feature(utils): add utilities to support priming logic

inference_utils.py

@@ -1,8 +1,17 @@
-from typing import Dict
-import numpy as np 
+from pathlib import Path
+from typing import Dict, NamedTuple, Union
+import numpy as np
+import torch 
 
 NULL_CHAR = '\x00'
 
+
+class PrimingData(NamedTuple):
+    """combines data required for priming the HandwritingRNN sampling"""
+    stroke_tensors: torch.Tensor # (batch_size, num_prime_strokes, 3) 
+    char_seq_tensors: torch.Tensor # (batch_size, num_prime_chars)
+    char_seq_lengths: torch.Tensor # (batch_size,)
+
 def construct_alphabet_list(alphabet_string: str) -> list[str]:
     if not isinstance(alphabet_string, str):
         raise TypeError("alphabet_string must be a string") 
@@ -45,4 +54,36 @@ def convert_offsets_to_absolute_coords(stroke_offsets: list[list[float]]) -> lis
     strokes_array[:, 0] = np.cumsum(strokes_array[:, 0])  # cumulative dx
     strokes_array[:, 1] = np.cumsum(strokes_array[:, 1])  # cumulative dy
     
-    return strokes_array.tolist()
+    return strokes_array.tolist()
+
+
+def load_np_strokes(stroke_path: Union[Path, str]) -> np.ndarray:
+    """loads stroke sequence from stroke_path"""
+    stroke_path = Path(stroke_path)
+    if not stroke_path.exists():
+        raise FileNotFoundError(f"style strokes file not found at {stroke_path}")
+    
+    return np.load(stroke_path)
+
+def load_text(text_path: Union[Path, str]) -> str:
+    """loads text from a text_path""" 
+    text_path = Path(text_path) 
+    if not text_path.exists():
+        raise FileNotFoundError(f"Text file not found at {text_path}")
+    if not text_path.is_file():
+        raise IsADirectoryError(f"Path is a directory, not a file.")
+    
+    try: 
+        with open(text_path, 'r', encoding='utf-8') as f:
+            content = f.read() 
+        return content 
+
+    except Exception as e:
+        raise IOError(f"Error reading text file {text_path}: {e}")
+
+def load_priming_data(style: int):
+    
+    priming_text = load_text(f"./data/samples/sample{style}.txt")
+    priming_strokes = load_np_strokes(f"./data/samples/sample{style}.npy")
+    
+    return priming_text, priming_strokes