:
data = onnx.numpy_helper.to_array(init)
print(f" {init.name}: shape={data.shape}, dtype={data.dtype}, "
f"range=[{data.min():.4f}, {data.max():.4f}]")
# All nodes
print(f"\n Nodes ({len(model.graph.node)}):")
for i, node in enumerate(model.graph.node):
domain_str = f" (domain={node.domain!r})" if node.domain else ""
print(f" [{i}] {node.op_type}{domain_str}")
print(f" in: {list(node.input)}")
print(f" out: {list(node.output)}")
for attr in node.attribute:
if attr.type == 2: # INT
print(f" {attr.name} = {attr.i}")
elif attr.type == 1: # FLOAT
print(f" {attr.name} = {attr.f}")
elif attr.type == 3: # STRING
val = attr.s
if len(val) > 100:
print(f" {attr.name} = bytes({len(val)})")
else:
print(f" {attr.name} = {val!r}")
elif attr.type == 4: # TENSOR
t = attr.t
print(f" {attr.name} = tensor(dtype={t.data_type}, dims={list(t.dims)}, "
f"raw_bytes={len(t.raw_data)})")
elif attr.type == 7: # INTS
print(f" {attr.name} = {list(attr.ints)}")
elif attr.type == 6: # FLOATS
print(f" {attr.name} = {list(attr.floats)[:10]}...")
# Show feature/config initializer details
for init in model.graph.initializer:
if "config" in init.name.lower() or "feature" in init.name.lower():
raw = init.raw_data
print(f"\n feature/config blob: {len(raw)} bytes")
print(f" First 64 bytes (hex): {raw[:64].hex()}")
print(f" Last 32 bytes (hex): {raw[-32:].hex()}")
# Try to interpret structure
# Check if starts with dimension info
print(f" As uint32 first 8 values: {[int.from_bytes(raw[i:i+4], 'little') for i in range(0, 32, 4)]}")
print(f" As float32 first 8 values: {list(np.frombuffer(raw[:32], dtype=np.float32))}")
|