ErrLookup › huggingface/transformers
huggingface/transformers
🤗 Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training. · Python · 3,011 source files
Analyzed at a597f97485 on 2026-08-14. 541 documented errors.
| Code / Message | Type | Severity | Tags |
|---|---|---|---|
| TP and DP cannot be used together | exception | error | bitsandbytes, quantization, hardware, cuda, installation |
| No baseline with name '{name}' in {RESULTS_DIR} | exception | error | deepgemm, kernels, cuda, torch, installation, network |
| Tensor parallelism was requested, but WORLD_SIZE is not set | exception | error | deepgemm, kernels, version-conflict, installation |
| Generated {results.size(-1)} tokens, expected {config.num_to | exception | error | deepgemm, fp4, gpu-architecture, hopper, blackwell |
| No benchmark was run successfully | exception | error | deepgemm, fp8, scale-format, ue8m0, blackwell |
| PUSH_TO_HUB_TOKEN is not set, cannot push results to the Hub | exception | error | deepgemm, tensor-shape, scale-factor, validation |
| All of the arguments --batch-size, --sequence-length, and -- | exception | error | deepgemm, fp8, block-quantization, checkpoint-config |
| --num_tokens_to_generate arguments should be larger than 1 | exception | error | deepgemm, fp8, block-size, checkpoint-config |
| Unsupported config file format: {args.config_file} | exception | error | deepgemm, fp8, static-quantization, activations |
| min should be < max (got min: {min}, max: {max}) | exception | error | deepgemm, dtype, bfloat16, activations |
| function {activation_string} not found in ACT2FN mapping {li | exception | error | deepgemm, bfloat16, moe, dtype |
| not supported filetype | exception | error | deepgemm, multi-gpu, device-map, moe, parallelism |
| Incorrect audio source. Must be a valid URL starting with `h | exception | error | deepgemm, moe, static-quantization, fp8 |
| Incorrect format used for `audio`. Should be a numpy array o | exception | error | deepgemm, bfloat16, moe, dtype |
| Unknown backend {backend!r}; expected 'auto', 'torchcodec', | exception | error | deepgemm, megamoe, tensor-shape, moe, model-config |
| The audio source is a '{filetype}' file, which librosa canno | exception | error | deepgemm, megamoe, fp4, checkpoint-config, moe |
| Invalid return_format: {return_format}. Must be 'base64', 'd | exception | error | deepspeed, trainer, lifecycle, dtype |
| File not found: {audio} | exception | error | deepspeed, zero, config, hidden-size, training |
| Error loading audio: {e} | exception | error | deepspeed, training, config-mismatch, trainer |
| Invalid input type. Must be a single audio or a list of audi | exception | error | audio, input-validation, type-error |
| mel_scale should be one of "htk", "slaney" or "kaldi". | exception | error | audio, mel-spectrogram, argument-validation |
| norm must be one of None or "slaney" | exception | error | audio, mel-spectrogram, argument-validation |
| Require num_frequency_bins: {num_frequency_bins} >= 2 | exception | error | audio, mel-spectrogram, argument-validation |
| Require min_frequency: {min_frequency} <= max_frequency: {ma | exception | error | audio, mel-spectrogram, argument-validation |
| Unknown window function '{name}' | exception | error | audio, stft, argument-validation |
| Length of the window ({window_length}) may not be larger tha | exception | error | audio, stft, argument-validation |
| frame_length ({frame_length}) may not be larger than fft_len | exception | error | audio, stft, argument-validation |
| Length of the window ({window_length}) must equal frame_leng | exception | error | audio, stft, argument-validation |
| hop_length must be greater than zero | exception | error | audio, stft, argument-validation |
| Input waveform must have only one dimension, shape is {wavef | exception | error | audio, stft, shape-error |
| Complex-valued input waveforms are not currently supported | exception | error | audio, stft, dtype-error |
| You have provided `mel_filters` but `power` is `None`. Mel s | exception | error | audio, mel-spectrogram, argument-validation |
| Cannot use log_mel option '{log_mel}' with power {power} | exception | error | audio, mel-spectrogram, argument-validation |
| Unknown log_mel option: {log_mel} | exception | error | audio, mel-spectrogram, argument-validation |
| reference must be greater than zero | exception | error | audio, validation, decibels, numerical |
| min_value must be greater than zero | exception | error | audio, validation, decibels, numerical |
| db_range must be greater than zero | exception | error | audio, validation, decibels, configuration |
| Stage_names must be set for transformers backbones | exception | error | backbone, configuration, validation, computer-vision |
| out_features must be a list got {type(self._out_features)} | exception | error | backbone, configuration, validation, type-error |
| out_features must be a subset of stage_names: {self.stage_na | exception | error | backbone, configuration, validation, naming |
| out_features must not contain any duplicates, got {self._out | exception | error | backbone, configuration, validation, duplicates |
| out_features must be in the same order as stage_names, expec | exception | error | backbone, configuration, validation, ordering |
| out_indices must be a list, got {type(self._out_indices)} | exception | error | config, backbone, validation, type-error |
| out_indices must be valid indices for stage_names {self.stag | exception | error | config, backbone, validation, index-error |
| out_indices must not contain any duplicates, got {self._out_ | exception | error | config, backbone, validation, duplicates |
| out_indices must be in the same order as stage_names, expect | exception | error | config, backbone, validation, ordering |
| out_features and out_indices should have the same length if | exception | error | config, backbone, validation, consistency |
| out_features and out_indices should correspond to the same s | exception | error | config, backbone, validation, consistency |
| backbone_type {self.backbone_type} not supported. | exception | error | backbone, internal, defensive, config |
| Config has `out_features` set to {out_features_from_config} | exception | error | backbone, timm, config, checkpoint-mismatch |
| Config has `stage_names` set to {stage_names_from_config} wh | exception | error | backbone, timm, config, checkpoint-mismatch |
| This method should be implemented by the derived class. | exception | error | backbone, not-implemented, abstract, subclassing |
| `crop` was called, but the current layer does not track past | exception | error | cache, kv-cache, sliding-window, generation |
| Once the sliding window size has been reached, `DynamicSlidi | exception | error | cache, kv-cache, sliding-window, generation |
| You need to install optimum-quanto in order to use KV cache | exception | error | dependencies, quantization, kv-cache, import-error |
| `nbits` for `quanto` backend has to be one of [`2`, `4`] but | validation | error | quantization, kv-cache, config, validation |
| `axis_key` for `quanto` backend has to be one of [`0`, `-1`] | validation | error | quantization, kv-cache, config, validation |
| `axis_value` for `quanto` backend has to be one of [`0`, `-1 | validation | error | quantization, kv-cache, config, validation |
| You need to install `HQQ` in order to use KV cache quantizat | exception | error | dependencies, quantization, kv-cache, import-error |
| `nbits` for `HQQ` backend has to be one of [`1`, `2`, `3`, ` | validation | error | quantization, kv-cache, config, validation |
| `axis_key` for `HQQ` backend has to be one of [`0`, `1`] but | validation | error | quantization, kv-cache, config, validation |
| `axis_value` for `HQQ` backend has to be one of [`0`, `1`] b | validation | error | quantization, hqq, cache, validation, valueerror |
| `crop` was called, but the current layer does not track past | exception | error | cache, linear-attention, generation, rollback, runtimeerror |
| Linear attention layers can only be cropped by passing a neg | exception | error | cache, linear-attention, crop, sign-convention, runtimeerror |
| You can construct a Cache either from a list `layers` of all | validation | error | cache, constructor, validation, valueerror |
| You should provide exactly one of `layers` or `layer_class_t | validation | error | cache, constructor, validation, valueerror |
| Cannot call `update_conv_state` on a non-LinearAttention lay | validation | error | cache, linear-attention, conv-state, layer-type, valueerror |
| Cannot call `update_indexer` on layer {layer_idx} which is a | validation | error | cache, indexer, layer-type, valueerror |
| `num_head` was provided as a list of length {len(num_heads)} | validation | error | cache, early-initialization, export, shape-mismatch, valueerror |
| `head_dim` was provided as a list of length {len(num_heads)} | validation | error | cache, early-initialization, export, shape-mismatch, valueerror, misleading-message |
| You called `get_seq_length` on layer index {layer_idx}, but | validation | error | cache, linear-attention, layer-type, get-seq-length, valueerror |
| `get_seq_length` can only be called on Attention layers, and | validation | error | cache, linear-attention, get-seq-length, valueerror |
| `has_previous_state` can only be called on LinearAttention l | validation | error | cache, linear-attention, has-previous-state, valueerror |
| You called `has_previous_state` on layer index {layer_idx}, | validation | error | cache, linear-attention, layer-type, valueerror |
| You called `get_mask_sizes` on layer index {layer_idx}, but | validation | error | cache, linear-attention, mask, layer-type, valueerror |
| `get_mask_sizes` can only be called on Attention layers, and | validation | error | cache, linear-attention, mask, valueerror |
| The batch size is not consistent across layers: {values} | validation | error | cache, batch-size, consistency, concat, valueerror |
| Unknown quantization backend `{backend}` | validation | error | cache, quantization, backend, valueerror |
| `QuantizedCache` is only supported for models with only full | validation | error | cache, quantization, layer-type, sliding-window, hybrid-model, valueerror |
| Expected {len(combined_cache_data) = } to be 4 or 6. {combin | validation | error | cache, encoder-decoder, ddp, data-layout, valueerror |
| One of the two arguments is not a Cache: {type(caches[0]) = | validation | error | cache, typeerror, encoder-decoder, constructor, transformers |
| Expected 1 or 2 arguments, got {len(caches)} | validation | error | cache, valueerror, encoder-decoder, arguments, transformers |
| `{method}` is only defined for dynamic cache, got {self.self | validation | error | cache, typeerror, dynamic-cache, static-cache, transformers |
| {lowercase_name} is not a valid model name | validation | error | cli, valueerror, add-new-model, model-type, developer-tooling |
| Could not find TOC entry for {old_lowercase_name} | validation | error | cli, docs, toctree, valueerror, developer-tooling |
| You need to install `libcst` to run this command -> `pip ins | validation | error | cli, missing-dependency, libcst, valueerror, developer-tooling |
| {x} is not a value that can be converted to a bool. | validation | error | cli, user-input, bool-parsing, valueerror, developer-tooling |
| Unknown error | exception | error | cli, chat, serving, model-loading, runtimeerror |
| You need to install rich to use the chat interface. (`pip in | exception | error | cli, missing-dependency, rich, importerror, chat |
| The server running on {url} returned status code {output.sta | validation | error | cli, chat, health-check, network, valueerror |
| No server currently running on {url}. To run a local server, | validation | error | cli, chat, connection-refused, network, serving |
| Failed to convert `generate_flags` into a valid JSON object. | validation | warning | cli, chat, json-parsing, generate-flags, valueerror |
| Missing dependencies for serving. Install with `pip install | exception | error | cli, serving, missing-dependency, importerror, extras |
| prompt must be a string. | http | error | serving, api, completions, validation, http-400 |
| Unsupported dtype: '{dtype}'. Must be 'auto' or a valid torc | validation | error | serving, dtype, cli, validation, valueerror |
| Unsupported quantization method: '{self.quantization}'. Must | validation | error | serving, quantization, cli, validation, valueerror |
| Unsupported attention implementation: '{self.attn_implementa | validation | error | serving, attention, cli, validation, valueerror |
| Unknown modality for: {model_classname} | exception | error | serving, model-registry, modality, valueerror, custom-model |
| 'input' must be a string or list | http | error | serving, api, responses, validation, http-422 |
| Unsupported input item type: {item_type!r} | http | error | serving, api, responses, validation, http-422 |
| Missing `model` field in the request body. | http | error | api, serving, validation, http-422 |
| Unexpected fields in the request: {unexpected} | http | error | api, serving, audio, validation, http-422 |
| Missing librosa dependency for audio transcription. Install | exception | error | dependency, audio, serving, importerror |
| Missing python-multipart dependency for file uploads. Instal | exception | error | dependency, serving, file-upload, importerror |
| Expected file upload, got string | http | error | api, serving, file-upload, validation, http-422 |
| Expected model name as string | http | error | api, serving, validation, http-422 |
| Audio transcription requires sequential generation (not CB) | http | error | serving, audio, continuous-batching, http-400 |
| CB worker is dead and cannot accept request {request_id}: {s | http | critical | serving, continuous-batching, crash, http-503 |
| CB worker died during request {request_id}: {result.error} | http | critical | serving, continuous-batching, crash, oom, http-503 |
| Unexpected fields in the request: {unexpected} | http | error | api, serving, validation, http-422 |
| Server is pinned to '{self.model_manager.force_model}'; requ | http | error | api, serving, model-pinning, http-400 |
| {cls.__name__} accepts only keyword arguments, but found `{l | validation | error | config, dataclass, validation |
| Missing required field - '{f.name}' | validation | error | config, dataclass, validation |
| `problem_type="single_label_classification"` requires `num_l | validation | error | config, classification, validation |
| The `output_attentions` attribute is not supported when usin | validation | error | config, attention, attn-implementation, validation |
| The embed_dim ({self.embed_dim}) is not a multiple of the nu | validation | error | config, architecture, validation |
| The `{layer_types}` entries must be in {allowed_types} but g | validation | error | config, architecture, layer-types, validation |
| `num_hidden_layers` ({self.num_hidden_layers}) must be equal | validation | error | config, architecture, layer-types, validation |
| Provided path ({save_directory}) should be a directory, not | exception | error | config, serialization, filesystem |
| Some generation parameters are set in the model config. Thes | validation | error | config, generation, save, migration |
| Can't load the configuration of '{pretrained_model_name_or_p | exception | error | hub, network, config, from-pretrained |
| It looks like the config file at '{resolved_config_file}' is | exception | error | json, config, corruption, from-pretrained |
| key {k} isn't in the original config dict | validation | error | config, update-from-string, validation |
| can't derive true or false from {v} (key {k}) | validation | error | config, update-from-string, boolean-parsing |
| You can only update int, float, bool or string values in the | validation | error | config, update-from-string, type-error |
| {auto_class} is not a valid auto class. | validation | error | config, auto-class, registration |
| Multiple valid text configs were found in the model config: | validation | error | config, text-config, multi-modal, ambiguity |
| Calling `get_mtp_config` on a config without `num_mtp_layers | exception | error | config, mtp, architecture |
| Calling `get_mtp_config` on a config containing `layer_types | exception | error | config, mtp, layer-types |
| The `layer_types` entries must be in {ALLOWED_LAYER_TYPES} | exception | error | config, layer-types, validation, deprecated |
| `num_hidden_layers` ({num_hidden_layers}) must be equal to t | exception | error | config, layer-types, validation, deprecated |
| Conversion mapping for '{model_type_or_class_name}' already | exception | error | conversion-mapping, registration, checkpoint-conversion |
| {error_message} requires the protobuf library but it was not | exception | error | dependencies, protobuf, tokenizer, conversion |
| You're trying to run a `Unigram` model but you're file was t | exception | error | tokenizer, sentencepiece, conversion |
| `tiktoken` is required to read a `tiktoken` file. Install it | exception | error | dependencies, tiktoken, tokenizer, conversion |
| Converting from SentencePiece and Tiktoken failed, if a conv | exception | error | tokenizer, conversion, tiktoken, fallback |
| Unrecognized tokenizer name, should be one of {list(TOKENIZE | exception | error | tooling, tokenizer, conversion, cli |
| Invalid checkpoint path: '{checkpoint}' attempts to escape ` | exception | error | security, path-traversal, tooling, tokenizer |
| Failed to convert {kwargs.get('full_layer_name')} | exception | error | weight-conversion, model-loading, chunk, checkpoint |
| Conv3dToLinear expects a 5D or 2D tensor, got {tensor.ndim}D | exception | error | weight-conversion, conv3d, model-loading, shape-mismatch |
| Cannot reshape tensor with shape {tensor.shape} into {target | exception | error | weight-conversion, conv3d, reshape, shape-mismatch |
| PermuteForRope expects a single tensor per key. | exception | error | weight-conversion, rope, permutation, model-loading |
| Expected pattern {key} in collected tensors but only found t | exception | error | weight-conversion, moe, pattern-mismatch, model-loading |
| Multiple different capturing groups found in target_patterns | exception | error | weight-conversion, regex, pattern-mismatch, programmer-error |
| Source pattern '{pattern}' contains \\1 backreference, but n | exception | error | weight-conversion, regex, backreference, programmer-error |
| Cannot assign to field {name}, you should create a new insta | exception | error | weight-conversion, immutability, api-misuse, programmer-error |
| Cannot reverse the transform with TP or quantization | exception | error | quantization, reverse-conversion, save-pretrained, unsupported-operation |
| GroupWeightRename requires N:N length matching, but found le | exception | error | weight-conversion, rename, validation, programmer-error |
| You must provide only one of `prefix_to_add` and `prefix_to_ | exception | error | weight-conversion, rename, prefix, validation, programmer-error |
| source keys={self.source_patterns}, target_patterns={self.ta | exception | error | weight-conversion, cardinality, validation, programmer-error |
| WeightConverter requires at least one operation. | exception | error | weight-conversion, validation, programmer-error |
| Framework '{return_tensors}' not recognized! | exception | error | data-collator, return-tensors, api-misuse, framework |
| You are attempting to pad samples but the tokenizer you are | exception | error | data-collator, padding, tokenizer, pad-token |
| This tokenizer does not have a mask token which is necessary | exception | error | data-collator, mlm, tokenizer, mask-token, training |
| mlm_probability should be between 0 and 1. | validation | error | data-collator, mlm, validation, probability, training |
| Whole word masking can only be used with mlm=True.If you wan | validation | error | data-collator, configuration, masked-language-modeling |
| The sum of mask_replace_prob and random_replace_prob should | validation | error | data-collator, configuration, validation, masked-language-modeling |
| mask_replace_prob should be between 0 and 1. | validation | error | data-collator, configuration, validation |
| random_replace_prob should be between 0 and 1. | validation | error | data-collator, configuration, validation |
| Worker process information is not available for seeding the | validation | error | data-collator, dataloader, multiprocessing, randomness, seed |
| This tokenizer does not have a mask token which is necessary | validation | error | data-collator, tokenizer, masked-language-modeling, sop |
| This tokenizer does not have a mask token which is necessary | validation | error | data-collator, tokenizer, permutation-language-modeling, xlnet |
| This collator requires that sequence lengths be even to crea | validation | error | data-collator, padding, sequence-length, permutation-language-modeling |
| return_tensors must be one of ("pt", "np"), {return_tensors= | validation | error | data-collator, sequence-packing, return-tensors, configuration |
| mode is not a valid split name | validation | error | glue, deprecated, dataset, split-name, keyerror |
| mode is not a valid split name | validation | error | squad, deprecated, dataset, split-name, keyerror |
| Predictions and labels have mismatched lengths {len(preds)} | validation | error | metrics, xnli, evaluation, shape-mismatch, deprecated |
| No valid predictions | validation | error | squad, metrics, question-answering, evaluation |
| PyTorch must be installed to return a PyTorch dataset. | exception | error | squad, environment, pytorch, optional-dependency, question-answering |
| SquadProcessor should be instantiated via SquadV1Processor o | validation | error | squad, processor, subclassing, question-answering, configuration |
| Text and labels have mismatched lengths {len(texts_or_text_a | validation | error | data-processing, validation, length-mismatch |
| Text and ids have mismatched lengths {len(texts_or_text_and_ | validation | error | data-processing, validation, length-mismatch, ids |
| Error with input length {len(input_ids)} vs {batch_length} | validation | error | tokenization, padding, max-length, truncation |
| Error with input length {len(attention_mask)} vs {batch_leng | validation | error | tokenization, padding, attention-mask, max-length |
| return_tensors set to 'pt' but PyTorch can't be imported | exception | error | pytorch, environment, dependencies, runtime |
| return_tensors should be `'pt'` or `None` | validation | error | api-misuse, validation, tensors |
| Training input {text_a} is not a string | validation | error | xnli, data-processing, file-format, validation |
| Training input {text_b} is not a string | validation | error | xnli, data-processing, file-format, validation |
| Training label {label} is not a string | validation | error | xnli, data-processing, labels, validation |
| DebugUnderflowOverflow: inf/nan detected, aborting as there | exception | error | debugging, numerical-stability, mixed-precision, nan |
| DebugUnderflowOverflow: aborting after {self.batch_number} b | exception | warning | debugging, intentional-abort, training-loop |
| can't find {pkg} in {deps.keys()}, check dependency_versions | exception | error | dependencies, version-check, contributor, configuration |
| FSDP+TP is not supported yet. Use DistributedConfig(fsdp_siz | exception | error | distributed, fsdp, tensor-parallelism, configuration |
| {type(model).__name__} does not have a FSDP2 plan declared. | exception | error | distributed, fsdp2, model-integration, sharding |
| Expert parallelism was requested (`enable_expert_parallel=Tr | exception | error | distributed, expert-parallelism, moe, tensor-parallelism |
| Can only set a dictionary as `tp_plan` | exception | error | distributed, tensor-parallelism, type-validation, api-misuse |
| Unsupported tensor parallel style '{parallel_style}' for lay | exception | error | tensor-parallel, distributed, config, validation |
| Can only set a dictionary as `pp_plan` | exception | error | pipeline-parallel, distributed, config, type-error |
| tp_size ({distributed_config.tp_size}) * fsdp_size ({distrib | exception | error | distributed, fsdp, tensor-parallel, world-size, config |
| save_pretrained(..., distributed_checkpoint=True) requires t | exception | error | torch-version, distributed, checkpointing, environment |
| save_pretrained(..., distributed_checkpoint=True) is only su | exception | error | fsdp, distributed, checkpointing, config |
| save_pretrained(..., distributed_checkpoint=True) requires t | exception | error | fsdp, distributed, device-mesh, initialization |
| Saving an FSDP-wrapped model requires torch.distributed to b | exception | error | fsdp, distributed, checkpointing, process-group |
| Current shard-on-read only supports disjoint ranges on a sin | exception | error | tensor-parallel, sharding, checkpoint-loading, distributed |
| We tried to initialize torch.distributed for you, but it fai | exception | error | distributed, process-group, environment, launch |
| FSDP2 requires `torch>=2.7` (distributed checkpoint save/loa | exception | error | fsdp, torch-version, distributed, environment |
| Distributed checkpointing requires `torch>=2.7`. | exception | error | torch-version, fsdp, checkpointing, distributed |
| This modeling file requires the following packages that were | exception | error | dependencies, trust-remote-code, environment, import |
| Loading this model requires you to execute custom code conta | exception | error | trust-remote-code, security, interactive, timeout |
| {error_message} You can inspect the repository content at ht | exception | error | trust-remote-code, windows, security, platform |
| Missing requirements in your local environment for `{path_or | exception | error | dependencies, requirements, trust-remote-code, environment |
| export_config_dict must contain key 'export_format' set to e | exception | error | export, config, validation |
| Unknown exporter type, got {name} - supported exporters are: | exception | error | export, config, registry, validation |
| Unsupported export config: {export_config_dict!r}. Registere | exception | error | export, config, registry, auto-class |
| AutoHfExporter.from_pretrained is not implemented yet. Load/ | exception | error | export, not-implemented, auto-class |
| Exporter must extend HfExporter | exception | error | export, registry, plugin, type-error |
| Export config must extend ExportConfigMixin | exception | error | export, registry, plugin, type-error |
| To use {type(self).__name__}, please install the following d | exception | error | export, dependencies, import-error, environment |
| {type(self).__name__} requires newer versions of: {', '.join | exception | error | export, dependencies, version-mismatch, environment |
| {type(self).__name__} does not implement `export`. Pick a co | exception | error | export, not-implemented, abstract-method, plugin |
| Per-component `config` dict is missing entries for: {sorted( | exception | error | export, generation, config, validation |
| {type(self).__name__}.export failed on component '{name}' (s | exception | error | export, generation, wrapper-exception, debugging |
| Expected config to be a DynamoConfig or dict, got {type(conf | exception | error | export, dynamo, type-error, config |
| Cannot flatten a bound method for pytree context | exception | error | export, dynamo, pytree, model-limitation |
| Cannot flatten {type(obj).__name__} for pytree context | exception | error | export, dynamo, pytree, serialization |
| Expected config to be an ExecutorchConfig or dict, got {type | exception | error | export, executorch, type-error, config |
| Unsupported backend {config.backend} for ExecuTorch export | exception | error | export, executorch, backend, config |
| CUDA is not available in this environment; cannot export to | exception | error | export, executorch, cuda, environment, hardware |
| 0 in strides is not supported for ExecuTorch. | exception | error | export, executorch, strides, tensor-ops |
| Expected config to be an OnnxConfig or dict, got {type(confi | exception | error | export, onnx, type-error, config |
| _aten_grouped_mm: number of experts (mat_b.shape[0]) must be | exception | error | export, onnx, moe, dynamic-shapes |
| Found '{label_key}' in inputs. Loss computation is not suppo | exception | error | export, inputs, training-data, validation |
| Found 'model.config.return_loss=True'. Loss computation is n | exception | error | export, config, loss, transformers |
| Found 'return_loss=True' in inputs. Loss computation is not | exception | error | export, inputs, loss, transformers |
| decompose_prefill_decode failed for {type(model).__name__}. | exception | error | export, generation, inputs, transformers |
| decompose_prefill_decode expected at least {num_new_tokens} | exception | error | export, generation, architecture, transformers |
| decompose_multimodal found no multi-modal submodules on {typ | exception | error | export, multimodal, model-structure, transformers |
| decompose_multimodal failed for {type(model).__name__}. Inpu | exception | error | export, multimodal, inputs, transformers |
| You should supply an instance of `transformers.BatchFeature` | exception | error | feature-extractor, padding, audio, transformers |
| type of {first_element} unknown: {type(first_element)}. Shou | exception | error | feature-extractor, dtype, audio, transformers |
| Some items in the output dictionary have a different batch s | exception | error | feature-extractor, batching, padding, transformers |
| Invalid padding strategy:{padding_side} | exception | error | feature-extractor, padding, config, transformers |
| When setting ``truncation=True``, make sure that ``max_lengt | exception | error | feature-extractor, truncation, audio, transformers |
| When setting ``padding={PaddingStrategy.MAX_LENGTH}``, make | exception | error | feature-extractor, padding, audio, transformers |
| Asking to pad but the feature_extractor does not have a padd | exception | error | feature-extractor, padding, config, transformers |
| only a single or a list of entries is supported but got type | exception | error | feature-extractor, audio, type-error, transformers |
| Indexing with integers is not available when using Python ba | exception | error | feature-extractor, batchfeature, api-misuse, transformers |
| Unable to convert output to PyTorch tensors format, PyTorch | exception | error | feature-extractor, environment, pytorch, transformers |
| Unable to create tensor for '{key}' with overflowing values | exception | error | feature-extractor, batchfeature, tensor-conversion, audio, transformers |
| Unable to convert output '{key}' (type: {type(value).__name_ | exception | error | feature-extractor, batchfeature, tensor-conversion, transformers |
| Attempting to cast a BatchFeature to type {str(arg)}. This i | exception | error | feature-extractor, batchfeature, dtype, device, transformers |
| Provided path ({save_directory}) should be a directory, not | exception | error | feature-extractor, save, filesystem, transformers |
| Can't load feature extractor for '{pretrained_model_name_or_ | exception | error | python, transformers, feature-extractor, hub-download, from-pretrained |
| Can't load feature extractor for '{pretrained_model_name_or_ | exception | error | python, transformers, feature-extractor, missing-config, auto-class |
| {auto_class} is not a valid auto class. | exception | error | python, transformers, feature-extractor, auto-class, validation |
| Model {cls.__name__} has no config class or model type | exception | error | python, transformers, fusion, model-config, model-type |
| Fusion {fusion_name} for model type {model_type} conflicts w | exception | error | python, transformers, fusion, weight-conversion, conflict |
| Unknown fusion type: {fusion_name} | exception | error | python, transformers, fusion, config-validation |
| Invalid fusion config for {fusion_name}: expected `True`, `F | exception | error | python, transformers, fusion, config-validation, type-error |
| {} is an abstract class. Only classes inheriting this class | exception | error | python, transformers, generation, assisted-decoding, abstract-class |
| {} is an abstract class. Only classes inheriting this class | exception | error | python, transformers, generation, assisted-decoding, abstract-class |
| Invalid max_matching_ngram_size or num_output_tokens | validation | error | python, transformers, generation, prompt-lookup, validation |
| Expected assistant_model to be a Gemma4AssistantForCausalLM | validation | error | python, transformers, generation, speculative-decoding, gemma4, assistant-model |
| `model_outputs` cannot be None, and they need to contain `hi | exception | error | python, transformers, generation, speculative-decoding, gemma4, hidden-states |
| Could not find `num_mtp_layers` in the model config. This mo | validation | error | python, transformers, generation, mtp, speculative-decoding, model-config |
| `model_outputs` cannot be None, and they need to contain `hi | exception | error | python, transformers, generation, mtp, hidden-states |
| `model_outputs` cannot be None and they need to contain `hid | exception | error | python, transformers, generation, diffusion, hidden-states |
| `early_stopping` must be a boolean or 'never', but is {}. | validation | error | python, transformers, generation, generation-config, validation |
| `max_new_tokens` must be greater than 0, but is {}. | validation | error | python, transformers, generation, generation-config, validation |
| `assistant_ensemble_weight` must be in the open interval `(0 | validation | error | python, transformers, generation, speculative-decoding, generation-config |
| Invalid `cache_implementation` ({}). Choose one of: {} | validation | error | python, transformers, generation, kv-cache, generation-config, validation |
| You provided `compile_config` as an instance of {}, but it m | validation | error | python, transformers, generation, torch-compile, generation-config, type-error |
| Greedy methods (do_sample != True) without beam search do no | validation | error | generation, generation-config, decoding, validation |
| `num_return_sequences` ({}) has to be smaller or equal to `n | validation | error | generation, beam-search, generation-config, validation |
| Argument `{}` is not a valid argument of `GenerationConfig`. | validation | error | generation-config, api-misuse, validation |
| GenerationConfig is invalid: {} | validation | error | generation-config, validation, save, strict-mode |
| {} Fix these issues to save the configuration. | validation | error | generation-config, save, validation |
| Provided path ({}) should be a directory, not a file | exception | error | generation-config, save, filesystem, path |
| Can't load the configuration of '{}'. If you were trying to | exception | error | generation-config, huggingface-hub, network, loading |
| It looks like the config file at '{}' is not a valid JSON fi | exception | error | generation-config, json, cache, loading, corruption |
| Some of the keys in `watermarking_config` are defined incorr | validation | error | watermarking, generation-config, validation |
| num_key_value_heads or num_attention_heads could not be foun | validation | error | continuous-batching, kv-cache, model-config, gqa |
| head_dim or (hidden_size and num_attention_heads) could not | validation | error | continuous-batching, kv-cache, model-config, head-dim |
| Block size must be at least {}, but got {} | validation | error | continuous-batching, kv-cache, block-size, flash-attention |
| Number of key value heads {} must be divisible by tensor par | validation | critical | continuous-batching, tensor-parallelism, multi-gpu, gqa, kv-cache |
| Invalid group type: {} | exception | error | continuous-batching, kv-cache, model-config, attention, validation |
| Failed to allocate {} blocks for request {} | exception | critical | continuous-batching, kv-cache, internal-invariant, concurrency, memory |
| flash_attn_with_kvcache_fn does not have a block_table or pa | exception | error | continuous-batching, flash-attention, version-mismatch, kv-cache, environment |
| m must be provided if max_batch_tokens and num_blocks are No | validation | error | continuous-batching, memory-planning, configuration, validation |
| Memory footprint {} is more than available memory {} | exception | critical | memory, continuous-batching, kv-cache, configuration |
| Invalid values: max_batch_tokens = {}, num_blocks = {} | validation | critical | memory, continuous-batching, configuration, solver |
| No real solution (discriminant = {}) | exception | critical | memory, solver, continuous-batching, math |
| No positive solution (root = {}) | exception | critical | memory, solver, continuous-batching, math |
| logit_processor_kwargs['{key}'] has type {type(value).__name | validation | error | validation, logits-processor, types, continuous-batching |
| Unknown logit_processor_kwargs: {unknown_keys}. {self.suppor | validation | error | validation, logits-processor, generation-config, continuous-batching |
| No requests can be scheduled and no requests can be offloade | exception | critical | runtime, scheduling, memory, continuous-batching, deadlock |
| Model must have 'config', 'device', and 'dtype' attributes. | validation | error | api-misuse, attributes, continuous-batching, model |
| A GenerationConfig must be provided or set in the model. | validation | error | generation-config, validation, continuous-batching |
| Distributed is off but received {device_mesh = }. | validation | error | distributed, device-mesh, configuration, continuous-batching |
| FSDP is not compatible with continuous batching but got {dev | validation | error | distributed, fsdp, compatibility, continuous-batching |
| Async batching requires CUDA, but {torch.cuda.is_available() | exception | error | cuda, environment, async, continuous-batching |
| cpu_offload_space=None requires psutil to auto-size the CPU | exception | error | dependencies, psutil, offloading, continuous-batching |
| Got {safety_margin = } but expected a value in [0, 1] | validation | error | validation, scheduler, configuration, continuous-batching |
| Got {max_requests_per_batch = } but expected a value >= 1 | validation | error | validation, scheduler, configuration, continuous-batching |
| Sliding window attention layers do not support block table | exception | error | not-implemented, sliding-window, attention, continuous-batching |
| {self.__class__} is an abstract class. Only classes inheriti | exception | error | abstract-class, logits-processor, api-misuse |
| Make sure that all the required parameters: {list(function_a | validation | error | logits-processor, kwargs, introspection, api-misuse |
| `min_length` has to be a non-negative integer, but is {min_l | validation | error | validation, logits-processor, types, generation-config |
| `{arg_name}` has to be a positive integer, but is {arg_value | validation | error | validation, logits-processor, types, error-message |
| `temperature` (={temperature}) has to be a strictly positive | validation | error | generation, sampling, temperature, argument-validation |
| `penalty` has to be a strictly positive float, but is {penal | validation | error | generation, repetition-penalty, sampling, argument-validation |
| `prompt_ignore_length` has to be a positive integer, but is | validation | error | generation, repetition-penalty, argument-validation |
| `top_p` has to be a float > 0 and < 1, but is {top_p} | validation | error | generation, top-p, nucleus-sampling, argument-validation |
| `min_tokens_to_keep` has to be a positive integer, but is {m | validation | error | generation, top-p, argument-validation |
| `top_k` has to be a strictly positive integer, but is {top_k | validation | error | generation, top-k, sampling, argument-validation |
| `top_h` must be in the range (0, 1]. | validation | error | generation, top-h, entropy-sampling, argument-validation |
| `min_p` has to be a float in the [0, 1] interval, but is {mi | validation | error | generation, min-p, sampling, argument-validation |
| `typical_p` has to be a float > 0 and < 1, but is {mass} | validation | error | generation, typical-sampling, typical-p, argument-validation |
| `epsilon_cutoff` has to be a float > 0 and < 1, but is {epsi | validation | error | generation, epsilon-sampling, cutoff, argument-validation |
| `min_tokens_to_keep` has to be a strictly positive integer, | validation | error | generation, epsilon-sampling, argument-validation |
| `eta_cutoff` has to be a float > 0 and < 1, but is {epsilon} | validation | error | generation, eta-sampling, cutoff, argument-validation |
| `ngram_size` has to be a strictly positive integer, but is { | validation | error | generation, no-repeat-ngram, argument-validation |
| `encoder_ngram_size` has to be a strictly positive integer, | validation | error | generation, encoder-no-repeat-ngram, seq2seq, argument-validation |
| The model vocabulary size is {vocabulary_size}, but the foll | validation | error | generation, sequence-bias, vocabulary, token-ids |
| `sequence_bias` has to be a non-empty dictionary, or non-emp | validation | error | generation, sequence-bias, argument-validation |
| `sequence_bias` has to be a dict with tuples as keys, but is | validation | error | |
| Each key in `sequence_bias` has to be a non-empty tuple of p | validation | error | |
| Each element in `sequence_bias` has to be a non-empty list o | validation | error | |
| `sequence_bias` has to be a dict with floats as values, but | validation | error | |
| `bad_words_ids` has to be a non-empty list, but is {bad_word | validation | error | |
| `bad_words_ids` has to be a list of lists, but is {bad_words | validation | error | |
| Each list in `bad_words_ids` has to be a list of positive in | validation | error | |
| `prefix_allowed_tokens_fn` returned an empty list for batch | validation | error | |
| `eos_token_id` has to be a list of positive integers, but is | validation | error | |
| `forced_decoder_ids` is deprecated in favor of `task` and `l | validation | error | |
| Require guidance scale >1 to use the classifier free guidanc | validation | error | |
| Logits should have twice the batch size of the input ids, th | validation | error | |
| `input_starting_length` has to be a non-negative integer, bu | validation | error | |
| `min_eos_p` has to be a positive float, but is {min_eos_p} | validation | error | |
| seeding_scheme has to be one of [`selfhash`, `lefthash`], bu | validation | error | |
| greenlist_ratio has be in range between 0.0 and 1.0, exclusi | validation | error | |
| Ngrams should be of shape (batch_size, num_ngrams, ngram_len | validation | error | |
| Ngrams should be of shape (batch_size, num_ngrams, ngram_len | validation | error | |
| Input ids should be of shape (batch_size, input_len), but is | validation | error | pytorch, shape-validation, generation, logits-processor |
| `guidance_top_k` has to be a strictly positive integer if gi | validation | error | classifier-free-guidance, top-k, generation, config-validation |
| Audio codebooks need at least one channel, but found {num_ch | validation | error | audio-generation, codebooks, config-validation, logits-processor |
| Expected `eos_token_id` to be a positive integer, found {eos | validation | error | audio-generation, eos-token, config-validation, logits-processor |
| StoppingCriteria needs to be subclassed | exception | error | stopping-criteria, abstract-class, generation, subclassing |
| Stop string preprocessing was unable to identify tokens matc | validation | error | stop-strings, tokenizer-vocabulary, unicode, stopping-criteria, generation |
| TextStreamer only supports batch size 1 | validation | error | streamer, batch-size, generation, text-streaming |
| TextDiffusionStreamer only supports batch size 1 | validation | error | streamer, text-diffusion, batch-size, generation |
| `{pretrained_model_name_or_path}` does not contain a `custom | exception | error | hub, custom-code, generate, remote-code, offline |
| `inputs`: {inputs}` were passed alongside {input_name} which | validation | error | generate, api-misuse, input-validation, multimodal |
| You passed `inputs_embeds` to `.generate()`, but the model c | validation | error | generate, inputs-embeds, model-architecture, custom-model |
| You passed `inputs_embeds` and `input_ids` to `.generate()`. | validation | error | generate, encoder-decoder, inputs-embeds, api-misuse |
| `bos_token_id` has to be defined when no `input_ids` are pro | validation | error | generate, bos-token, config, unconditional-generation |
| `decoder_start_token_id` expected to have length {batch_size | validation | error | generate, encoder-decoder, batch-size, decoder-start-token |
| If `is_encoder_decoder` is True, make sure that `encoder_out | validation | error | generate, encoder-decoder, beam-search, internal-api |
| Setting `assistant_ensemble_weight` requires candidate logit | validation | error | generate, assisted-generation, speculative-decoding, prompt-lookup, config-conflict |
| Expected class name to start with Gemma4 or Gemma3n. Got {se | validation | error | generate, assisted-generation, gemma, model-compatibility, mtp |
| Invalid value for `do_sample`: expected a boolean, got {type | validation | error | generation, assisted-decoding, type-validation, do-sample |
| There are one or more stop strings, either in the arguments | validation | error | generation, stop-strings, tokenizer, stopping-criteria |
| {self.__class__.__name__} only supports {supported_modes}, b | validation | error | generation, generation-mode, beam-search, model-capabilities |
| `streamer` cannot be used with beam search (yet!). Make sure | validation | error | generation, streamer, beam-search, incompatible-arguments |
| num_return_sequences has to be 1 when doing assisted generat | validation | error | generation, assisted-decoding, num-return-sequences, incompatible-arguments |
| assisted generation is not supported with stateful models, s | validation | error | generation, assisted-decoding, stateful-model, incompatible-arguments |
| The main model and the assistant don't have compatible encod | validation | error | generation, assisted-decoding, encoder-decoder, whisper, model-loading |
| `assistant_tokenizer` is not required when the main and assi | validation | error | generation, assisted-decoding, tokenizer, universal-assisted-decoding |
| The main and assistant models have different tokenizers. Ple | validation | error | generation, assisted-decoding, tokenizer, universal-assisted-decoding |
| The following `model_kwargs` are not used by the model: {unu | validation | error | generation, kwargs, typo, argument-validation |
| Input length of {input_ids_string} is {input_ids_length}, bu | validation | error | generation, max-length, truncation, context-length |
| You have modified the pretrained model configuration to cont | validation | error | generation, generation-config, legacy-migration, breaking-change |
| Passing both `cache_implementation` (used to initialize cert | validation | error | generation, cache, past-key-values, cache-implementation, conflicting-arguments |
| Passing a tuple of `past_key_values` is not supported anymor | validation | error | generation, cache, past-key-values, legacy-migration, breaking-change |
| This model does not support the quantized cache. If you want | validation | error | generation, cache, quantized-cache, encoder-decoder, unsupported-feature |
| `decoder_start_token_id` or `bos_token_id` has to be defined | validation | error | generation, encoder-decoder, decoder-start-token, config |
| {generation_mode.name.replace('_', ' ').title()} requires `t | validation | error | generation, trust-remote-code, custom-generate, deprecation, security |
| inputs or input_ids must be provided for CB generation. | validation | error | generation, continuous-batching, paged-cache, missing-input |
| inputs must be a 1D or 2D tensor, got {inputs.dim() = } | validation | error | generation, continuous-batching, paged-cache, tensor-shape |
| stopping_criteria is not supported for continuous batching. | validation | error | generation, continuous-batching, paged-cache, stopping-criteria, unsupported-feature |
| prefix_allowed_tokens_fn is not supported for continuous bat | validation | error | |
| assistant_model is not supported for continuous batching. Go | validation | error | |
| streaming is not supported for continuous batching. Got {str | validation | error | |
| `attention_mask` passed to `generate` must be 2D. | exception | error | |
| When generating with token healing, you must pass the model | exception | error | |
| `low_memory=True` is not supported after the beam search ref | exception | error | |
| {self.__class__.__name__} cannot use beam search with a cach | exception | error | |
| assisted generate requires `use_cache=True` | exception | error | |
| assisted generate is not supported with Static cache classes | exception | error | |
| assisted decoding requires a cache | exception | error | |
| assisted generate is only supported for batch_size = 1 | exception | error | |
| Cannot use prefill chunking without a cache | exception | error | |
| Must have at least `1` token to score after the first min_pr | exception | error | |
| Truthy value expected: got {v} but expected one of yes/no, t | exception | error | |
| Unresolved type detected, which should have been done with t | exception | error | |
| Only `Union[X, NoneType]` (i.e., `Optional[X]`) is allowed f | exception | error | |
| Type resolution failed for {dtype}. Try declaring the class | exception | error | |
| Some specified arguments are not used by the HfArgumentParse | exception | error | |
| Some keys are not used by the HfArgumentParser: {sorted(unus | exception | error | |
| You picked the {self.name} backend, but it is not installed. | exception | error | |
| No hyperparameter search backend available.\n - To install { | exception | error | |
| only a single or a list of entries is supported but got type | exception | error | |
| Unsupported input image type {image_type} | exception | error | |
| Pad size must contain 'height' and 'width' keys only. Got pa | exception | error | |
| Padding dimensions are negative. Please make sure that the ` | exception | error | |
| Size must contain 'height' and 'width' keys, or 'max_height' | exception | error | |
| The size dictionary must have keys 'height' and 'width'. Got | exception | error | |
| Padding dimensions are negative. Please make sure that the ` | exception | error | |
| Provided path ({save_directory}) should be a directory, not | exception | error | |
| Can't load image processor for '{pretrained_model_name_or_pa | exception | error | |
| {auto_class} is not a valid auto class. | exception | error | |
| only a single or a list of entries is supported but got type | exception | error | |
| The size dictionary must have keys 'height' and 'width'. Got | exception | error | |
| Cannot specify both size as an int, with default_to_square=T | exception | error | image-processing, config, resize, valueerror |
| Cannot specify both default_to_square=True and max_size | exception | error | image-processing, config, resize, valueerror |
| Could not convert size input to size dict: {size} | exception | error | image-processing, type-coercion, config, valueerror |
| {param_name} must have one of the following set of keys: {VA | validation | error | image-processing, config, validation, valueerror |
| Input image must be of type np.ndarray, got {type(image)} | validation | error | image-processing, numpy, typeerror, channel-dimension |
| Unsupported channel dimension format: {channel_dim} | validation | error | image-processing, channel-dimension, valueerror, validation |
| The image to be converted to a PIL image contains values out | validation | error | image-processing, pil, range-validation, valueerror |
| The image to be converted to a PIL image contains values out | validation | error | image-processing, pil, range-validation, normalization, valueerror |
| Input image type not supported: {type(image)} | validation | error | image-processing, pil, typeerror, input-validation |
| size must have 1 or 2 elements if it is a list or tuple | validation | error | image-processing, resize, validation, valueerror |
| max_size = {max_size} must be strictly greater than the requ | validation | error | image-processing, resize, config, valueerror |
| size must have 2 elements | validation | error | image-processing, resize, validation, valueerror |
| image must be a numpy array | validation | error | image-processing, numpy, normalization, typeerror |
| mean must have {num_channels} elements if it is an iterable, | validation | error | image-processing, normalization, validation, valueerror |
| std must have {num_channels} elements if it is an iterable, | validation | error | image-processing, normalization, validation, valueerror |
| size must have 2 elements representing the height and width | validation | error | image-processing, crop, validation, valueerror |
| Unsupported input type {type(bboxes_center)} | validation | error | object-detection, bounding-boxes, typeerror, valueerror |
| Unsupported input type {type(bboxes_corners)} | validation | error | object-detection, bounding-boxes, typeerror, valueerror |
| Unsupported format: {values} | validation | error | image-processing, padding, argument-validation, numpy |
| Invalid padding mode: {mode} | validation | error | image-processing, padding, enum, argument-validation |
| Unsupported channel dimension: {input_data_format} | validation | error | image-processing, channel-dimension, enum, argument-validation |
| Unrecognized image type {type(image)} | validation | error | image-processing, type-checking, argument-validation |
| Invalid image shape. Expected either {expected_ndims + 1} or | validation | error | image-processing, shape-validation, batching |
| Invalid image type. Expected either PIL.Image.Image, numpy.n | validation | error | image-processing, type-checking, argument-validation |
| Could not make a flat list of images from {images} | validation | error | image-processing, shape-validation, nesting, batching |
| Invalid input type. Must be a single image, a list of images | validation | error | image-processing, shape-validation, nesting, batching |
| Invalid image type: {type(img)} | validation | error | image-processing, type-checking, numpy, argument-validation |
| Unsupported number of image dimensions: {image.ndim} | validation | error | image-processing, shape-validation, channel-dimension |
| Unable to infer channel dimension format | validation | error | image-processing, channel-dimension, shape-inference |
| Unsupported data format: {input_data_format} | validation | error | image-processing, channel-dimension, enum, argument-validation |
| Unsupported data format: {channel_dim} | validation | error | image-processing, channel-dimension, enum, argument-validation |
| Invalid channel dimension format: {input_data_format} | validation | error | image-processing, channel-dimension, enum, batching |
| Incorrect image source. Must be a valid URL starting with `h | validation | error | image-processing, io, base64, file-path, loading |
| Incorrect format used for image. Should be an url linking to | validation | error | image-processing, type-checking, loading |
| Incorrect format used for image. Should be a URL, a local pa | validation | error | image-processing, type-checking, torch, loading |
| `rescale_factor` must be specified if `do_rescale` is `True` | validation | error | image-processing, preprocess, argument-validation, configuration |
| Depending on the model, `size_divisor` or `pad_size` or `siz | validation | error | image-processing, preprocess, padding, argument-validation, configuration |
| `image_mean` and `image_std` must both be specified if `do_n | validation | error | |
| `crop_size` must be specified if `do_center_crop` is `True`. | validation | error | |
| `size` and `resample` must be specified if `do_resize` is `T | validation | error | |
| Got type {type(image)} which is not supported, only `PIL.Ima | validation | error | |
| max_size = {max_size} must be strictly greater than the requ | validation | error | |
| Unsupported annotation format: {format} must be one of {supp | validation | error | |
| Invalid COCO detection annotations. Annotations must a dict | validation | error | |
| Invalid COCO panoptic annotations. Annotations must a dict ( | validation | error | |
| Key {key} not found in SizeDict. | validation | error | |
| Key {key} is not a valid field of SizeDict. | validation | error | |
| invalid distribution {distribution} | validation | error | |
| You are using `from_pretrained` with a meta device context m | exception | error | |
| When passing device_map as a string, the value needs to be a | validation | error | |
| You can't pass device_map as a negative int. If you want to | validation | error | |
| DeepSpeed Zero-3 is not compatible with passing a `device_ma | validation | error | |
| Using a `device_map`, `tp_plan`, `torch.device` context mana | validation | error | |
| The current `device_map` had weights offloaded to the disk, | validation | error | |
| {param_name} is on the meta device because it was offloaded, | validation | error | |
| The `device_map` does not contain the module {param}. | validation | error | |
| Input weight should be of type nn.Parameter, got {type(weigh | validation | error | |
| None of the available devices `available_devices = {availabl | exception | error | |
| Failed to load `kernels-community/deep-gemm` — check that a | exception | error | |
| DeepGEMM kernel is missing required symbols: {', '.join(miss | exception | error | |
| DeepGEMM's FP4 (int8-packed) path requires a Blackwell (SM10 | exception | error | |
| DeepGEMM has no float32 scale-factor path on Blackwell (SM10 | exception | error | |
| DeepGEMM SF must be 2D or 3D, got {sf.dim()}D | validation | error | |
| DeepGEMM requires block-wise quantized FP8 weights, but the | validation | error | |
| DeepGEMM requires `block_size` ∈ {(128, 128), (1, 128)}, got | validation | error | |
| DeepGEMM linear does not support static activation quantizat | exception | error | |
| DeepGEMM linear requires FP16 or BF16 activations, got {inpu | validation | error | |
| DeepGEMM experts path requires bfloat16 hidden states, got { | validation | error | |
| DeepGEMM experts selected on a model spanning multiple CUDA | exception | error | |
| DeepGEMM experts dispatch does not support static activation | exception | error | |
| DeepGEMM Mega MoE requires `hidden_dim` and `intermediate_hi | validation | error | |
| DeepGEMM Mega MoE requires FP4-packed expert weights (dtype= | exception | error | |
| DeepGEMM Mega MoE requires a `process_group` for the EP grou | validation | error | |
| trainer_config_process() wasn't called yet to tell dtype | validation | error | |
| The model's config file has neither `hidden_size` nor `hidde | validation | error | |
| Please correct the following DeepSpeed config values that mi | validation | error | |
| Weight conversions (e.g., MoE expert fusion) with DeepSpeed | exception | error | |
| Failed to apply weight conversion for '{renamed_key}'. This | exception | error | |
| ZeRO inference only makes sense with ZeRO Stage 3 - please a | exception | error | |
| [deepspeed] failed to resume from checkpoint {checkpoint_pat | exception | error | |
| Can't find a valid checkpoint at {checkpoint_path} | exception | error | |
| Sequence parallelism is enabled but no SP process group is a | exception | error | |
| The model must have caching enabled to be performant. | exception | error | |
| Need to specify either input_ids or inputs_embeds. | exception | error | |
| The model must have a generation config to be exported with | exception | error | |
| The model must have caching enabled to be exported with stat | exception | error | |
| The model must use a 'static' caching implementation to be e | exception | error | |
| batch_size must be provided, either as an argument or in cac | exception | error | |
| max_cache_len must be provided, either as an argument or in | exception | error | |
| Model must have caching enabled. | exception | error | |
| This pytree flattening function should only be applied to Dy | exception | error | |
| scale_fmt='ue8m0' requires torch.float8_e8m0fnu, which is on | exception | error | |
| {type(obj).__name__} has none of: {names} | exception | error | |
| finegrained-fp8 kernel unavailable: {_MISSING_KERNELS_MESSAG | exception | error | fp8, quantization, kernels, import-error, optional-dependency |
| Failed to load the finegrained-fp8 kernel — check that `kern | exception | error | fp8, quantization, kernels, torch-cuda-mismatch, import-error |
| finegrained-fp8 kernel is missing required symbols: {', '.jo | exception | error | fp8, quantization, kernels, version-mismatch, import-error, cache |
| batched_mm experts dispatch does not support activation_sche | exception | error | fp8, moe, quantization, config-mismatch, not-implemented |
| grouped_mm experts dispatch does not support activation_sche | exception | error | fp8, moe, quantization, config-mismatch, not-implemented |
| Weight shape ({rows}, {cols}) not divisible by scale grid ({ | exception | error | fp8, mxfp4, dequantization, checkpoint-mismatch, shape-validation |
| Fp8Dequantize: weight/scale count mismatch for {key} ({len(w | exception | error | fp8, moe, checkpoint-corruption, state-dict, dequantization |
| Tensor query has shape with a zero dimension. FlashAttentio | exception | error | flash-attention, attention, empty-tensor, shape-validation |
| `flex_attention` does not support `dropout`. Please use it w | exception | error | flex-attention, attention, dropout, config, training-vs-inference |
| Attention sinks cannot be run on CPU with flex attention. Pl | exception | error | flex-attention, attention-sinks, cpu, device, lse |
| Unsupported forward dtype: {config.forward_dtype} | exception | error | fp-quant, quantization, config, dtype, validation |
| Unsupported backward dtype: {config.backward_dtype} | exception | error | fp-quant, quantization, config, dtype, validation |
| Received multiple types, therefore expected the first type t | exception | error | gguf, ggml, metadata-parsing, file-corruption |
| tokens and scores need to be passed for a LLaMa tokenizer wi | exception | error | gguf, tokenizer, metadata-parsing, llama-spm |
| `skip` must be an iterable of strings. | exception | error | heterogeneity, config, type-validation, per-layer |
| `skip` must contain only strings. | exception | error | heterogeneity, config, type-validation, per-layer |
| `per_layer_config` keys must be integer layer indices in the | exception | error | heterogeneity, config, index-validation, per-layer |
| The following layers have the mutually exclusive `sliding_wi | exception | error | heterogeneity, config, sliding-window, attention, mutually-exclusive |
| The following attributes are missing: {sorted(missing_requir | exception | error | heterogeneity, config, missing-attribute, per-layer, validation |
| Layer type '{layer_idx}' requested, but config.layer_types i | exception | error | heterogeneity, config, layer-types, indexing |
| Layer type '{layer_idx}' not found in config.layer_types: {l | exception | error | config, heterogeneity, layer-types, validation |
| Layer type '{layer_idx}' is not homogeneous across layers (l | validation | error | config, heterogeneity, per-layer-overrides, validation |
| list index out of range | validation | error | config, index-error, heterogeneity, off-by-one |
| '{key}' is a per-layer attribute and may vary across layers. | validation | error | config, heterogeneity, attribute-access, ambiguity |
| Unsupported p={p}, n={n} | exception | error | quantization, higgs, not-implemented, unsupported-arguments |
| Workspace must be set before calling forward | exception | error | quantization, higgs, vllm-kernel, missing-setup, runtime-state |
| LayerRepository requires `kernels` to be installed. Run `pip | exception | error | dependencies, kernels, optional-install, runtime-error |
| LocalLayerRepository requires `kernels` to be installed. Run | exception | error | dependencies, kernels, local-kernels, offline |
| FuncRepository requires `kernels` to be installed. Run `pip | exception | error | dependencies, kernels, optional-install |
| replace_kernel_forward_from_hub requires `kernels` to be ins | exception | error | dependencies, kernels, decorator, import-time |
| register_kernel_mapping requires `kernels` to be installed. | exception | error | dependencies, kernels, kernel-mapping |
| register_kernel_mapping_transformers requires `kernels` to b | exception | error | dependencies, kernels, kernel-mapping, startup |
| `kernels` is either not installed or uses an incompatible ve | exception | error | dependencies, kernels, attention, version-mismatch |
| An error occurred while trying to load from '{repo_id}': {e} | validation | error | huggingface-hub, kernels, network, repo-not-found, wrapped-error |
| Model {cls.__name__} has no config_class or model_type. | validation | error | kernels, model-registration, config-class, custom-model |
| Expected exactly one kernel repo regardless of device/mode s | validation | error | kernels, kernel-mapping, config-validation, kernelize |
| Invalid hub repo {hub_repo!r} for layer {layer_name!r} | validation | error | kernel-config, validation, hub-kernels |
| Invalid kernel repo string {repo_str!r} for layer {layer_nam | validation | error | kernel-config, hub-kernels, validation |
| Could not load kernel class from hub_repo={hub_repo!r} | validation | error | hub-kernels, loading, network |
| Fused kernel {kernel_cls.__name__!r} requires a companion la | validation | error | hub-kernels, fusion, kernel-authoring |
| All patterns for a fused kernel must share the same parent m | validation | error | hub-kernels, fusion, kernel-config |
| Module {name!r} does not have the expected child modules {ch | validation | error | hub-kernels, fusion, pattern-matching |
| No module matched pattern {parent_pattern!r} for fused kerne | validation | error | hub-kernels, fusion, pattern-matching, model-structure |
| Unknown type for trial {trial.__class__} | exception | error | hpo, optuna, dependencies, trainer |
| only support DDP optuna HPO for ParallelMode.DISTRIBUTED cur | exception | error | hpo, optuna, distributed, trainer |