GGUF metadata keys
Key templates only; gguf-py does not record value types here. {arch} = general.architecture value; {id} = 0-based index; other {placeholders} are arch-specific.
ggml-org/llama.cpp gguf-py/gguf/constants.py class Keys @ 6050b467c07a
llama.cpp gguf-py/gguf/constants.py (Keys, GGMLQuantizationType, LlamaFileType, GGML_QUANT_SIZES)
| group | const | key |
|---|---|---|
| General | TYPE | general.type |
| General | ARCHITECTURE | general.architecture |
| General | QUANTIZATION_VERSION | general.quantization_version |
| General | ALIGNMENT | general.alignment |
| General | FILE_TYPE | general.file_type |
| General | TENSOR_EXTRA_NAME | general.tensor_extra.name |
| General | TENSOR_EXTRA_PREC_A4 | general.tensor_extra.prec_a4 |
| General | SAMPLING_SEQUENCE | general.sampling.sequence |
| General | SAMPLING_TOP_K | general.sampling.top_k |
| General | SAMPLING_TOP_P | general.sampling.top_p |
| General | SAMPLING_MIN_P | general.sampling.min_p |
| General | SAMPLING_XTC_PROBABILITY | general.sampling.xtc_probability |
| General | SAMPLING_XTC_THRESHOLD | general.sampling.xtc_threshold |
| General | SAMPLING_TEMP | general.sampling.temp |
| General | SAMPLING_PENALTY_LAST_N | general.sampling.penalty_last_n |
| General | SAMPLING_PENALTY_REPEAT | general.sampling.penalty_repeat |
| General | SAMPLING_MIROSTAT | general.sampling.mirostat |
| General | SAMPLING_MIROSTAT_TAU | general.sampling.mirostat_tau |
| General | SAMPLING_MIROSTAT_ETA | general.sampling.mirostat_eta |
| General | NAME | general.name |
| General | AUTHOR | general.author |
| General | VERSION | general.version |
| General | ORGANIZATION | general.organization |
| General | FINETUNE | general.finetune |
| General | BASENAME | general.basename |
| General | DESCRIPTION | general.description |
| General | QUANTIZED_BY | general.quantized_by |
| General | SIZE_LABEL | general.size_label |
| General | LICENSE | general.license |
| General | LICENSE_NAME | general.license.name |
| General | LICENSE_LINK | general.license.link |
| General | URL | general.url |
| General | DOI | general.doi |
| General | UUID | general.uuid |
| General | REPO_URL | general.repo_url |
| General | SOURCE_URL | general.source.url |
| General | SOURCE_DOI | general.source.doi |
| General | SOURCE_UUID | general.source.uuid |
| General | SOURCE_REPO_URL | general.source.repo_url |
| General | BASE_MODEL_COUNT | general.base_model.count |
| General | BASE_MODEL_NAME | general.base_model.{id}.name |
| General | BASE_MODEL_AUTHOR | general.base_model.{id}.author |
| General | BASE_MODEL_VERSION | general.base_model.{id}.version |
| General | BASE_MODEL_ORGANIZATION | general.base_model.{id}.organization |
| General | BASE_MODEL_DESCRIPTION | general.base_model.{id}.description |
| General | BASE_MODEL_URL | general.base_model.{id}.url |
| General | BASE_MODEL_DOI | general.base_model.{id}.doi |
| General | BASE_MODEL_UUID | general.base_model.{id}.uuid |
| General | BASE_MODEL_REPO_URL | general.base_model.{id}.repo_url |
| General | DATASET_COUNT | general.dataset.count |
| General | DATASET_NAME | general.dataset.{id}.name |
| General | DATASET_AUTHOR | general.dataset.{id}.author |
| General | DATASET_VERSION | general.dataset.{id}.version |
| General | DATASET_ORGANIZATION | general.dataset.{id}.organization |
| General | DATASET_DESCRIPTION | general.dataset.{id}.description |
| General | DATASET_URL | general.dataset.{id}.url |
| General | DATASET_DOI | general.dataset.{id}.doi |
| General | DATASET_UUID | general.dataset.{id}.uuid |
| General | DATASET_REPO_URL | general.dataset.{id}.repo_url |
| General | TAGS | general.tags |
| General | LANGUAGES | general.languages |
| LLM | VOCAB_SIZE | {arch}.vocab_size |
| LLM | CONTEXT_LENGTH | {arch}.context_length |
| LLM | EMBEDDING_LENGTH | {arch}.embedding_length |
| LLM | EMBEDDING_LENGTH_OUT | {arch}.embedding_length_out |
| LLM | FEATURES_LENGTH | {arch}.features_length |
| LLM | BLOCK_COUNT | {arch}.block_count |
| LLM | LEADING_DENSE_BLOCK_COUNT | {arch}.leading_dense_block_count |
| LLM | FEED_FORWARD_LENGTH | {arch}.feed_forward_length |
| LLM | EXPERT_FEED_FORWARD_LENGTH | {arch}.expert_feed_forward_length |
| LLM | EXPERT_SHARED_FEED_FORWARD_LENGTH | {arch}.expert_shared_feed_forward_length |
| LLM | EXPERT_CHUNK_FEED_FORWARD_LENGTH | {arch}.expert_chunk_feed_forward_length |
| LLM | USE_PARALLEL_RESIDUAL | {arch}.use_parallel_residual |
| LLM | TENSOR_DATA_LAYOUT | {arch}.tensor_data_layout |
| LLM | EXPERT_COUNT | {arch}.expert_count |
| LLM | EXPERT_USED_COUNT | {arch}.expert_used_count |
| LLM | EXPERT_SHARED_COUNT | {arch}.expert_shared_count |
| LLM | EXPERT_GROUP_COUNT | {arch}.expert_group_count |
| LLM | EXPERT_GROUP_USED_COUNT | {arch}.expert_group_used_count |
| LLM | EXPERT_WEIGHTS_SCALE | {arch}.expert_weights_scale |
| LLM | EXPERT_WEIGHTS_NORM | {arch}.expert_weights_norm |
| LLM | EXPERT_GATING_FUNC | {arch}.expert_gating_func |
| LLM | EXPERT_GROUP_SCALE | {arch}.expert_group_scale |
| LLM | EXPERT_LATENT_LENGTH | {arch}.expert_latent_length |
| LLM | EXPERTS_PER_GROUP | {arch}.experts_per_group |
| LLM | MOE_EVERY_N_LAYERS | {arch}.moe_every_n_layers |
| LLM | MOE_LATENT_SIZE | {arch}.moe_latent_size |
| LLM | POST_NORM_EPS | {arch}.post_norm_epsilon |
| LLM | NEXTN_PREDICT_LAYERS | {arch}.nextn_predict_layers |
| LLM | NUM_DEEPSTACK_LAYERS | {arch}.n_deepstack_layers |
| LLM | DEEPSTACK_MAPPING | {arch}.deepstack_mapping |
| LLM | POOLING_TYPE | {arch}.pooling_type |
| LLM | LOGIT_SCALE | {arch}.logit_scale |
| LLM | DECODER_START_TOKEN_ID | {arch}.decoder_start_token_id |
| LLM | DECODER_BLOCK_COUNT | {arch}.decoder_block_count |
| LLM | ATTN_LOGIT_SOFTCAPPING | {arch}.attn_logit_softcapping |
| LLM | ROUTER_LOGIT_SOFTCAPPING | {arch}.router_logit_softcapping |
| LLM | FINAL_LOGIT_SOFTCAPPING | {arch}.final_logit_softcapping |
| LLM | SWIN_NORM | {arch}.swin_norm |
| LLM | RESCALE_EVERY_N_LAYERS | {arch}.rescale_every_n_layers |
| LLM | TIME_MIX_EXTRA_DIM | {arch}.time_mix_extra_dim |
| LLM | TIME_DECAY_EXTRA_DIM | {arch}.time_decay_extra_dim |
| LLM | RESIDUAL_SCALE | {arch}.residual_scale |
| LLM | EMBEDDING_SCALE | {arch}.embedding_scale |
| LLM | TOKEN_SHIFT_COUNT | {arch}.token_shift_count |
| LLM | INTERLEAVE_MOE_LAYER_STEP | {arch}.interleave_moe_layer_step |
| LLM | FULL_ATTENTION_INTERVAL | {arch}.full_attention_interval |
| LLM | NUM_LOOPS | {arch}.num_loops |
| LLM | SKIP_LOOP_FINAL_NORM | {arch}.skip_loop_final_norm |
| LLM | HASH_LAYER_COUNT | {arch}.hash_layer_count |
| LLM | ACTIVATION_SPARSITY_SCALE | {arch}.activation_sparsity_scale |
| LLM | ALTUP_ACTIVE_IDX | {arch}.altup.active_idx |
| LLM | ALTUP_NUM_INPUTS | {arch}.altup.num_inputs |
| LLM | EMBD_LENGTH_PER_LAYER_INP | {arch}.embedding_length_per_layer_input |
| LLM | SWIGLU_CLAMP_EXP | {arch}.swiglu_clamp_exp |
| LLM | SWIGLU_CLAMP_SHEXP | {arch}.swiglu_clamp_shexp |
| LLM | HIDDEN_ACT | {arch}.hidden_activation |
| LLM | DENSE_FEAT_IN_SIZE | {arch}.{dense}_feat_in |
| LLM | DENSE_FEAT_OUT_SIZE | {arch}.{dense}_feat_out |
| LLM | TARGET_LAYERS | {arch}.target_layers |
| LLM | TARGET_HIDDEN_SIZE | {arch}.target_hidden_size |
| LLM | BLOCK_SIZE | {arch}.block_size |
| LLM | CONV_KERNEL_SIZE | {arch}.conv_kernel_size |
| LLM | CONV_GROUP_SIZE | {arch}.conv_group_size |
| LLM | SELECTOR_RANK | {arch}.selector_rank |
| LLM | SELECTOR_TOP_K | {arch}.selector_top_k |
| LLM | SAMPLE_FROM_ANCHOR | {arch}.sample_from_anchor |
| LLM | HAS_CONFIDENCE_HEAD | {arch}.has_confidence_head |
| LLM | NORM_BEFORE_RESIDUAL | {arch}.norm_before_residual |
| LLM | NORM_BEFORE_FC | {arch}.norm_before_fc |
| LLM | REALTIME_DELAY_TOKENS | {arch}.realtime_delay_tokens |
| Adapters | COUNT | {arch}.adapters.count |
| Adapters | TOKEN_IDS_ACTIVATE | {arch}.adapters.token_ids_activate |
| Adapters | TOKEN_IDS_SUBSTITUTE | {arch}.adapters.token_ids_substitute |
| Adapters | LORA_RANK | {arch}.adapters.lora_rank |
| Adapters | ROUTER_GAIN | {arch}.adapters.router_gain |
| Attention | HEAD_COUNT | {arch}.attention.head_count |
| Attention | HEAD_COUNT_KV | {arch}.attention.head_count_kv |
| Attention | MAX_ALIBI_BIAS | {arch}.attention.max_alibi_bias |
| Attention | CLAMP_KQV | {arch}.attention.clamp_kqv |
| Attention | KEY_LENGTH | {arch}.attention.key_length |
| Attention | VALUE_LENGTH | {arch}.attention.value_length |
| Attention | LAYERNORM_EPS | {arch}.attention.layer_norm_epsilon |
| Attention | LAYERNORM_RMS_EPS | {arch}.attention.layer_norm_rms_epsilon |
| Attention | GROUPNORM_EPS | {arch}.attention.group_norm_epsilon |
| Attention | GROUPNORM_GROUPS | {arch}.attention.group_norm_groups |
| Attention | CAUSAL | {arch}.attention.causal |
| Attention | Q_LORA_RANK | {arch}.attention.q_lora_rank |
| Attention | KV_LORA_RANK | {arch}.attention.kv_lora_rank |
| Attention | DECAY_LORA_RANK | {arch}.attention.decay_lora_rank |
| Attention | ICLR_LORA_RANK | {arch}.attention.iclr_lora_rank |
| Attention | VALUE_RESIDUAL_MIX_LORA_RANK | {arch}.attention.value_residual_mix_lora_rank |
| Attention | GATE_LORA_RANK | {arch}.attention.gate_lora_rank |
| Attention | REL_BUCKETS_COUNT | {arch}.attention.relative_buckets_count |
| Attention | SLIDING_WINDOW | {arch}.attention.sliding_window |
| Attention | SCALE | {arch}.attention.scale |
| Attention | OUTPUT_GROUP_COUNT | {arch}.attention.output_group_count |
| Attention | OUTPUT_LORA_RANK | {arch}.attention.output_lora_rank |
| Attention | OUTPUT_SCALE | {arch}.attention.output_scale |
| Attention | VALUE_SCALE | {arch}.attention.value_scale |
| Attention | COMPRESS_RATIOS | {arch}.attention.compress_ratios |
| Attention | COMPRESS_ROPE_FREQ_BASE | {arch}.attention.compress_rope_freq_base |
| Attention | KV_SOURCE_LAYERS | {arch}.attention.kv_source_layers |
| Attention | TEMPERATURE_LENGTH | {arch}.attention.temperature_length |
| Attention | KEY_LENGTH_MLA | {arch}.attention.key_length_mla |
| Attention | VALUE_LENGTH_MLA | {arch}.attention.value_length_mla |
| Attention | KEY_LENGTH_SWA | {arch}.attention.key_length_swa |
| Attention | VALUE_LENGTH_SWA | {arch}.attention.value_length_swa |
| Attention | KEY_LENGTH_MLA_SWA | {arch}.attention.key_length_mla_swa |
| Attention | VALUE_LENGTH_MLA_SWA | {arch}.attention.value_length_mla_swa |
| Attention | KV_LORA_RANK_SWA | {arch}.attention.kv_lora_rank_swa |
| Attention | SHARED_KV_LAYERS | {arch}.attention.shared_kv_layers |
| Attention | SLIDING_WINDOW_PATTERN | {arch}.attention.sliding_window_pattern |
| Attention | RECURRENT_LAYERS | {arch}.attention.recurrent_layers |
| Attention | TEMPERATURE_SCALE | {arch}.attention.temperature_scale |
| Attention | ROPE_PATTERN | {arch}.attention.rope_pattern |
| Attention | VALUE_EXPERT_COUNT | {arch}.attention.value_expert_count |
| Attention | VALUE_EXPERT_USED_COUNT | {arch}.attention.value_expert_used_count |
| Attention.Indexer | HEAD_COUNT | {arch}.attention.indexer.head_count |
| Attention.Indexer | KEY_LENGTH | {arch}.attention.indexer.key_length |
| Attention.Indexer | TOP_K | {arch}.attention.indexer.top_k |
| Attention.Indexer | BLOCK_SIZE | {arch}.attention.indexer.block_size |
| Attention.Indexer | LOCAL_BLOCKS | {arch}.attention.indexer.local_blocks |
| Attention.Indexer | TYPES | {arch}.attention.indexer.types |
| Attention.Indexer | KPOOL | {arch}.attention.indexer.kpool |
| Attention.Indexer | KPOOL_SELECT_TAIL | {arch}.attention.indexer.kpool_select_tail |
| Attention.Indexer | SOURCE_LAYERS | {arch}.attention.indexer.source_layers |
| Attention.Indexer | CANDIDATE_SOURCE_LAYER | {arch}.attention.indexer.candidate_source_layer |
| Attention.Indexer | CANDIDATE_TOP_K_BLOCKS | {arch}.attention.indexer.candidate_top_k_blocks |
| Attention.Indexer | CANDIDATE_BLOCK_SIZE | {arch}.attention.indexer.candidate_block_size |
| HyperConnection | COUNT | {arch}.hyper_connection.count |
| HyperConnection | SINKHORN_ITERATIONS | {arch}.hyper_connection.sinkhorn_iterations |
| HyperConnection | EPSILON | {arch}.hyper_connection.epsilon |
| HyperConnection | MAGNITUDE | {arch}.hyper_connection.magnitude |
| HyperConnection | LOW_RANK | {arch}.hyper_connection.low_rank |
| HyperConnection | SINGLE_PASS | {arch}.hyper_connection.single_pass |
| Engram | PAD_TOKEN_ID | {arch}.engram.pad_token_id |
| Engram | TOKEN_MAP | {arch}.engram.token_map |
| PerLayerEmbedding | LAYERS | {arch}.ple.layers |
| PerLayerEmbedding | NGRAM_SIZE | {arch}.ple.ngram_size |
| PerLayerEmbedding | HEADS_PER_NGRAM | {arch}.ple.heads_per_ngram |
| PerLayerEmbedding | CONV_KERNEL | {arch}.ple.conv_kernel |
| PerLayerEmbedding | LAYER_MULTIPLIERS | {arch}.ple.layer_multipliers |
| PerLayerEmbedding | HEAD_OFFSETS | {arch}.ple.head_offsets |
| PerLayerEmbedding | HEAD_VOCAB_SIZES | {arch}.ple.head_vocab_sizes |
| PerLayerEmbedding | EOS_TOKEN_ID | {arch}.ple.eos_token_id |
| PerLayerEmbedding | IMAGE_TOKEN_ID | {arch}.ple.image_token_id |
| Rope | DIMENSION_COUNT | {arch}.rope.dimension_count |
| Rope | DIMENSION_COUNT_SWA | {arch}.rope.dimension_count_swa |
| Rope | DIMENSION_SECTIONS | {arch}.rope.dimension_sections |
| Rope | SECTION_ORDER | {arch}.rope.section_order |
| Rope | FREQ_BASE | {arch}.rope.freq_base |
| Rope | FREQ_BASE_SPATIAL | {arch}.rope.freq_base_spatial |
| Rope | FREQ_BASE_SWA | {arch}.rope.freq_base_swa |
| Rope | SCALING_TYPE | {arch}.rope.scaling.type |
| Rope | SCALING_FACTOR | {arch}.rope.scaling.factor |
| Rope | SCALING_ALPHA | {arch}.rope.scaling.alpha |
| Rope | SCALING_ATTN_FACTOR | {arch}.rope.scaling.attn_factor |
| Rope | SCALING_ORIG_CTX_LEN | {arch}.rope.scaling.original_context_length |
| Rope | SCALING_FINETUNED | {arch}.rope.scaling.finetuned |
| Rope | SCALING_YARN_TRUNCATE | {arch}.rope.scaling.yarn_truncate |
| Rope | SCALING_YARN_LOG_MUL | {arch}.rope.scaling.yarn_log_multiplier |
| Rope | SCALING_YARN_EXT_FACTOR | {arch}.rope.scaling.yarn_ext_factor |
| Rope | SCALING_YARN_ATTN_FACTOR | {arch}.rope.scaling.yarn_attn_factor |
| Rope | SCALING_YARN_BETA_FAST | {arch}.rope.scaling.yarn_beta_fast |
| Rope | SCALING_YARN_BETA_SLOW | {arch}.rope.scaling.yarn_beta_slow |
| Activation | SITU_BETA | {arch}.activation.situ_beta |
| Activation | SITU_LINEAR_BETA | {arch}.activation.situ_linear_beta |
| AttnRes | BLOCK_SIZE | {arch}.attn_res.block_size |
| Split | LLM_KV_SPLIT_NO | split.no |
| Split | LLM_KV_SPLIT_COUNT | split.count |
| Split | LLM_KV_SPLIT_TENSORS_COUNT | split.tensors.count |
| HRM | LAYERS_PER_STACK | {arch}.hrm.layers_per_stack |
| HRM | H_CYCLES | {arch}.hrm.h_cycles |
| HRM | L_CYCLES | {arch}.hrm.l_cycles |
| HRM | PREFIX_LM | {arch}.hrm.prefix_lm |
| SSM | CONV_KERNEL | {arch}.ssm.conv_kernel |
| SSM | INNER_SIZE | {arch}.ssm.inner_size |
| SSM | STATE_SIZE | {arch}.ssm.state_size |
| SSM | TIME_STEP_RANK | {arch}.ssm.time_step_rank |
| SSM | GROUP_COUNT | {arch}.ssm.group_count |
| SSM | DT_B_C_RMS | {arch}.ssm.dt_b_c_rms |
| KDA | HEAD_DIM | {arch}.kda.head_dim |
| KDA | SAFE_GATE | {arch}.kda.safe_gate |
| KDA | GATE_LOWER_BOUND | {arch}.kda.gate_lower_bound |
| WKV | HEAD_SIZE | {arch}.wkv.head_size |
| LiliCorr | HIDDEN_SIZE | lilicorr.hidden_size |
| LiliCorr | CANDIDATE_TOPK | lilicorr.candidate_topk |
| LiliCorr | NUM_HEADS | lilicorr.num_heads |
| LiliCorr | NUM_LAYERS | lilicorr.num_layers |
| LiliCorr | FACTOR_DIM | lilicorr.factor_dim |
| LiliCorr | MLP_HIDDEN_SIZE | lilicorr.mlp_hidden_size |
| LiliCorr | VECTOR_EPS | lilicorr.vector_eps |
| LiliCorr | LOGIT_SCALE | lilicorr.logit_scale |
| PosNet | EMBEDDING_LENGTH | {arch}.posnet.embedding_length |
| PosNet | BLOCK_COUNT | {arch}.posnet.block_count |
| ConvNext | EMBEDDING_LENGTH | {arch}.convnext.embedding_length |
| ConvNext | BLOCK_COUNT | {arch}.convnext.block_count |
| Classifier | OUTPUT_LABELS | {arch}.classifier.output_labels |
| Classifier | POOLING_TYPE | {arch}.classifier.pooling_type |
| Classifier | ACTIVATION | {arch}.classifier.activation |
| ShortConv | L_CACHE | {arch}.shortconv.l_cache |
| Decision | TYPE | {arch}.decision.type |
| Decision | BLOCK_COUNT | {arch}.decision.block_count |
| Decision | ROUTING_BLOCK_COUNT | {arch}.decision.routing_block_count |
| Decision | HEAD_COUNT | {arch}.decision.head_count |
| Decision | MAX_HEAD_TOKENS | {arch}.decision.max_head_tokens |
| Decision | LABELS | {arch}.decision.labels |
| Decision | TEMPERATURE | {arch}.decision.temperature.{name} |
| Decision | SCORE_BIAS | {arch}.decision.score_bias.{n}.{i} |
| Tokenizer | MODEL | tokenizer.ggml.model |
| Tokenizer | PRE | tokenizer.ggml.pre |
| Tokenizer | LIST | tokenizer.ggml.tokens |
| Tokenizer | TOKEN_TYPE | tokenizer.ggml.token_type |
| Tokenizer | TOKEN_TYPE_COUNT | tokenizer.ggml.token_type_count |
| Tokenizer | SCORES | tokenizer.ggml.scores |
| Tokenizer | MERGES | tokenizer.ggml.merges |
| Tokenizer | BOS_ID | tokenizer.ggml.bos_token_id |
| Tokenizer | EOS_ID | tokenizer.ggml.eos_token_id |
| Tokenizer | EOT_ID | tokenizer.ggml.eot_token_id |
| Tokenizer | EOM_ID | tokenizer.ggml.eom_token_id |
| Tokenizer | UNK_ID | tokenizer.ggml.unknown_token_id |
| Tokenizer | SEP_ID | tokenizer.ggml.seperator_token_id |
| Tokenizer | PAD_ID | tokenizer.ggml.padding_token_id |
| Tokenizer | MASK_ID | tokenizer.ggml.mask_token_id |
| Tokenizer | ADD_BOS | tokenizer.ggml.add_bos_token |
| Tokenizer | ADD_EOS | tokenizer.ggml.add_eos_token |
| Tokenizer | ADD_SEP | tokenizer.ggml.add_sep_token |
| Tokenizer | DNA_AUTO_TAGS | tokenizer.ggml.dna_auto_tags |
| Tokenizer | ADD_PREFIX | tokenizer.ggml.add_space_prefix |
| Tokenizer | REMOVE_EXTRA_WS | tokenizer.ggml.remove_extra_whitespaces |
| Tokenizer | CLEAN_SPACES | tokenizer.ggml.clean_spaces |
| Tokenizer | PRECOMPILED_CHARSMAP | tokenizer.ggml.precompiled_charsmap |
| Tokenizer | SUPPRESS_TOKENS | tokenizer.ggml.suppress_tokens |
| Tokenizer | HF_JSON | tokenizer.huggingface.json |
| Tokenizer | RWKV | tokenizer.rwkv.world |
| Tokenizer | CHAT_TEMPLATE | tokenizer.chat_template |
| Tokenizer | CHAT_TEMPLATE_N | tokenizer.chat_template.{name} |
| Tokenizer | CHAT_TEMPLATES | tokenizer.chat_templates |
| Tokenizer | NORMALIZER_NFC | tokenizer.ggml.normalizer.nfc |
| Tokenizer | NORMALIZER_LOWERCASE | tokenizer.ggml.normalizer.lowercase |
| Tokenizer | NORMALIZER_STRIP_ACCENTS | tokenizer.ggml.normalizer.strip_accents |
| Tokenizer | METASPACE_PREPEND_SCHEME | tokenizer.ggml.metaspace.prepend_scheme |
| Tokenizer | FIM_PRE_ID | tokenizer.ggml.fim_pre_token_id |
| Tokenizer | FIM_SUF_ID | tokenizer.ggml.fim_suf_token_id |
| Tokenizer | FIM_MID_ID | tokenizer.ggml.fim_mid_token_id |
| Tokenizer | FIM_PAD_ID | tokenizer.ggml.fim_pad_token_id |
| Tokenizer | FIM_REP_ID | tokenizer.ggml.fim_rep_token_id |
| Tokenizer | FIM_SEP_ID | tokenizer.ggml.fim_sep_token_id |
| Tokenizer | PREFIX_ID | tokenizer.ggml.prefix_token_id |
| Tokenizer | SUFFIX_ID | tokenizer.ggml.suffix_token_id |
| Tokenizer | MIDDLE_ID | tokenizer.ggml.middle_token_id |
| Adapter | TYPE | adapter.type |
| Adapter | LORA_ALPHA | adapter.lora.alpha |
| Adapter | LORA_TASK_NAME | adapter.lora.task_name |
| Adapter | LORA_PROMPT_PREFIX | adapter.lora.prompt_prefix |
| Adapter | ALORA_INVOCATION_TOKENS | adapter.alora.invocation_tokens |
| IMatrix | CHUNK_COUNT | imatrix.chunk_count |
| IMatrix | CHUNK_SIZE | imatrix.chunk_size |
| IMatrix | DATASETS | imatrix.datasets |
| Clip | PROJECTOR_TYPE | clip.projector_type |
| Clip | HAS_VISION_ENCODER | clip.has_vision_encoder |
| Clip | HAS_AUDIO_ENCODER | clip.has_audio_encoder |
| Clip | HAS_GEN_AUDIO_ENCODER | clip.has_gen_audio_encoder |
| Clip | HAS_LLAVA_PROJECTOR | clip.has_llava_projector |
| ClipVision | PROJECTOR_TYPE | clip.vision.projector_type |
| ClipVision | IMAGE_SIZE | clip.vision.image_size |
| ClipVision | IMAGE_MIN_PIXELS | clip.vision.image_min_pixels |
| ClipVision | IMAGE_MAX_PIXELS | clip.vision.image_max_pixels |
| ClipVision | PREPROC_MIN_TILES | clip.vision.preproc_min_tiles |
| ClipVision | PREPROC_MAX_TILES | clip.vision.preproc_max_tiles |
| ClipVision | PREPROC_IMAGE_SIZE | clip.vision.preproc_image_size |
| ClipVision | PREPROC_TILE_SIZE | clip.vision.preproc_tile_size |
| ClipVision | PREPROC_MAX_PIXELS_TOLERANCE | clip.vision.preproc_max_pixels_tolerance |
| ClipVision | PATCH_SIZE | clip.vision.patch_size |
| ClipVision | EMBEDDING_LENGTH | clip.vision.embedding_length |
| ClipVision | FEED_FORWARD_LENGTH | clip.vision.feed_forward_length |
| ClipVision | PROJECTION_DIM | clip.vision.projection_dim |
| ClipVision | BLOCK_COUNT | clip.vision.block_count |
| ClipVision | IMAGE_MEAN | clip.vision.image_mean |
| ClipVision | IMAGE_STD | clip.vision.image_std |
| ClipVision | MAX_SLICE_NUMS | clip.vision.max_slice_nums |
| ClipVision | IMAGE_RESIZE_ALGO | clip.vision.image_resize_algo |
| ClipVision | SPATIAL_MERGE_SIZE | clip.vision.spatial_merge_size |
| ClipVision | SWIGLU_CLAMP | clip.vision.swiglu_clamp |
| ClipVision | EXPERT_COUNT_PER_LAYER | clip.vision.expert_count_per_layer |
| ClipVision | EXPERT_USED_COUNT | clip.vision.expert_used_count |
| ClipVision | USE_GELU | clip.use_gelu |
| ClipVision | USE_SILU | clip.use_silu |
| ClipVision | N_WA_PATTERN | clip.vision.n_wa_pattern |
| ClipVision | WA_LAYER_INDEXES | clip.vision.wa_layer_indexes |
| ClipVision | WA_PATTERN_MODE | clip.vision.wa_pattern_mode |
| ClipVision | IS_DEEPSTACK_LAYERS | clip.vision.is_deepstack_layers |
| ClipVision | WINDOW_SIZE | clip.vision.window_size |
| ClipVision | FEATURE_LAYERS | clip.vision.feature_layer |
| ClipVision | IMAGE_GRID_PINPOINTS | clip.vision.image_grid_pinpoints |
| ClipVision.Attention | HEAD_COUNT | clip.vision.attention.head_count |
| ClipVision.Attention | HEAD_COUNT_KV | clip.vision.attention.head_count_kv |
| ClipVision.Attention | HEAD_DIM | clip.vision.attention.head_dim |
| ClipVision.Attention | LAYERNORM_EPS | clip.vision.attention.layer_norm_epsilon |
| ClipVision.Projector | SCALE_FACTOR | clip.vision.projector.scale_factor |
| ClipVision.Projector | QUERY_SIDE | clip.vision.projector.query_side |
| ClipVision.Projector | WINDOW_SIDE | clip.vision.projector.window_side |
| ClipVision.Projector | SPATIAL_OFFSETS | clip.vision.projector.spatial_offsets |
| ClipVision.SAM | BLOCK_COUNT | clip.vision.sam.block_count |
| ClipVision.SAM | EMBEDDING_LENGTH | clip.vision.sam.embedding_length |
| ClipVision.SAM | HEAD_COUNT | clip.vision.sam.head_count |
| ClipAudio | PROJECTOR_TYPE | clip.audio.projector_type |
| ClipAudio | NUM_MEL_BINS | clip.audio.num_mel_bins |
| ClipAudio | EMBEDDING_LENGTH | clip.audio.embedding_length |
| ClipAudio | FEED_FORWARD_LENGTH | clip.audio.feed_forward_length |
| ClipAudio | PROJECTION_DIM | clip.audio.projection_dim |
| ClipAudio | BLOCK_COUNT | clip.audio.block_count |
| ClipAudio | SUBSAMPLING_FACTOR | clip.audio.subsampling_factor |
| ClipAudio | CHUNK_SIZE | clip.audio.chunk_size |
| ClipAudio | CONV_KERNEL_SIZE | clip.audio.conv_kernel_size |
| ClipAudio | HIDDEN_SIZES | clip.audio.hidden_sizes |
| ClipAudio | PATCH_SIZES | clip.audio.patch_sizes |
| ClipAudio | DEPTHS | clip.audio.depths |
| ClipAudio | MAX_POS_EMB | clip.audio.max_pos_emb |
| ClipAudio | FEATURE_LAYERS | clip.audio.feature_layer |
| ClipAudio | RVQ_NUM_QUANTIZERS | clip.audio.rvq.num_quantizers |
| ClipAudio | RVQ_CODEBOOK_SIZE | clip.audio.rvq.codebook_size |
| ClipAudio | WA_PATTERN_MODE | clip.audio.wa_pattern_mode |
| ClipAudio | WINDOW_SIZE | clip.audio.window_size |
| ClipAudio | LOCAL_BLOCK_COUNT | clip.audio.local_block_count |
| ClipAudio | LOCAL_GROUP_SIZE | clip.audio.local_group_size |
| ClipAudio.Attention | HEAD_COUNT | clip.audio.attention.head_count |
| ClipAudio.Attention | HEAD_DIM | clip.audio.attention.head_dim |
| ClipAudio.Attention | LAYERNORM_EPS | clip.audio.attention.layer_norm_epsilon |
| ClipAudio.Projector | STACK_FACTOR | clip.audio.projector.stack_factor |
| ClipAudio.Projector | WINDOW_SIZE | clip.audio.projector.window_size |
| ClipAudio.Projector | DOWNSAMPLE_RATE | clip.audio.projector.downsample_rate |
| ClipAudio.Projector | HEAD_COUNT | clip.audio.projector.head_count |
| ClipGenAudio | PROJECTOR_TYPE | clip.gen.audio.projector_type |
| ClipGenAudio | MODEL_VARIANT | clip.gen.audio.model_variant |
| ClipGenAudio | EMBEDDING_LENGTH | clip.gen.audio.embedding_length |
| ClipGenAudio | FEED_FORWARD_LENGTH | clip.gen.audio.feed_forward_length |
| ClipGenAudio | BLOCK_COUNT | clip.gen.audio.block_count |
| ClipGenAudio | PROJECTION_DIM | clip.gen.audio.projection_dim |
| ClipGenAudio.Attention | HEAD_COUNT | clip.gen.audio.attention.head_count |
| ClipGenAudio.Attention | HEAD_COUNT_KV | clip.gen.audio.attention.head_count_kv |
| ClipGenAudio.Attention | LAYERNORM_EPS | clip.gen.audio.attention.layer_norm_epsilon |
| Diffusion | SHIFT_LOGITS | diffusion.shift_logits |
| xIELU | ALPHA_P | xielu.alpha_p |
| xIELU | ALPHA_N | xielu.alpha_n |
| xIELU | BETA | xielu.beta |
| xIELU | EPS | xielu.eps |
| PrismHadamard | VERSION | prism.hadamard.version |
| PrismHadamard | TIED_OUTPUT | prism.hadamard.tied_output |
| PrismHadamard | BLOCK_SIZE | prism.hadamard.block_size |
| PrismHadamard | TRANSFORM | prism.hadamard.transform |
| PrismHadamard | AXIS | prism.hadamard.axis |
| PrismHadamard | SIGN_MODE | prism.hadamard.sign_mode |
| PrismHadamard | SIGN_WIDTHS | prism.hadamard.sign_widths |
| PrismHadamard | SIGN_VALUES | prism.hadamard.sign_values |
| PrismHadamard | WEIGHT_NAMES | prism.hadamard.weight_names |
| PrismHadamard | INVERSE_WEIGHT_NAMES | prism.hadamard.inverse_weight_names |
| PrismHadamard | GDN_V_GROUPED | prism.hadamard.gdn_v_grouped |
Nick Reynolds · @goldnode · rev 2.5