[Model] Remove native LongCat implementations

CMakeLists.txt

@@ -429,7 +429,6 @@ if(VLLM_GPU_LANG STREQUAL "CUDA" OR VLLM_GPU_LANG STREQUAL "HIP")
     "csrc/libtorch_stable/cuda_view.cu"
     "csrc/libtorch_stable/cuda_utils_kernels.cu"
     "csrc/libtorch_stable/activation_kernels.cu"
-    "csrc/libtorch_stable/ngram_embedding_kernels.cu"
     "csrc/libtorch_stable/quantization/activation_kernels.cu"
     "csrc/libtorch_stable/quantization/w8a8/int8/scaled_quant.cu"
     "csrc/libtorch_stable/quantization/w8a8/fp8/common.cu"

csrc/libtorch_stable/ops.h

@@ -693,12 +693,3 @@ void cp_gather_indexer_k_quant_cache(
                                                 // quant_block_size * 4]
     const torch::stable::Tensor& block_table,   // [batch_size, num_blocks]
     const torch::stable::Tensor& cu_seq_lens);  // [batch_size + 1]
-
-// LongCat n-gram embedding index kernel (see ngram_embedding_kernels.cu).
-void ngram_compute_n_gram_ids(
-    int64_t ne_n, int64_t ne_k, torch::stable::Tensor& ne_weights,
-    torch::stable::Tensor& ne_mods,
-    torch::stable::Tensor& exclusive_ne_embedder_size_sums,
-    torch::stable::Tensor& exclusive_req_len_sums,
-    torch::stable::Tensor& ne_token_table, torch::stable::Tensor& row_indices,
-    torch::stable::Tensor& column_starts, torch::stable::Tensor& n_gram_ids);

csrc/libtorch_stable/torch_bindings.cpp

@@ -695,22 +695,9 @@ STABLE_TORCH_LIBRARY_FRAGMENT(_C, ops) {
       "Tensor? initial_state_idx,"
       "Tensor? cu_chunk_seqlen,"
       "Tensor? last_chunk_indices) -> ()");
-
-  // LongCat n-gram embedding index kernel. All tensor args are marked mutable
-  // to match the (non-const) stable-Tensor& C++ signature; only ne_token_table
-  // and n_gram_ids are actually written in place.
-  ops.def(
-      "ngram_compute_n_gram_ids(int ne_n, int ne_k, Tensor(a!) ne_weights, "
-      "Tensor(b!) ne_mods, Tensor(c!) exclusive_ne_embedder_size_sums, "
-      "Tensor(d!) exclusive_req_len_sums, Tensor(e!) ne_token_table, "
-      "Tensor(f!) row_indices, Tensor(g!) column_starts, "
-      "Tensor(h!) n_gram_ids) -> ()");
 }
 
 STABLE_TORCH_LIBRARY_IMPL(_C, CUDA, ops) {
-  // LongCat n-gram embedding index kernel.
-  ops.impl("ngram_compute_n_gram_ids", TORCH_BOX(&ngram_compute_n_gram_ids));
-
   // Per-token group quantization
   ops.impl("per_token_group_fp8_quant", TORCH_BOX(&per_token_group_quant_fp8));
   ops.impl("per_token_group_fp8_quant_packed",

docs/models/supported_models.md

@@ -417,7 +417,6 @@ th {
 | `Lfm2ForCausalLM` | LFM2 | `LiquidAI/LFM2-1.2B`, `LiquidAI/LFM2-700M`, `LiquidAI/LFM2-350M`, etc. | ✅︎ | ✅︎ |
 | `Lfm2MoeForCausalLM` | LFM2MoE | `LiquidAI/LFM2-8B-A1B-preview`, etc. | ✅︎ | ✅︎ |
 | `LlamaForCausalLM` | Llama 3.1, Llama 3, Llama 2, LLaMA, Yi | `meta-llama/Meta-Llama-3.1-405B-Instruct`, `meta-llama/Meta-Llama-3.1-70B`, `meta-llama/Meta-Llama-3-70B-Instruct`, `meta-llama/Llama-2-70b-hf`, `01-ai/Yi-34B`, etc. | ✅︎ | ✅︎ |
-| `LongcatFlashForCausalLM` | LongCat-Flash | `meituan-longcat/LongCat-Flash-Chat`, `meituan-longcat/LongCat-Flash-Chat-FP8` | ✅︎ | ✅︎ |
 | `MambaForCausalLM` | Mamba | `state-spaces/mamba-130m-hf`, `state-spaces/mamba-790m-hf`, `state-spaces/mamba-2.8b-hf`, etc. | | ✅︎ |
 | `Mamba2ForCausalLM` | Mamba2 | `mistralai/Mamba-Codestral-7B-v0.1`, etc. | | ✅︎ |
 | `MellumForCausalLM` | Mellum 2 | `JetBrains/Mellum2-12B-A2.5B-Base`, etc. | | ✅︎ |
@@ -471,13 +470,17 @@ Some models are supported only via the [Transformers modeling backend](#transfor
 | `GPTBigCodeForCausalLM` | StarCoder, SantaCoder, WizardCoder | `bigcode/starcoder`, `bigcode/gpt_bigcode-santacoder`, `WizardLM/WizardCoder-15B-V1.0`, etc. | ✅︎ | |
 | `HunYuanDenseV1ForCausalLM` | Hunyuan Dense | `tencent/Hunyuan-7B-Instruct` | ✅︎ | ✅︎ |
 | `HunYuanMoEV1ForCausalLM` | Hunyuan-A13B | `tencent/Hunyuan-A13B-Instruct`, `tencent/Hunyuan-A13B-Pretrain`, `tencent/Hunyuan-A13B-Instruct-FP8`, etc. | ✅︎ | ✅︎ |
+| `LongcatFlashForCausalLM` | LongCat-Flash | `meituan-longcat/LongCat-Flash-Chat`, `meituan-longcat/LongCat-Flash-Chat-FP8` | ✅︎ | ✅︎ |
+| `LongcatFlashNgramForCausalLM` | LongCat-Flash-Lite | `meituan-longcat/LongCat-Flash-Lite` | ✅︎ | ✅︎ |
 | `OlmoForCausalLM` | OLMo | `allenai/OLMo-1B-hf`, `allenai/OLMo-7B-hf`, etc. | ✅︎ | ✅︎ |
 | `Olmo2ForCausalLM` | OLMo2 | `allenai/OLMo-2-0425-1B`, etc. | ✅︎ | ✅︎ |
 | `Olmo3ForCausalLM` | OLMo3 | `allenai/Olmo-3-7B-Instruct`, `allenai/Olmo-3-32B-Think`, etc. | ✅︎ | ✅︎ |
 | `SmolLM3ForCausalLM` | SmolLM3 | `HuggingFaceTB/SmolLM3-3B` | ✅︎ | ✅︎ |
 | `Starcoder2ForCausalLM` | Starcoder2 | `bigcode/starcoder2-3b`, `bigcode/starcoder2-7b`, `bigcode/starcoder2-15b`, etc. | ✅︎ | ✅︎ |
 | `VaultGemmaForCausalLM` | VaultGemma | `google/vaultgemma-1b` | ✅︎ | ✅︎ |
 
+LongCat-Flash-Lite requires `--trust-remote-code` for its custom n-gram model.
+
 !!! note
     Currently, the ROCm version of vLLM supports Mistral and Mixtral only for context lengths up to 4096.
 

tests/config/test_model_arch_config.py

@@ -28,7 +28,6 @@
     "stepfun-ai/Step-3.5-Flash",
     # Excluded: Not available online right now
     # "FreedomIntelligence/openPangu-Ultra-MoE-718B-V1.1",
-    "meituan-longcat/LongCat-Flash-Chat",
 }
 
 BASE_MODELS_TO_TEST = [
@@ -51,6 +50,7 @@
     "lmsys/gpt-oss-20b-bf16",
     "deepseek-ai/DeepSeek-V3.2-Exp",
     "meta-llama/Llama-4-Scout-17B-16E-Instruct",
+    "meituan-longcat/LongCat-Flash-Chat",
 ] + list(BASE_TRUST_REMOTE_CODE_MODELS)
 
 # (target_model, draft_model, trust_remote_code)

tests/models/registry.py

@@ -392,11 +392,13 @@ def check_available_online(
         "meta-llama/Llama-4-Scout-17B-16E-Instruct",
     ),
     "LongcatFlashForCausalLM": _HfExamplesInfo(
-        "meituan-longcat/LongCat-Flash-Chat", trust_remote_code=True
+        "meituan-longcat/LongCat-Flash-Chat",
+        min_transformers_version="5.15.0",
     ),
     "LongcatFlashNgramForCausalLM": _HfExamplesInfo(
         "meituan-longcat/LongCat-Flash-Lite",
         trust_remote_code=True,
+        min_transformers_version="5.15.0",
         # Shrink the ~62GB n-gram tables (ngram_vocab_size_ratio * vocab_size)
         # so the dummy-weight init test fits in CI memory.
         hf_overrides={"ngram_vocab_size_ratio": 1},
@@ -1742,11 +1744,6 @@ def check_available_online(
         trust_remote_code=True,
         is_available_online=False,
     ),
-    "LongCatFlashMTPModel": _HfExamplesInfo(
-        "meituan-longcat/LongCat-Flash-Chat",
-        trust_remote_code=True,
-        speculative_model="meituan-longcat/LongCat-Flash-Chat",
-    ),
     "MiMoMTPModel": _HfExamplesInfo(
         "XiaomiMiMo/MiMo-7B-RL",
         trust_remote_code=True,

tests/models/transformers/fusers/test_moe.py

@@ -84,6 +84,23 @@ def forward(self, hidden_states):
         return logits, scores, index
 
 
+class NestedRouter(nn.Module):
+    """LongCat-style router with its projection under `classifier`."""
+
+    def __init__(self, num_experts=8, hidden=16, top_k=2):
+        super().__init__()
+        self.top_k = top_k
+        self.classifier = nn.Linear(hidden, num_experts, bias=False)
+        self.register_buffer("e_score_correction_bias", torch.zeros(num_experts))
+
+    def forward(self, hidden_states):
+        logits = F.linear(hidden_states.float(), self.classifier.weight.float())
+        scores = F.softmax(logits, dim=-1)
+        _, index = torch.topk(scores + self.e_score_correction_bias, self.top_k, dim=-1)
+        weights = scores.gather(1, index)
+        return weights.to(hidden_states.dtype), index
+
+
 class BiasedRouter(TopKRouter):
     """A valid top-k router but not `weight`-only (extra `bias` param) -> declined."""
 
@@ -295,6 +312,17 @@ def test_moe_fuser_matches_correction_router():
     assert fuser.scoring_func == "sigmoid"
 
 
+def test_moe_fuser_matches_nested_router_projection():
+    with torch.device("meta"):
+        block = MoEBlock(NestedRouter)
+    fuser = MoEBlockFuser.match(block, "experts")
+    assert isinstance(fuser, MoEBlockFuser)
+    assert fuser.gate_name == "gate"
+    assert fuser.gate_linear_name == "classifier"
+    assert fuser.scoring_func == "softmax"
+    assert fuser.router_dtype == torch.float32
+
+
 def test_moe_fuser_reads_router_dtype_from_the_gate():
     """A router that computes its logits in fp32 must keep routing in fp32 when
     rebuilt, even though no config field names the dtype. The cast back to the

tests/models/utils.py

@@ -545,7 +545,6 @@ def _supports_multimodal_for_mm_prefix() -> bool:
     # from num_layers for its dual-attention layers).
     if model_arch not in (
         "LongcatFlashForCausalLM",
-        "LongCatFlashMTPModel",
         "LongcatFlashNgramForCausalLM",
     ):
         update_dict["num_hidden_layers"] = num_hidden_layers

tests/transformers_utils/test_config.py

@@ -6,6 +6,7 @@
 `BaseRenderer.get_eos_token_id`.
 """
 
+import json
 from types import SimpleNamespace
 from typing import cast
 from unittest.mock import MagicMock, patch
@@ -16,11 +17,39 @@
 from vllm.tokenizers import get_tokenizer
 from vllm.transformers_utils import config as config_module
 from vllm.transformers_utils.config import (
+    get_config,
     get_safetensors_params_metadata,
     try_get_generation_config,
 )
 
 
+def test_longcat_config_without_model_type_uses_builtin_transformers(tmp_path):
+    config_path = tmp_path / "config.json"
+    config_path.write_text(
+        json.dumps(
+            {
+                "architectures": ["LongcatFlashForCausalLM"],
+                "expert_ffn_hidden_size": 1024,
+                "num_hidden_layers": 56,
+                "num_layers": 14,
+                "auto_map": {
+                    "AutoConfig": "configuration_longcat_flash.LongcatFlashConfig",
+                    "AutoModel": "modeling_longcat_flash.LongcatFlashModel",
+                },
+            }
+        ),
+        encoding="utf-8",
+    )
+
+    config = get_config(tmp_path, trust_remote_code=False)
+
+    assert type(config).__name__ == "LongcatFlashConfig"
+    assert config.model_type == "longcat_flash"
+    assert config.auto_map == {}
+    assert config.num_hidden_layers == 28
+    assert config.moe_intermediate_size == 1024
+
+
 def test_get_llama3_eos_token():
     model_name = "meta-llama/Llama-3.2-1B-Instruct"
 

vllm/_custom_ops.py

@@ -235,37 +235,6 @@ def rms_norm(
     torch.ops._C.rms_norm(out, input, weight, epsilon)
 
 
-# LongCat n-gram embedding index kernel (see csrc/.../ngram_embedding_kernels.cu).
-def ngram_compute_n_gram_ids(
-    ne_n: int,
-    ne_k: int,
-    ne_weights: torch.Tensor,
-    ne_mods: torch.Tensor,
-    exclusive_ne_embedder_size_sums: torch.Tensor,
-    exclusive_req_len_sums: torch.Tensor,
-    ne_token_table: torch.Tensor,
-    row_indices: torch.Tensor,
-    column_starts: torch.Tensor,
-    n_gram_ids: torch.Tensor,
-) -> None:
-    """Compute concatenated (offset) n-gram ids for a ragged prefill batch.
-
-    Writes ``n_gram_ids`` of shape ``[token_num, (ne_n-1)*ne_k]``.
-    """
-    torch.ops._C.ngram_compute_n_gram_ids(
-        ne_n,
-        ne_k,
-        ne_weights,
-        ne_mods,
-        exclusive_ne_embedder_size_sums,
-        exclusive_req_len_sums,
-        ne_token_table,
-        row_indices,
-        column_starts,
-        n_gram_ids,
-    )
-
-
 def fused_add_rms_norm(
     input: torch.Tensor,
     residual: torch.Tensor,

vllm/config/speculative.py

@@ -48,7 +48,6 @@
     "exaone4_5_mtp",
     "qwen3_next_mtp",
     "qwen3_5_mtp",
-    "longcat_flash_mtp",
     "bailing_hybrid_v3_mtp",
     "minimax_m3_mtp",
     "bailing_hybrid_mtp",
@@ -886,13 +885,6 @@ def hf_config_override(hf_config: PretrainedConfig) -> PretrainedConfig:
                     "architectures": [architecture],
                 }
             )
-        if hf_config.model_type in ("longcat_flash", "longcat_flash_ngram"):
-            hf_config.model_type = "longcat_flash_mtp"
-            n_predict = getattr(hf_config, "num_nextn_predict_layers", 1)
-            hf_config.update(
-                {"n_predict": n_predict, "architectures": ["LongCatFlashMTPModel"]}
-            )
-
         if hf_config.model_type in ("step3p5", "step3p7") or hf_config.architectures[
             0
         ] in ("Step3p5ForCausalLM", "Step3p7ForConditionalGeneration"):