Fix _generate_seed_sequence sampling (#21399)

Author: SkafteNicki

src/lightning/fabric/utilities/seed.py

@@ -111,10 +111,19 @@ def pl_worker_init_function(worker_id: int, rank: Optional[int] = None) -> None:
 
 
 def _generate_seed_sequence(base_seed: int, worker_id: int, global_rank: int, count: int) -> list[int]:
-    """Generates a sequence of seeds from a base seed, worker id and rank using the linear congruential generator (LCG)
-    algorithm."""
+    """Generates a sequence of seeds from a base seed, worker id and rank using hash-based mixing followed by the
+    linear congruential generator (LCG) algorithm."""
     # Combine base seed, worker id and rank into a unique 64-bit number
     combined_seed = (base_seed << 32) | (worker_id << 16) | global_rank
+
+    # Apply hash-based mixing (MurmurHash3 finalizer) to distribute bits uniformly
+    # This ensures that small base seeds don't result in zeros in lower bits
+    combined_seed ^= combined_seed >> 33
+    combined_seed = (combined_seed * 0xFF51AFD7ED558CCD) & ((1 << 64) - 1)
+    combined_seed ^= combined_seed >> 33
+    combined_seed = (combined_seed * 0xC4CEB9FE1A85EC53) & ((1 << 64) - 1)
+    combined_seed ^= combined_seed >> 33
+
     seeds = []
     for _ in range(count):
         # x_(n+1) = (a * x_n + c) mod m. With c=1, m=2^64 and a is D. Knuth's constant

src/lightning/pytorch/CHANGELOG.md

@@ -20,6 +20,10 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 
 -
 
+### Fixed
+
+- Fix `_generate_seed_sequence_sampling` function not producing unique seeds ([#21399](https://github.com/Lightning-AI/pytorch-lightning/pull/21399))
+
 
 ## [2.6.0] - 2025-11-28
 

tests/tests_fabric/utilities/test_seed.py

@@ -10,6 +10,7 @@
 
 from lightning.fabric.utilities.seed import (
     _collect_rng_states,
+    _generate_seed_sequence,
     _set_rng_states,
     pl_worker_init_function,
     reset_seed,
@@ -153,3 +154,23 @@ def test_pl_worker_init_function(base_seed, num_workers, num_ranks):
     assert len(stdlib_rands) == num_ranks * num_workers
     assert len(numpy_rands) == num_ranks * num_workers
     assert len(torch_rands | stdlib_rands | numpy_rands) == 3 * num_workers * num_ranks
+
+
+def test_generate_seed_sequence_no_collision():
+    """Test that _generate_seed_sequence produces unique seeds for different base seeds."""
+    base_seeds = [0, 1, 42, 123, 999, 12345]
+    generated_seeds = []
+    random_outputs = []
+
+    for base_seed in base_seeds:
+        seed_everything(base_seed)
+        process_seed = torch.initial_seed()
+        generated_seed = _generate_seed_sequence(process_seed, worker_id=0, global_rank=0, count=1)[0]
+        generated_seeds.append(generated_seed)
+        torch.manual_seed(generated_seed)
+        random_outputs.append(tuple(torch.randn(10).tolist()))
+
+    assert len(set(generated_seeds)) == len(generated_seeds), (
+        "Generated seeds should be unique for different base seeds"
+    )
+    assert len(set(random_outputs)) == len(random_outputs), "Random outputs should be unique for different base seeds"