File size: 13,926 Bytes
72c0672 | 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 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 | import json
import base64
from offline_utils import packed_bytes_to_pseudo
def pack_compressed_spans(data, bits_per_compressed: int, compression_bit_threshold: int, compression_offset: int = 256):
"""
Convert consecutive compressed values into larger integers.
Args:
data: List of integers where 0-255 are raw bytes, compression_offset+ are compressed bytes
bits_per_compressed: Number of bits to use for each packed value
compression_bit_threshold: Number of bits each compressed value actually uses
compression_offset: Offset that marks start of compressed values (default 256)
Returns:
List with consecutive compressed spans packed into larger integers
"""
if not data:
return []
result = []
i = 0
assert compression_bit_threshold % bits_per_compressed == 0, "compression_bit_threshold must be divisible by bits_per_compressed"
packing_mask = (1 << bits_per_compressed) - 1
compression_mask = (1 << compression_bit_threshold) - 1
# Calculate byte-aligned padded size
padded_compression_bit_threshold = ((compression_bit_threshold + 7) // 8) * 8
padded_mask = (1 << padded_compression_bit_threshold) - 1
padding_bits = padded_compression_bit_threshold - compression_bit_threshold
while i < len(data):
if data[i] >= compression_offset:
# Find the end of consecutive compressed bytes
span_start = i
while i < len(data) and data[i] >= compression_offset:
i += 1
# Extract the span of compressed bytes
compressed_span = data[span_start:i]
base_values = [x - compression_offset for x in compressed_span]
# Process bytes incrementally to avoid large numbers
bit_buffer = 0
bits_in_buffer = 0
packed_values = []
for val in base_values:
# Add this byte to bit buffer
bit_buffer = (bit_buffer << 8) | val
bits_in_buffer += 8
# Extract padded chunks as soon as we have enough bits
while bits_in_buffer >= padded_compression_bit_threshold:
shift_amount = bits_in_buffer - padded_compression_bit_threshold
padded_val = (bit_buffer >> shift_amount) & padded_mask
# Remove the extracted bits from buffer
bit_buffer &= (1 << shift_amount) - 1
bits_in_buffer -= padded_compression_bit_threshold
# Strip padding by extracting only the meaningful bits
extracted_val = (padded_val >> padding_bits) & compression_mask
pack_buffer = extracted_val
pack_bits = compression_bit_threshold
# Pack values as soon as we have enough bits
while pack_bits >= bits_per_compressed:
pack_shift = pack_bits - bits_per_compressed
packed_val = (pack_buffer >> pack_shift) & packing_mask
packed_values.append(packed_val + compression_offset)
# Remove packed bits from pack buffer
pack_buffer &= (1 << pack_shift) - 1
pack_bits -= bits_per_compressed
assert bits_in_buffer == 0, "bits_in_buffer must be 0 after processing compressed span"
assert pack_bits == 0, "pack_bits must be 0 after packing"
result.extend(packed_values)
else:
# Raw byte (0-255), keep as is
result.append(data[i])
i += 1
return result
def unpack_compressed_spans(packed_data, bits_per_compressed: int, compression_bit_threshold: int, compression_offset: int = 256):
"""
Reverse operation: unpack larger integers back to consecutive compressed bytes.
Args:
packed_data: List with packed compressed spans
bits_per_compressed: Number of bits used for packing
compression_bit_threshold: Number of bits each compressed value actually uses
compression_offset: Offset used for compressed values
Returns:
Original format with consecutive compressed bytes
"""
result = []
i = 0
# Calculate byte-aligned padded size
padded_compression_bit_threshold = ((compression_bit_threshold + 7) // 8) * 8
padding_bits = padded_compression_bit_threshold - compression_bit_threshold
while i < len(packed_data):
if packed_data[i] >= compression_offset: # Start of compressed span
# Find consecutive packed values
span_start = i
while i < len(packed_data) and packed_data[i] >= compression_offset:
i += 1
packed_span = packed_data[span_start:i]
base_values = [x - compression_offset for x in packed_span]
# Unpack using two-phase process to handle padding
unpacked_bytes = []
bit_buffer = 0
bits_in_buffer = 0
for val in base_values:
# Add this packed value to our bit buffer
bit_buffer = (bit_buffer << bits_per_compressed) | val
bits_in_buffer += bits_per_compressed
# Extract compression_bit_threshold values as soon as we have enough bits
while bits_in_buffer >= compression_bit_threshold:
# Extract the top compression_bit_threshold bits
shift_amount = bits_in_buffer - compression_bit_threshold
compressed_val = (bit_buffer >> shift_amount) & ((1 << compression_bit_threshold) - 1)
# Remove the extracted bits from buffer
bit_buffer &= (1 << shift_amount) - 1
bits_in_buffer -= compression_bit_threshold
# Add padding back to make it byte-aligned
padded_val = compressed_val << padding_bits
# Convert padded value back to bytes
bytes_needed = padded_compression_bit_threshold // 8
for byte_idx in range(bytes_needed):
shift = (bytes_needed - 1 - byte_idx) * 8
byte_val = (padded_val >> shift) & 0xFF
unpacked_bytes.append(byte_val + compression_offset)
# Verify all bits were processed cleanly
assert bits_in_buffer == 0, "bits_in_buffer must be 0 after unpacking compressed span"
result.extend(unpacked_bytes)
else:
# Raw byte, keep as is
result.append(packed_data[i])
i += 1
return result
def run_test_case(test_name: str, data: list, bits_per_compressed: int, compression_bit_threshold: int):
"""Run a single test case with comprehensive validation."""
print(f"π§ͺ {test_name}")
print(f" Original: {data}")
print(f" Config: bits_per_compressed={bits_per_compressed}, compression_bit_threshold={compression_bit_threshold}")
try:
# Test packing
packed = pack_compressed_spans(data, bits_per_compressed, compression_bit_threshold)
print(f" Packed: {packed}")
# Test unpacking
unpacked = unpack_compressed_spans(packed, bits_per_compressed, compression_bit_threshold)
print(f" Unpacked: {unpacked}")
# Verify round-trip
success = data == unpacked
print(f" Result: {'β
PASS' if success else 'β FAIL'}")
# Show compression stats
original_compressed = len([x for x in data if x >= 256])
packed_compressed = len([x for x in packed if x >= 256])
if original_compressed > 0:
ratio = original_compressed / packed_compressed if packed_compressed > 0 else 0
print(f" Stats: {original_compressed} β {packed_compressed} compressed values ({ratio:.2f}x)")
return success
except Exception as e:
print(f" Result: β ERROR: {e}")
return False
def test_packing_comprehensive():
from m1_compression import utils
import random
def random_bytes_generator(n: int, bit_threshold: int):
ret = []
length = random.randint(n // 2, n)
for _ in range(length):
bits = ""
for _ in range(bit_threshold):
bits += "0" if random.random() < 0.5 else "1"
compressed_bytes, _ = utils.bits_to_bytes_padding_to_threshold(bits, bit_threshold)
ret.extend([c + 256 for c in list(compressed_bytes)])
ret.extend([random.randint(0, 255)])
return ret
"""Comprehensive test suite for packing functions."""
print("=" * 60)
print("π COMPREHENSIVE PACKING TESTS")
print("=" * 60)
test_results = []
# Test 1: Basic functionality - 16-bit alignment (no padding)
test_results.append(run_test_case(
"Basic 16-bit packing (no padding)",
random_bytes_generator(100, 16),
bits_per_compressed=16,
compression_bit_threshold=16
))
print()
# Test 2: 12-bit values with padding (12 bits stored in 16 bits)
test_results.append(run_test_case(
"12-bit values with 4-bit padding",
random_bytes_generator(100, 12),
bits_per_compressed=12,
compression_bit_threshold=12
))
print()
# Test 3: 20-bit values with padding (20 bits stored in 24 bits)
test_results.append(run_test_case(
"20-bit values with 4-bit padding",
random_bytes_generator(100, 20),
bits_per_compressed=20,
compression_bit_threshold=20
))
print()
# Test 5: Edge case - single compressed byte
test_results.append(run_test_case(
"Single compressed byte",
[100, 256, 200],
bits_per_compressed=8,
compression_bit_threshold=8
))
print()
# Test 6: Edge case - no compressed bytes
test_results.append(run_test_case(
"No compressed bytes",
[100, 200, 50, 150],
bits_per_compressed=16,
compression_bit_threshold=16
))
print()
# Test 7: Edge case - all compressed bytes
test_results.append(run_test_case(
"All compressed bytes",
[256, 257, 258, 259, 260, 261],
bits_per_compressed=8,
compression_bit_threshold=8
))
print()
# Test 8: Mixed compression ratios
test_results.append(run_test_case(
"24-bit to 12-bit packing (2:1 ratio)",
random_bytes_generator(100, 24),
bits_per_compressed=12,
compression_bit_threshold=24
))
print()
# Summary
passed = sum(test_results)
total = len(test_results)
print("=" * 60)
print(f"π TEST SUMMARY: {passed}/{total} tests passed")
print("=" * 60)
if passed == total:
print("π All tests passed! The implementation is working correctly.")
else:
print("β οΈ Some tests failed. Please review the implementation.")
return passed, total
def test_real_data():
print("=" * 40)
print("π§ REAL DATA TESTS")
print("=" * 40)
key = "m1_ac_ow20_escapefb-False_iterative-True"
with open("output_compress/m1.chunk.0_out_0_out_0_writer_0.jsonl", "r") as f:
for i, line in enumerate(f):
data = json.loads(line)
# NOTE: for visualization purposes, we replace values > 256 with byte '_'
# bytes_array = packed_bytes_to_pseudo(base64.b64decode(data[key]))
# bytes_array = [b if b < 256 else ord('_') for b in bytes_array]
# bytes_string = bytes(bytes_array).decode("utf-8", errors="replace")
# print(bytes_string)
# extract bit_threshold in key
key_splits = key.split("_")
bit_threshold = None
for k in key_splits:
if k.startswith(
):
bit_threshold = int(k[len("ow"):])
break
assert bit_threshold is not None
print(f"Bit threshold: {bit_threshold}")
# NEW: Apply the packing function to the original bytes_array (before replacement)
original_bytes_array = packed_bytes_to_pseudo(base64.b64decode(data[key]))
run_test_case(
f"Packing {bit_threshold}-bit values",
original_bytes_array,
10,
bit_threshold
)
if i > 4:
break
def test_error_conditions():
"""Test error conditions and edge cases."""
print("\nπ§ ERROR CONDITION TESTS")
print("=" * 40)
# Test invalid bit alignment
try:
pack_compressed_spans([256, 257], 10, 15) # 15 % 10 != 0
print("β Should have failed on invalid bit alignment")
except AssertionError:
print("β
Correctly caught invalid bit alignment")
# Test empty data
result = pack_compressed_spans([], 16, 16)
print(f"β
Empty data handling: {result == []}")
print()
if __name__ == "__main__":
# Run comprehensive tests
passed, total = test_packing_comprehensive()
# Final result
if passed == total:
print("π ALL TESTS COMPLETED SUCCESSFULLY!")
else:
print("π₯ SOME TESTS FAILED!")
test_real_data()
# Run error condition tests
test_error_conditions()
|