huggingface/candle
Documented errors, page 2 of 3. Back to huggingface/candle
| Code / Message | Type | Severity | Tags |
|---|---|---|---|
| prediction_type not implemented yet: sample | exception | error | rust, candle, stable-diffusion, scheduler, unsupported-feature |
| GroupNorm doesn't support causal evaluation. | validation | error | rust, candle, mimi, audio, streaming, incompatible-config |
| unexpected dtype for | exception | error | onnx, dtype-mismatch, eval, input-validation |
| convtr1d: shape mismatch on c_in | validation | error | metal, gpu, conv-transpose, shape-mismatch, candle |
| missing input | exception | error | onnx, input-validation, missing-value, eval |
| bool_masked_pos set without mask_token | validation | error | rust, candle, vit, vision, missing-parameter |
| both hidden_act and hidden_activation are set | validation | error | config, gemma, model-loading |
| dimension index is too large for tensor rank | validation | error | rust, pyo3, tensor, dimension-out-of-range |
| input rank for GroupNorm should be at least 3 | exception | error | tensor, shape, normalization, rust |
| swiglu-multiple-of has to be set | validation | error | rust, candle, metavoice, config, missing-field |
| gguf: header declares | exception | error | gguf, truncated-file, file-format |
| is a dummy type and does not support storage | panic | error | rust, panic, unsupported-dtype, storage, candle |
| Unsupported activation function | validation | error | rust, candle, voxtral, config, unsupported-feature |
| backward not supported for upsample_bilinear2d | exception | error | autograd, unsupported-op, tensor, rust |
| strides have to be the same on both axis | validation | error | onnx, conv2d, strides |
| unsupported type for ' ' attribute in ' ' for | exception | error | rust, candle, onnx, attribute, type-mismatch |
| attribute of type TENSOR has an unsupported data_type | exception | error | rust, candle, onnx, dtype, unsupported |
| not a f64 | exception | error | candle, gguf, type-mismatch |
| Unsupported projector activation | validation | error | rust, candle, voxtral, config, unsupported-feature |
| Reshape: at most one dimension of the target shape can be -1 | exception | error | onnx, reshape, shape-mismatch, eval |
| expected numeric value for token type/id, got | validation | error | candle, gguf, tokenizer, metadata |
| gguf: tensor_count exceeds max | exception | error | gguf, corrupt-file, file-format, security |
| Input size is too large for the position embedding | exception | error | rust, candle, mmdit, diffusion, resolution-limit |
| block_table must be a cuda tensor | validation | error | cuda, flash-attention, device-mismatch, paged-attention |
| If node is malformed: branch outputs ( ) don't match node… | validation | error | onnx, if, control-flow, malformed-graph |
| new_decoder_architecture is not supported | validation | error | config, falcon, unsupported-architecture, candle |
| INTERNAL ERROR inconsistent lhs and rhs | error_code | error | streaming, invariant-violation, internal |
| ScatterND expects k (indices.shape[-1]) to be at most the… | validation | error | onnx, scatternd, shape |
| is a dummy type and cannot be converted | panic | error | rust, panic, unsupported-dtype, candle |
| cannot find the ' ' attribute in ' ' for | exception | error | rust, candle, onnx, missing-attribute |
| dtype mismatch | exception | error | dtype, cat, tensor, rust |
| `k` and `v` head dims must match | validation | error | sdpa, shape-mismatch, gqa, metal |
| more strides than expected in conv2d | validation | error | onnx, conv2d, strides, attribute-arity |
| quantized embedding requires contiguous ids | exception | error | cuda, quantization, embedding, layout, contiguity |
| Trilu expects input with at least 2 dimensions | validation | error | onnx, trilu, shape |
| batch size mismatch between inp | exception | error | loss, shape, batching, rust |
| Metal contiguous unary | exception | error | metal, gpu, unary-op, dtype, candle |
| flash-attn is only supported for f16/bf16 | validation | error | cuda, flash-attention, dtype, precision |
| flash-attn-v3-varlen expects input tensors of rank 3 | validation | error | shape, rank-mismatch, flash-attention, varlen |
| only kv-repeat = 1 is supported | exception | error | rust, candle, mimi, transformer, kv-cache |
| seqlens_k has to be contiguous | validation | error | cuda, flash-attention, contiguity, varlen |
| unexpected shape for input | exception | error | cuda, quantization, matmul, shape, rank |
| No data in file | exception | error | imatrix, empty-file, parsing |
| Reshape: -1 cannot be inferred when another dimension is… | exception | error | onnx, reshape, shape-inference |
| different inner dimensions in broadcast matmul | validation | error | candle, matmul, shape, dimension-mismatch |
| dim is odd | validation | error | rope, flux, tensor-shape, candle |
| flash-attn-varlen paged expects k/v tensors of rank 4 | validation | error | cuda, flash-attention, paged-attention, tensor-shape |
| Invalid nval for entry | exception | error | imatrix, corrupt-file, parsing |
| GroupNorm: num_groups | exception | error | config, normalization, validation, rust |
| not a string | exception | error | candle, gguf, type-mismatch |
| empty last-dim in arg-sort | validation | error | shape, argsort, empty-dim |
| input and target must have the same shape, got inp | exception | error | loss, shape, regression, rust |
| error in prelu: unexpected number of channels for the… | exception | error | candle-nn, prelu, shape-mismatch, channels |
| not a vec | exception | error | candle, gguf, type-mismatch |
| the target tensor should have a single dimension | exception | error | loss, shape, tensor, rust |
| Video embedding length | validation | error | rust, candle, qwen3-vl, video, shape-mismatch |
| Metal strided to_dtype | exception | error | metal, gpu, dtype-cast, strided, candle |
| beta has to be contiguous | exception | error | cpu, contiguity, layernorm, weights |
| mismatch on matmul dim | exception | error | cuda, quantization, matmul, shape-mismatch, dimension |
| cannot reshape tensor with | validation | error | candle, reshape, shape |
| SpectralNorm is not supported yet. | validation | error | rust, candle, mimi, audio, unsupported-feature |
| unsupported op_type for op | exception | error | onnx, unsupported-operator |
| Unsupported resize mode | validation | error | onnx, resize, unsupported-feature |
| the accelerate backend does not support f16 matmul | exception | error | cpu-backend, dtype, matmul, f16, rust |
| attribute of type TENSOR was an invalid data_type number | exception | error | rust, candle, onnx, dtype |
| layer_types length does not match num_hidden_layers | validation | error | config, granitemoe, validation |
| query `n_heads` must be a multiple of `n_kv_heads` | validation | error | sdpa, gqa, metal, attention |
| slice-assign requires input with the same rank as there are… | validation | error | rust, candle, slice-assign, ranges |
| input has to be contiguous | validation | error | candle, argsort, contiguity, cpu |
| Invalid quantize source storage locations: not on cpu | exception | error | device-mismatch, quantized, cpu |
| Metal strided unary not implemented | exception | error | metal, gpu, unary-op, strided, dtype, candle |
| seanet lstm is not supported | exception | error | rust, candle, mimi, audio, unsupported-config |
| image is too large ( , ), maximum size | validation | error | image-size, preprocessing, input-validation, candle |
| in_channel mismatch between input | exception | error | shape-mismatch, convolution, groups, rust |
| the target tensor should have two dimensions | exception | error | loss, shape, tensor, rust |
| Wrong shape for input_ids or attention_mask | validation | error | rust, candle, bert, attention-mask, shape-mismatch |
| cross_entropy expects an input tensor of rank 2 | exception | error | loss, shape, tensor, rust |
| Expand: incompatible shapes for broadcast | validation | error | onnx, expand, broadcast, shape |
| in_channel is not divisible by the number of groups | exception | error | shape-mismatch, convolution, groups, rust |
| input rank ( ) must be >= weight rank ( ) | exception | error | metal, quantized, shape-mismatch, matmul |
| Metal where_cond not implemented | exception | error | metal, gpu, where-cond, dtype, candle |
| page_block_size does not match k shape | validation | error | cuda, flash-attention, paged-attention, shape-mismatch, config |
| cannot copy kv-caches as the transformers have different… | exception | error | rust, candle, mimi, kv-cache, state-mismatch |
| dimension index is too low for tensor rank | validation | error | rust, pyo3, tensor, negative-dimension |
| Invalid where: different dtypes for values | validation | error | metal, gpu, where-cond, dtype-mismatch, candle |
| Non contiguous softmax-last-dim is not implemented | exception | error | metal, softmax, contiguity, gpu, candle |
| unexpected shape for qkv | exception | error | rust, candle, mixformer, tensor-shape, rotary-embedding |
| Expected CPU | exception | error | device, cpu-only, kv-cache, candle |
| all q, k, v dtypes must match. | validation | error | sdpa, dtype-mismatch, metal, candle |
| not a bool | exception | error | candle, gguf, type-mismatch |
| Reshape: invalid dimension | exception | error | onnx, reshape, shape-mismatch |
| rmsnorm is not implemented for | exception | error | metal, rmsnorm, dtype, gpu, candle |
| unexpected rhs shape in dmmv | exception | error | cuda, quantization, matmul, shape, rank |
| replication-pad with a size of | exception | error | padding, unsupported-operation, candle |
| axis is too small, tensor rank | validation | error | rust, candle, axis, negative-index |
| input2 has to be contiguous | exception | error | cpu, contiguity, custom-op, candle |
| Slice node is invalid, expected 3-5 inputs, got | validation | error | onnx, slice, unsupported-input, opset |
| empty concat | validation | error | onnx, concat, malformed-model |
| gguf: tensor ' ' has dimensions, max is | exception | error | gguf, corrupt-file, file-format |
| only 2d matrixes are supported | validation | error | candle, matmul, shape, rank |