Public/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-08-26 15:16:23 +02:00
Code Issues Packages Projects Releases Wiki Activity
Files
d2fd589f514b13d2526fd9467f1ae53bff2aca46
llama.cpp/conversion/__init__.py
T
Piotr Wilkin (ilintar) ad1de39e07 model: add Kimi-K3 text model (#26185) 
* model: add Kimi-K3 text model

Hybrid KDA (linear) + MLA (full) attention as in Kimi-Linear-48B, plus five
things that architecture does not have:

  1. cross-layer residual attention  (attn_res_block_size)
  2. latent MoE                      (routed experts run at n_expert_latent)
  3. situ activation                 (replaces SwiGLU everywhere)
  4. MLA output gate                 (sigmoid gate before o_proj)
  5. full-rank KDA gate              (single ssm_g instead of ssm_g_a/ssm_g_b)

K3's text_config reports KimiLinearForCausalLM - the older 48B architecture -
so get_model_architecture routes on the top-level name instead.

The KDA decay gate has two forms, selected by linear_attn_config's
gate_lower_bound. It is not a clamp: when set it swaps the activation entirely
(fla/ops/kda/gate.py), from -exp(A_log)*softplus(x) to
lower_bound*sigmoid(exp(A_log)*x). K3 sets it to -5.0; kimi-linear leaves it
unset, so that path is unchanged.

Cross-layer residuals reuse ggml_dsv4_hc_pre for the weighted sum. That op is
CPU + CUDA only, so Metal/Vulkan will fall back per-node until those kernels
exist.

The routed experts ship as compressed-tensors "mxfp4-pack-quantized". That is
bit-compatible with ggml's MXFP4 - same E2M1 code assignment, same E8M0 scale
byte, only the nibble positions within a block differ - so they are repacked
rather than dequantized, losslessly and without a ~5.5 TB bf16 round-trip.
The repack is built lazily because gguf_writer holds every added tensor until
the final write. DeepSeek-V4 was already doing the identical bit-shuffling, so
it now shares the helper.

Verified against Moonshot's own code path (transformers + fla's Triton KDA
kernels) on a tiny model exercising every K3-specific feature. Final-position
logits vs the fp32 reference: 6.7e-05 rel / corr 1.00000000 for both the
chunked and the recurrent delta-net path. MXFP4 blocks dequantize to the source
weights with 0.0e+00 error.

Assisted-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

* model: fix ty errors in the Kimi-K3 converter

- `_res_parts` buffers (kind, tensor) pairs, not bare tensors
- `get_tensors` must return an Iterator, matching ModelBase
- LazyBase's `func` takes one argument, so pass the expert loaders through
  `args` instead of the closure
- borrowing KimiLinearModel.set_vocab from an unrelated TextModel is
  deliberate and safe, but not expressible in the signature

No behaviour change: the MXFP4 repack still dequantizes to the source weights
with 0.0e+00 error and end-to-end logits are unchanged (8.386e-03 rel,
corr 0.99996630).

Assisted-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

* Update conversion/kimi_k3.py

Co-authored-by: Boris Dvorkin  <b_dvorkin@niuitmo.ru>

* Increase LLAMA_MAX_EXPERTS from 512 to 1024

* tests : support for Kimi K3 in archs test

* chat : add Kimi K3 chat format (reasoning, content, typed tool calls)

K3's assistant output is an XTML-ish tagged format built by the template's
open_tag/close_tag macros. Two properties break generic parsing:

1. The generation prompt ends with open_tag('think'), so the completion
   starts inside the think section with no opening marker in the output
   (thinking_forced_open).
2. Only <|open|>/<|close|>/<|sep|>/<|end_of_msg|> are special tokens; tag
   names ("think", "response", "message") are ordinary text tokens.

Adds common_chat_params_init_kimi_k3 (PEG_NATIVE) with detection on the
marker trio, reasoning extraction, response unwrapping, and tool-call
parsing of the tools/call/argument tag structure with argument types
taken from the tool schema. Includes the K3 chat template fixture and 9
test-chat cases derived from real generations of the full 2.8T model.

Verified end-to-end against Kimi-K3-Q2_K (GrEarl/Kimi-K3-GGUF) on 8x B200:
content, reasoning_content, streaming deltas, and tool_calls all correct;
finish_reason stop/tool_calls as appropriate.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

* chat : add message_delimiters for Kimi K3

Per-role message-start markers for token-level span splitting. User and
assistant messages carry only the role attribute, so their full opener
(through <|sep|>) is used; system and tool messages continue with more
attributes (type=/tool=/index=), so those delimiters stop after the
role's closing quote. Verified against the K3 tiktoken vocabulary that
the closing quote is always a standalone token across all attribute
variants, so the token-level prefix match stays exact.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

* fix: apply nits from @ngxson and text fixes from @danielhanchen

* tests : added missing hyperparameters and tensors for Kimi K3 in test-llama-archs

* chore : move overly verbose header file comments to Kimi K3 source file

* tests : re-enabled KIMI_K3 in test-llama-archs for WebGPU backend

* model-saver : emit kda_gate_lower_bound for Kimi K3

Quick fix. The Kimi K3 loader reads kda_gate_lower_bound and gates a graph branch on it (it scales the KDA gate when the bound is above -INFINITY), but the model
saver never wrote the key, so a save->load roundtrip silently dropped it back to the -INFINITY default and changed the model's output. The real K3 config sets gate_lower_bound = -5.0.

I propose to emit it from the saver, and set it to -5.0 in the test-llama-archs K3 case so the roundtrip check exercises it (the roundtrip fails without the saver line).

* Refactor conditional for model architecture check

* tests : re-enabled (again) KIMI_K3 and MINIMAX_M3 in test-llama-archs for WebGPU backend

* fix code comments

* add template on conversion

* move repack_mxfp4_blocks to model base

* nits

* add_value_length

* optimize res_stack construction

* nits

---------

Co-authored-by: Boris Dvorkin <b_dvorkin@niuitmo.ru>
Co-authored-by: Stanisław Szymczyk <sszymczy@gmail.com>
Co-authored-by: Deepankar Singh <singh.deepankar39@gmail.com>
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: Caleb DeLeeuw <caleb.deleeuw@gmail.com>
Co-authored-by: Xuan Son Nguyen <son@huggingface.co>
2026-08-15 17:11:05 +02:00

383 lines
15 KiB
Python
Raw Blame History

from __future__ import annotations
from .base import (
ModelBase, TextModel, MmprojModel, ModelType, SentencePieceTokenTypes,
logger, _mistral_common_installed, _mistral_import_error_msg,
get_model_architecture, LazyTorchTensor,
)
from typing import Type
__all__ = [
"ModelBase", "TextModel", "MmprojModel", "ModelType", "SentencePieceTokenTypes",
"get_model_architecture", "LazyTorchTensor", "logger",
"_mistral_common_installed", "_mistral_import_error_msg",
"get_model_class", "print_registered_models", "load_all_models",
]
TEXT_MODEL_MAP: dict[str, str] = {
"AfmoeForCausalLM": "afmoe",
"LagunaForCausalLM": "laguna",
"ApertusForCausalLM": "llama",
"ArceeForCausalLM": "llama",
"ArcticForCausalLM": "arctic",
"AudioFlamingo3ForConditionalGeneration": "qwen",
"BaiChuanForCausalLM": "baichuan",
"BaichuanForCausalLM": "baichuan",
"BailingMoeForCausalLM": "bailingmoe",
"BailingMoeV2ForCausalLM": "bailingmoe",
"BambaForCausalLM": "granite",
"BertForMaskedLM": "bert",
"BertForSequenceClassification": "bert",
"BertModel": "bert",
"BitnetForCausalLM": "bitnet",
"BitNetForCausalLM": "bitnet",
"BloomForCausalLM": "bloom",
"BloomModel": "bloom",
"CamembertModel": "bert",
"ChameleonForCausalLM": "chameleon",
"ChameleonForConditionalGeneration": "chameleon",
"ChatGLMForConditionalGeneration": "chatglm",
"ChatGLMModel": "chatglm",
"CodeShellForCausalLM": "codeshell",
"CogVLMForCausalLM": "cogvlm",
"Cohere2MoeForCausalLM": "command_r",
"Cohere2ForCausalLM": "command_r",
"CohereForCausalLM": "command_r",
"DbrxForCausalLM": "dbrx",
"DeciLMForCausalLM": "deci",
"DeepseekForCausalLM": "deepseek",
"DeepseekOCRForCausalLM": "deepseek",
"DeepseekV2ForCausalLM": "deepseek",
"DeepseekV3ForCausalLM": "deepseek",
"DeepseekV32ForCausalLM": "deepseek",
"DFlashDraftModel": "qwen",
"Qwen3DSparkModel": "qwen",
"DeepseekV4ForCausalLM": "deepseek",
"DeepseekV4DSparkModel": "deepseek",
"DistilBertForMaskedLM": "bert",
"DistilBertForSequenceClassification": "bert",
"DistilBertModel": "bert",
"Dots1ForCausalLM": "dots1",
"DotsOCRForCausalLM": "qwen",
"DreamModel": "dream",
"Ernie4_5ForCausalLM": "ernie",
"Ernie4_5_ForCausalLM": "ernie",
"Ernie4_5_MoeForCausalLM": "ernie",
"EuroBertModel": "bert",
"Exaone4_5_ForConditionalGeneration": "exaone",
"Exaone4ForCausalLM": "exaone",
"ExaoneForCausalLM": "exaone",
"ExaoneMoEForCausalLM": "exaone",
"ExaoneMoeForCausalLM": "exaone",
"FalconForCausalLM": "falcon",
"FalconH1ForCausalLM": "falcon_h1",
"FalconMambaForCausalLM": "mamba",
"GPT2LMHeadModel": "gpt2",
"GPTBigCodeForCausalLM": "starcoder",
"GPTNeoXForCausalLM": "gptneox",
"GPTRefactForCausalLM": "refact",
"Gemma2ForCausalLM": "gemma",
"Gemma3ForCausalLM": "gemma",
"Gemma3ForConditionalGeneration": "gemma",
"Gemma3TextModel": "gemma",
"Gemma3nForCausalLM": "gemma",
"Gemma3nForConditionalGeneration": "gemma",
"Gemma4AssistantForCausalLM": "gemma",
"Gemma4ForConditionalGeneration": "gemma",
"Gemma4ForCausalLM": "gemma",
"Gemma4UnifiedForConditionalGeneration": "gemma",
"Gemma4UnifiedAssistantForCausalLM": "gemma",
"GemmaForCausalLM": "gemma",
"Glm4ForCausalLM": "glm",
"Glm4MoeForCausalLM": "glm",
"Glm4MoeLiteForCausalLM": "glm",
"Glm4vForConditionalGeneration": "glm",
"Glm4vMoeForConditionalGeneration": "glm",
"GlmForCausalLM": "chatglm",
"GlmMoeDsaForCausalLM": "glm",
"GlmOcrForConditionalGeneration": "glm",
"GptOssForCausalLM": "gpt_oss",
"GraniteForCausalLM": "granite",
"GraniteMoeForCausalLM": "granite",
"GraniteMoeHybridForCausalLM": "granite",
"GraniteMoeSharedForCausalLM": "granite",
"GraniteSwitchForCausalLM": "granite",
"GraniteSpeechForConditionalGeneration": "granite",
"GraniteSpeechPlusForConditionalGeneration": "granite",
"Grok1ForCausalLM": "grok",
"GrokForCausalLM": "grok",
"GroveMoeForCausalLM": "grovemoe",
"HunYuanDenseV1ForCausalLM": "hunyuan",
"HunYuanMoEV1ForCausalLM": "hunyuan",
"HunYuanVLForConditionalGeneration": "hunyuan",
"HYV3ForCausalLM": "hunyuan",
"IQuestCoderForCausalLM": "llama",
"InternLM2ForCausalLM": "internlm",
"InternLM3ForCausalLM": "internlm",
"JAISLMHeadModel": "jais",
"Jais2ForCausalLM": "jais",
"JambaForCausalLM": "jamba",
"JanusForConditionalGeneration": "januspro",
"JinaBertForMaskedLM": "bert",
"JinaBertModel": "bert",
"JinaEmbeddingsV5Model": "bert",
"KORMoForCausalLM": "qwen",
"KimiK25ForConditionalGeneration": "deepseek",
"KimiK3ForConditionalGeneration": "kimi_k3",
"KimiLinearForCausalLM": "kimi_linear",
"KimiLinearModel": "kimi_linear",
"KimiVLForConditionalGeneration": "deepseek",
"LFM2ForCausalLM": "lfm2",
"LLaDAMoEModel": "llada",
"LLaDAMoEModelLM": "llada",
"LLaDAModelLM": "llada",
"LLaMAForCausalLM": "llama",
"Lfm25AudioTokenizer": "lfm2",
"Lfm2BidirectionalModel": "lfm2",
"Lfm2ForCausalLM": "lfm2",
"Lfm2Model": "lfm2",
"Lfm2MoeForCausalLM": "lfm2",
"Llama4ForCausalLM": "llama",
"Llama4ForConditionalGeneration": "llama",
"LlamaBidirectionalModel": "llama",
"LlamaForCausalLM": "llama",
"LlamaModel": "llama",
"Eagle3DraftModel": "llama",
"Eagle3Speculator": "llama",
"Eagle3LlamaForCausalLM": "llama",
"LlamaForCausalLMEagle3": "llama",
"LlavaForConditionalGeneration": "llama",
"LlavaStableLMEpochForCausalLM": "stablelm",
"MPTForCausalLM": "mpt",
"MT5ForConditionalGeneration": "t5",
"MaincoderForCausalLM": "maincoder",
"Mamba2ForCausalLM": "mamba",
"MambaForCausalLM": "mamba",
"MambaLMHeadModel": "mamba",
"MellumForCausalLM": "mellum",
"MiMoV2FlashForCausalLM": "mimo",
"MiMoV2ForCausalLM": "mimo",
"MiniCPM3ForCausalLM": "minicpm",
"MiniCPMForCausalLM": "minicpm",
"MiniCPMV4_6ForConditionalGeneration": "minicpm",
"MiniMaxText01ForCausalLM": "minimax",
"MiniMaxM1ForCausalLM": "minimax",
"MiniMaxM2ForCausalLM": "minimax",
"MiniMaxM3SparseForCausalLM": "minimax",
"MiniMaxM3SparseForConditionalGeneration": "minimax",
"Ministral3ForCausalLM": "mistral3",
"Mistral3ForConditionalGeneration": "mistral3",
"MistralForCausalLM": "llama",
"MixtralForCausalLM": "llama",
"ModernBertForMaskedLM": "bert",
"ModernBertForSequenceClassification": "bert",
"ModernBertModel": "bert",
"NanbeigeForCausalLM": "nanbeige",
"NemotronForCausalLM": "nemotron",
"NemotronHForCausalLM": "nemotron",
"NeoBERT": "bert",
"NeoBERTForSequenceClassification": "bert",
"NeoBERTLMHead": "bert",
"NomicBertModel": "bert",
"OLMoForCausalLM": "olmo",
"Olmo2ForCausalLM": "olmo",
"Olmo3ForCausalLM": "olmo",
"OlmoForCausalLM": "olmo",
"OlmoeForCausalLM": "olmo",
"MuseGlimmerAssistantModel": "muse_glimmer",
"MuseGlimmerForConditionalGeneration": "muse_glimmer",
"OpenELMForCausalLM": "openelm",
"OrionForCausalLM": "orion",
"PLMForCausalLM": "plm",
"PLaMo2ForCausalLM": "plamo",
"PLaMo3ForCausalLM": "plamo",
"PaddleOCRVLForConditionalGeneration": "ernie",
"PanguEmbeddedForCausalLM": "pangu",
"Phi3ForCausalLM": "phi",
"Phi4ForCausalLMV": "phi",
"PhiForCausalLM": "phi",
"PhiMoEForCausalLM": "phi",
"Plamo2ForCausalLM": "plamo",
"Plamo3ForCausalLM": "plamo",
"PlamoForCausalLM": "plamo",
"QWenLMHeadModel": "qwen",
"Qwen2AudioForConditionalGeneration": "qwen",
"Qwen2ForCausalLM": "qwen",
"Qwen2Model": "qwen",
"Qwen2MoeForCausalLM": "qwen",
"Qwen2VLForConditionalGeneration": "qwenvl",
"Qwen2VLModel": "qwenvl",
"Qwen2_5OmniModel": "qwenvl",
"Qwen2_5_VLForConditionalGeneration": "qwenvl",
"Qwen3ASRForConditionalGeneration": "qwen3vl",
"Qwen3ForCausalLM": "qwen",
"Qwen3Model": "qwen",
"Qwen3MoeForCausalLM": "qwen",
"Qwen3NextForCausalLM": "qwen",
"Qwen3OmniMoeForConditionalGeneration": "qwen3vl",
"PocketTTSModel": "pockettts",
"Qwen3TTSForConditionalGeneration": "qwen3tts",
"Qwen3VLForConditionalGeneration": "qwen3vl",
"Qwen3VLMoeForConditionalGeneration": "qwen3vl",
"Qwen3_5ForCausalLM": "qwen",
"Qwen3_5ForConditionalGeneration": "qwen",
"Qwen3_5MoeForCausalLM": "qwen",
"Qwen3_5MoeForConditionalGeneration": "qwen",
"RND1": "qwen",
"RWForCausalLM": "falcon",
"RWKV6Qwen2ForCausalLM": "rwkv",
"RWKV7ForCausalLM": "rwkv",
"RobertaForSequenceClassification": "bert",
"RobertaModel": "bert",
"RuGPT3XLForCausalLM": "gpt2",
"Rwkv6ForCausalLM": "rwkv",
"Rwkv7ForCausalLM": "rwkv",
"RwkvHybridForCausalLM": "rwkv",
"Sarashina2VisionForCausalLM": "sarashina2",
"SarvamMoEForCausalLM": "bailingmoe",
"SeedOssForCausalLM": "olmo",
"SmallThinkerForCausalLM": "smallthinker",
"SmolLM3ForCausalLM": "llama",
"SolarOpenForCausalLM": "glm",
"StableLMEpochForCausalLM": "stablelm",
"StableLmForCausalLM": "stablelm",
"Starcoder2ForCausalLM": "starcoder",
"Step3p5ForCausalLM": "step3",
"StepVLForConditionalGeneration": "step3",
"Step3p7ForConditionalGeneration": "step3",
"T5EncoderModel": "t5",
"T5ForConditionalGeneration": "t5",
"T5WithLMHeadModel": "t5",
"TalkieForCausalLM": "talkie",
"UMT5ForConditionalGeneration": "t5",
"UMT5Model": "t5",
"UltravoxModel": "ultravox",
"UnlimitedOCRForCausalLM": "deepseek",
"VLlama3ForCausalLM": "llama",
"VoxtralForConditionalGeneration": "llama",
"WavTokenizerDec": "wavtokenizer",
"XLMRobertaForSequenceClassification": "bert",
"XLMRobertaModel": "bert",
"XverseForCausalLM": "xverse",
"YoutuForCausalLM": "deepseek",
"YoutuVLForConditionalGeneration": "deepseek",
"modeling_grove_moe.GroveMoeForCausalLM": "grovemoe",
"modeling_sarvam_moe.SarvamMoEForCausalLM": "bailingmoe",
}
MMPROJ_MODEL_MAP: dict[str, str] = {
"AudioFlamingo3ForConditionalGeneration": "ultravox",
"CogVLMForCausalLM": "cogvlm",
"DeepseekOCR2ForCausalLM": "deepseek",
"DeepseekOCRForCausalLM": "deepseek",
"DotsOCRForCausalLM": "dotsocr",
"Exaone4_5_ForConditionalGeneration": "exaone",
"Gemma3ForConditionalGeneration": "gemma",
"Gemma3nForConditionalGeneration": "gemma",
"Gemma4ForConditionalGeneration": "gemma",
"Gemma4UnifiedForConditionalGeneration": "gemma",
"Glm4vForConditionalGeneration": "qwen3vl",
"Glm4vMoeForConditionalGeneration": "qwen3vl",
"Glm5vForConditionalGeneration": "kimivl",
"GlmOcrForConditionalGeneration": "qwen3vl",
"GlmasrModel": "ultravox",
"Granite4VisionForConditionalGeneration": "granite",
"GraniteSpeechForConditionalGeneration": "granite",
"GraniteSpeechPlusForConditionalGeneration": "granite",
"HunYuanVLForConditionalGeneration": "hunyuan",
"Idefics3ForConditionalGeneration": "smolvlm",
"InternVisionModel": "internvl",
"JanusForConditionalGeneration": "januspro",
"KimiK25ForConditionalGeneration": "kimivl",
"KimiVLForConditionalGeneration": "kimivl",
"Lfm2AudioForConditionalGeneration": "lfm2",
"Lfm2VlForConditionalGeneration": "lfm2",
"LightOnOCRForConditionalGeneration": "lighton_ocr",
"Llama4ForConditionalGeneration": "llama4",
"LlavaForConditionalGeneration": "llava",
"MERaLiON2ForConditionalGeneration": "ultravox",
"MiMoV2ForCausalLM": "mimo",
"MiniMaxM3SparseForConditionalGeneration": "minimax",
"MiniCPMV4_6ForConditionalGeneration": "minicpm",
"Mistral3ForConditionalGeneration": "llava",
"NemotronH_Nano_VL_V2": "nemotron",
"MuseGlimmerForConditionalGeneration": "muse_glimmer",
"PaddleOCRVisionModel": "ernie",
"Phi4ForCausalLMV": "phi",
"Qwen2AudioForConditionalGeneration": "ultravox",
"Qwen2VLForConditionalGeneration": "qwenvl",
"Qwen2VLModel": "qwenvl",
"Qwen2_5OmniModel": "qwenvl",
"Qwen2_5_VLForConditionalGeneration": "qwenvl",
"Qwen3ASRForConditionalGeneration": "qwen3vl",
"Qwen3OmniMoeForConditionalGeneration": "qwen3vl",
"PocketTTSModel": "pockettts",
"Qwen3TTSForConditionalGeneration": "qwen3tts",
"Qwen3VLForConditionalGeneration": "qwen3vl",
"Qwen3VLMoeForConditionalGeneration": "qwen3vl",
"Qwen3_5ForConditionalGeneration": "qwen3vl",
"Qwen3_5MoeForConditionalGeneration": "qwen3vl",
"RADIOModel": "nemotron",
"Sarashina2VisionForCausalLM": "sarashina2",
"SmolVLMForConditionalGeneration": "smolvlm",
"StepVLForConditionalGeneration": "step3",
"Step3p7ForConditionalGeneration": "step3",
"UltravoxModel": "ultravox",
"UnlimitedOCRForCausalLM": "deepseek",
"VoxtralForConditionalGeneration": "ultravox",
"YoutuVLForConditionalGeneration": "youtuvl",
}
_TEXT_MODEL_MODULES = sorted(set(TEXT_MODEL_MAP.values()))
_MMPROJ_MODEL_MODULES = sorted(set(MMPROJ_MODEL_MAP.values()))
_loaded_text_modules: set[str] = set()
_loaded_mmproj_modules: set[str] = set()
def load_all_models() -> None:
"""Import all model modules to trigger @ModelBase.register() decorators."""
if len(_loaded_text_modules) != len(_TEXT_MODEL_MODULES):
for module_name in _TEXT_MODEL_MODULES:
if module_name not in _loaded_text_modules:
try:
__import__(f"conversion.{module_name}")
_loaded_text_modules.add(module_name)
except Exception as e:
logger.warning(f"Failed to load model module {module_name}: {e}")
if len(_loaded_mmproj_modules) != len(_MMPROJ_MODEL_MODULES):
for module_name in _MMPROJ_MODEL_MODULES:
if module_name not in _loaded_mmproj_modules:
try:
__import__(f"conversion.{module_name}")
_loaded_mmproj_modules.add(module_name)
except Exception as e:
logger.warning(f"Failed to load model module {module_name}: {e}")
def get_model_class(name: str, mmproj: bool = False) -> Type[ModelBase]:
"""Dynamically import and return a model class by its HuggingFace architecture name."""
relevant_map = MMPROJ_MODEL_MAP if mmproj else TEXT_MODEL_MAP
if name not in relevant_map:
raise NotImplementedError(f"Architecture {name!r} not supported!")
module_name = relevant_map[name]
__import__(f"conversion.{module_name}")
model_type = ModelType.MMPROJ if mmproj else ModelType.TEXT
return ModelBase._model_classes[model_type][name]
def print_registered_models() -> None:
load_all_models()
logger.error("TEXT models:")
for name in sorted(TEXT_MODEL_MAP.keys()):
logger.error(f" - {name}")
logger.error("MMPROJ models:")
for name in sorted(MMPROJ_MODEL_MAP.keys()):
logger.error(f" - {name}")
Reference in New Issue View Git Blame Copy Permalink