huggingface/candle
Documented errors, page 3 of 3. Back to huggingface/candle
| Code / Message | Type | Severity | Tags |
|---|---|---|---|
| is a dummy type and does not support operations | panic | error | rust, panic, arithmetic, unsupported-dtype, candle |
| attribute of type TENSOR has a negative dimension, which is… | exception | error | rust, candle, onnx, shape, dimensions |
| image width is not a multiple of patch width | validation | error | rust, candle, vision, beit, shape-constraint |
| rope_thd is not implemented for | exception | error | candle, rope, metal, dtype, unsupported-kernel |
| pixel_values_videos require video_grid_thw | validation | error | rust, candle, qwen3-vl, video, missing-argument |
| input is not a f32 tensor | exception | error | metal, dtype, gpu, custom-op |
| shape mismatch alibi_slopes | validation | error | cuda, flash-attention, shape-mismatch, tensor-shape |
| upcasting is not supported | validation | error | rust, candle, bigcode, dtype, f16 |
| unsupported 'value' type | validation | error | onnx, constant, attribute-type |
| conv-block is not supported | exception | error | rust, candle, mimi, transformer, unsupported-config |
| only 2d MaxPool is supported, strides | validation | error | onnx, maxpool, unsupported-op |
| only alibi is supported as a position-embedding-type | validation | error | config, jina-bert, position-embedding |
| n_head_kv is not supported | validation | error | config, falcon, grouped-query-attention, candle |
| is a dummy type and cannot be converted to scalar | panic | error | rust, panic, unsupported-dtype, candle |
| layernorm is not implemented for | exception | error | metal, dtype, layernorm, mixed-precision |
| alpha has to be contiguous | exception | error | cpu, rmsnorm, contiguity, candle |
| Meta SDPA does not support q head dim | validation | error | sdpa, metal, head-dim, unsupported |
| two elements have different shapes | validation | error | cpu, tensor-construction, shape-mismatch |
| kernel-size is larger than the input size | validation | error | shape, avgpool, cnn, kernel-size |
| gguf: metadata_kv_count | exception | error | gguf, corrupt-file, file-format, security |
| Invalid quantize storage locations do not match | exception | error | device-mismatch, quantized, cpu, metal, cuda |
| more dilations than expected in conv2d | validation | error | onnx, conv, unsupported-rank |
| mamba layers are not yet supported in GraniteMoeHybrid… | validation | error | granitemoe, unsupported-feature, mamba |
| metal col2im1d not implemented | exception | error | metal, gpu, conv-transpose, col2im, dtype, candle |
| Metal device does not yet support F8E4M3. | error_code | error | candle, metal, argsort, fp8, unsupported |
| causal::run_causal_attn_cpu is B=1 only | exception | error | candle-nn, cpu, attention, batch-size, varlen |
| norm not supported for integer dtypes | validation | error | dtype, norm, integer |
| kv_cache_enabled=true is not supported | validation | error | rust, candle, metavoice, kv-cache, unsupported-feature |
| softmax-last-dim is not implemented for | exception | error | metal, softmax, dtype, gpu, candle |
| Unsupported MM projector type | validation | error | config, llava, multimodal |
| unsupported 'mode' value | validation | error | onnx, pad, unsupported-attribute, limitation |
| unsupported number of inputs | validation | error | onnx, pad, unsupported-input, opset |
| Qwen3 does not support f64; load weights as f32 or bf16… | validation | error | rust, candle, dtype, f64, qwen3 |
| Embedding dimension must be even | exception | error | rust, candle, mmdit, diffusion, tensor-shape |
| no graph defined in proto | exception | error | onnx, model-loading, missing-field, protobuf |
| paged flash-attn requires page_block_size to be a multiple… | validation | error | cuda, flash-attention, paged-attention, constraint |
| `q` and `k` last dims must match | validation | error | sdpa, shape-mismatch, metal, attention |
| step cannot be zero | validation | error | arange, zero-division, range |
| unsupported dtype for rmsnorm | exception | error | dtype, rmsnorm, cpu, candle |
| unsupported tokenizer model | validation | error | candle, gguf, tokenizer, unsupported |
| SDPA has no cpu impl | exception | error | cpu, sdpa, unimplemented, candle |
| two elements have different len | exception | error | cpu, tensor-construction, shape-mismatch |
| not implemented | exception | error | not-implemented, quantized, cuda, metal, cpu |
| empty array | exception | error | cpu, tensor-construction, empty-input |
| input3 has to be contiguous | exception | error | cpu, contiguity, custom-op, candle |
| weight rank ( ) must be <= 4 | exception | error | metal, quantized, rank-limit |
| seqlens_q must be a cuda tensor | validation | error | cuda, flash-attention, device-mismatch, varlen |
| block_table must be u32 or i32, got | exception | error | cuda, flash-attn, dtype, paged-attention, i32 |
| unsupported auto_pad | exception | error | onnx, maxpool, unsupported-op |
| only exclusive == 0 is supported in CumSum | validation | error | onnx, cumsum, unsupported-attribute, limitation |
| Metal conv1d not implemented | exception | error | metal, gpu, conv1d, dtype, candle |
| shape mismatch k and v | validation | error | cuda, flash-attention, paged-attention, shape-mismatch |
| is odd | validation | error | timestep-embedding, flux, tensor-shape, candle |
| none of hidden_act and hidden_activation are set | validation | error | config, gemma, missing-field |