Rewrite asserts as exceptions

Mechanically rewrite `assert cond[, msg]` in library code as
`if not cond: raise AssertionError[(msg)]` so the checks still fire
under `python -O`, which strips asserts. Test files are left alone
since pytest relies on `assert` for introspection.

Closes #1408.

Assisted-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: rapsealk

bitsandbytes/autograd/_functions.py

@@ -381,7 +381,8 @@ def matmul_4bit(
     out: Optional[torch.Tensor] = None,
     bias: Optional[torch.Tensor] = None,
 ):
-    assert quant_state is not None
+    if quant_state is None:
+        raise AssertionError
     if A.device.type == "cpu":
         if getattr(quant_state, "packing_format_for_cpu", False):
             out = F.gemv_4bit(A, B, out, state=quant_state)

bitsandbytes/diagnostics/main.py

@@ -39,7 +39,8 @@ def sanity_check():
     loss.backward()
     adam.step()
     p2 = p.data.sum().item()
-    assert p1 != p2
+    if p1 == p2:
+        raise AssertionError
 
 
 def get_package_version(name: str) -> str:

bitsandbytes/functional.py

@@ -99,7 +99,8 @@ def get_paged(*shape, dtype=torch.float32, device=FIRST_CUDA_DEVICE):
 
 
 def prefetch_tensor(A: torch.Tensor, to_cpu=False):
-    assert A.is_paged, "Only paged tensors can be prefetched!"
+    if not A.is_paged:
+        raise AssertionError("Only paged tensors can be prefetched!")
     if to_cpu:
         deviceid = -1
     else:
@@ -218,7 +219,8 @@ def create_normal_map(offset=0.9677083, use_extra_value=True):
     values = values.sort().values
     values /= values.max()
 
-    assert values.numel() == 256
+    if values.numel() != 256:
+        raise AssertionError
 
     return values
 
@@ -254,7 +256,8 @@ def create_fp8_map(signed=True, exponent_bits=5, precision_bits=2, total_bits=8)
     e = exponent_bits
     p = precision_bits
     has_sign = 1 if signed else 0
-    assert e + p == total_bits - has_sign
+    if e + p != total_bits - has_sign:
+        raise AssertionError
     # the exponent is biased to 2^(e-1) -1 == 0
     evalues = []
     for i, val in enumerate(range(-(2 ** (exponent_bits - has_sign)), 2 ** (exponent_bits - has_sign), 1)):
@@ -279,7 +282,8 @@ def create_fp8_map(signed=True, exponent_bits=5, precision_bits=2, total_bits=8)
             if signed:
                 values.append(-value)
 
-    assert len(values) == 2**total_bits
+    if len(values) != 2**total_bits:
+        raise AssertionError
     values.sort()
     if total_bits < 8:
         gap = 256 - len(values)
@@ -337,7 +341,8 @@ def create_dynamic_map(signed=True, max_exponent_bits=7, total_bits=8):
     data.append(0)
     data.append(1.0)
 
-    assert len(data) == 2**total_bits
+    if len(data) != 2**total_bits:
+        raise AssertionError
 
     gap = 256 - len(data)
     for i in range(gap):
@@ -516,7 +521,8 @@ def from_dict(cls, qs_dict: dict[str, Any], device: torch.device) -> "QuantState
             qs_dict.update(unpack_tensor_to_dict(qs_dict.pop(first_qs_key)))
 
         qs_dict = {k.split(".")[-1]: v for k, v in qs_dict.items()}  # strip prefixes
-        assert set(qs_dict.keys()).issubset(cls.valid_qs_keys)
+        if not set(qs_dict.keys()).issubset(cls.valid_qs_keys):
+            raise AssertionError
 
         if "nested_absmax" in qs_dict:
             offset = torch.tensor(float(qs_dict["nested_offset"])).to(device)
@@ -721,7 +727,8 @@ def dequantize_blockwise(
             The dequantized tensor. The datatype is indicated by `quant_state.dtype` and defaults to `torch.float32`.
     """
 
-    assert quant_state is not None or absmax is not None
+    if quant_state is None and absmax is None:
+        raise AssertionError
     if code is None and quant_state is None:
         if "dynamic" not in name2qmap:
             name2qmap["dynamic"] = create_dynamic_map().to(A.device)
@@ -842,7 +849,8 @@ def get_4bit_type(typename, device=None, blocksize=64):
     data = torch.tensor(data, device=device)
     data.div_(data.abs().max())
 
-    assert data.numel() == 16
+    if data.numel() != 16:
+        raise AssertionError
 
     return data
 
@@ -1009,7 +1017,8 @@ def dequantize_4bit(
         blocksize = 64
 
     if quant_state is None:
-        assert absmax is not None and out is not None
+        if absmax is None or out is None:
+            raise AssertionError
 
         quant_state = QuantState(
             absmax=absmax,
@@ -1365,7 +1374,8 @@ def igemm(
         ldc = sB[1]
     elif len(sB) == 3:
         # special case
-        assert len(sA) == 3
+        if len(sA) != 3:
+            raise AssertionError
         if not (sA[0] == sB[0] and sA[1] == sB[1]):
             raise ValueError(
                 f"Only bsi,bso->io supported for tensor contractions, but dims for A x B were: {sA} x {sB}",
@@ -1658,10 +1668,13 @@ def _convert_weight_packed_for_cpu(qweight: torch.Tensor, quant_state: QuantStat
     unpacked_w[::2] = qweight >> 4
     qweight_final = unpacked_w.reshape(quant_state.shape).to(torch.uint8)  # (*, N, K)
     # pack weight: [*, N, K] -> [*, N, K/2] combine low and high bit
-    assert len(qweight_final.shape) == 2
+    if len(qweight_final.shape) != 2:
+        raise AssertionError
     N, K = qweight_final.shape[0], qweight_final.shape[1]
-    assert N % block_n == 0, "N must be divisible by block_n"
-    assert K % 2 == 0, "K must be even"
+    if N % block_n != 0:
+        raise AssertionError("N must be divisible by block_n")
+    if K % 2 != 0:
+        raise AssertionError("K must be even")
     BLOCK_N = block_n
     BIT_COUNT = 32  # (=32 low +32 high)
     new_shape = [N // BLOCK_N, BLOCK_N, K // 2, 2]
@@ -1706,18 +1719,23 @@ def _convert_weight_packed_for_cpu_inverse(
         qweight: [*, N, K] uint8, original qweight shape (quant_state.shape)
         recovered_state: QuantState with partially restored fields (best-effort inverse)
     """
-    assert quant_state.packing_format_for_cpu, "only for packing format"
-    assert packed_weight.dtype == torch.uint8
-    assert len(packed_weight.shape) == 2, "packed_weight should be [N, K/2]"
+    if not quant_state.packing_format_for_cpu:
+        raise AssertionError("only for packing format")
+    if packed_weight.dtype != torch.uint8:
+        raise AssertionError
+    if len(packed_weight.shape) != 2:
+        raise AssertionError("packed_weight should be [N, K/2]")
     N, K_half = packed_weight.shape
     K = K_half * 2
 
     # 1) packed [N, K/2] -> [N//BLOCK_N, BLOCK_N, K/2, 2]
     BLOCK_N = block_n
     BIT_COUNT = 32  # (=32 low + 32 high)
 
-    assert N % BLOCK_N == 0, "N must be divisible by block_n"
-    assert K % 2 == 0, "K must be even"
+    if N % BLOCK_N != 0:
+        raise AssertionError("N must be divisible by block_n")
+    if K % 2 != 0:
+        raise AssertionError("K must be even")
 
     # [N, K/2] -> [-1, 64] (32 low + 32 high)
     packed = packed_weight.reshape(-1, BIT_COUNT)  # [-1, 64]

bitsandbytes/nn/modules.py

@@ -495,7 +495,8 @@ def fix_4bit_weight_quant_state_from_module(module: Union["Embedding4bit", "Line
 
     # the quant state got lost when the parameter got converted. This happens for example for fsdp
     # since we registered the module, we can recover the state here
-    assert module.weight.shape[1] == 1
+    if module.weight.shape[1] != 1:
+        raise AssertionError
     if not isinstance(module.weight, Params4bit):
         module.weight = Params4bit(module.weight, quant_storage=module.quant_storage, bnb_quantized=True)
     module.weight.quant_state = module.quant_state
@@ -866,8 +867,10 @@ def forward(self, input: Tensor) -> Tensor:
         rows = self.weight.data
         row_stats = self.weight.SCB
 
-        assert rows.shape == (self.num_embeddings, self.embedding_dim)
-        assert row_stats.shape == (self.num_embeddings,)
+        if rows.shape != (self.num_embeddings, self.embedding_dim):
+            raise AssertionError
+        if row_stats.shape != (self.num_embeddings,):
+            raise AssertionError
 
         compressed_output = F.embedding(input, rows)
         compressed_output_stats = F.embedding(input, row_stats.view(self.num_embeddings, 1))
@@ -928,35 +931,40 @@ def __init__(
             )
 
     def _forward_with_partial_dequantize(self, input: Tensor):
-        assert self.embedding_dim % self.weight.quant_state.blocksize == 0
+        if self.embedding_dim % self.weight.quant_state.blocksize != 0:
+            raise AssertionError
 
         w_4bit_uint8 = self.weight.data.view(torch.uint8).view(self.num_embeddings * self.embedding_dim // 2, 1)
 
         output_4bit = torch.nn.functional.embedding(
             weight=w_4bit_uint8.view(self.num_embeddings, self.embedding_dim // 2),
             input=input,
         ).view(-1, 1)
-        assert output_4bit.shape == (input.numel() * self.embedding_dim // 2, 1)
+        if output_4bit.shape != (input.numel() * self.embedding_dim // 2, 1):
+            raise AssertionError
 
         blocks_per_emb = self.embedding_dim // self.weight.blocksize
 
         absmax = self.weight.quant_state.absmax
-        assert absmax.shape == (self.num_embeddings * blocks_per_emb,)
+        if absmax.shape != (self.num_embeddings * blocks_per_emb,):
+            raise AssertionError
 
         output_absmax = torch.nn.functional.embedding(
             weight=absmax.view(self.num_embeddings, blocks_per_emb),
             input=input,
         ).view(
             -1,
         )
-        assert output_absmax.shape == (input.numel() * blocks_per_emb,)
+        if output_absmax.shape != (input.numel() * blocks_per_emb,):
+            raise AssertionError
 
         output_quant_state = copy.deepcopy(self.weight.quant_state)
         output_quant_state.absmax = output_absmax
         output_quant_state.shape = torch.Size((*input.shape, self.embedding_dim))
 
         output = bnb.functional.dequantize_4bit(output_4bit, output_quant_state)
-        assert output.shape == (*input.shape, self.embedding_dim)
+        if output.shape != (*input.shape, self.embedding_dim):
+            raise AssertionError
 
         return output.to(self.dtype)
 

bitsandbytes/nn/parametrize.py

@@ -175,14 +175,16 @@ def _parametrized_state_dict_post_hook(
         clean_key = f"{prefix}{param_name}"
         state_dict[clean_key] = state_dict.pop(original_key)
 
-        assert P.is_parametrized(module, param_name)
+        if not P.is_parametrized(module, param_name):
+            raise AssertionError
 
         # Find the parametrization, which should have the quantization state.
         parametrization: Bnb4bitParametrization = next(
             filter(lambda x: isinstance(x, Bnb4bitParametrization), module.parametrizations[param_name]), None
         )
 
-        assert parametrization is not None, "Parametrization not found for the parameter."
+        if parametrization is None:
+            raise AssertionError("Parametrization not found for the parameter.")
 
         quant_state = parametrization.quant_state
 

bitsandbytes/optim/lars.py

@@ -248,7 +248,8 @@ def step(self, closure=None):
 
                 update_scale = 1.0
                 if max_unorm > 0.0:
-                    assert p.dtype == torch.float32
+                    if p.dtype != torch.float32:
+                        raise AssertionError
                     pnorm = torch.norm(p.detach())
                     unorm = torch.norm(update)
                     if unorm > max_unorm * pnorm:

bitsandbytes/optim/optimizer.py

@@ -100,7 +100,8 @@ def override_config(self, parameters, key=None, value=None, key_value_dict=None)
         if isinstance(parameters, torch.Tensor):
             parameters = [parameters]
         if key is not None and value is not None:
-            assert key_value_dict is None
+            if key_value_dict is not None:
+                raise AssertionError
             key_value_dict = {key: value}
 
         if key_value_dict is not None:
@@ -286,8 +287,10 @@ def to_gpu(self):
     def check_overrides(self):
         for module, attr, config in self.mng.module_weight_config_triple:
             pmodule = getattr(module, attr)
-            assert pmodule is not None
-            assert isinstance(pmodule, torch.Tensor) or isinstance(pmodule, torch.Parameter)
+            if pmodule is None:
+                raise AssertionError
+            if not (isinstance(pmodule, torch.Tensor) or isinstance(pmodule, torch.Parameter)):
+                raise AssertionError
             found = False
             for gindex, group in enumerate(self.param_groups):
                 if found:

bitsandbytes/utils.py

@@ -9,7 +9,8 @@
 
 
 def outlier_hook(module, input):
-    assert isinstance(module, torch.nn.Linear)
+    if not isinstance(module, torch.nn.Linear):
+        raise AssertionError
     tracer = OutlierTracer.get_instance()
     hvalue = tracer.get_hvalue(module.weight)
     if hvalue not in tracer.hvalue2outlier_idx:
@@ -20,7 +21,8 @@ def outlier_hook(module, input):
             # assign the current layer the outlier idx found from the weight
             # of the previous linear layer
             if tracer.outliers[-1].numel() > 0:
-                assert tracer.outliers[-1].max() < module.weight.shape[1]
+                if tracer.outliers[-1].max() >= module.weight.shape[1]:
+                    raise AssertionError
             tracer.hvalue2outlier_idx[hvalue] = tracer.outliers[-1]
 
         else:

check_bnb_install.py

@@ -16,6 +16,7 @@
 
 p2 = p.data.sum().item()
 
-assert p1 != p2
+if p1 == p2:
+    raise AssertionError
 print("SUCCESS!")
 print("Installation was successful!")