[None][feat] Add Inkling NVFP4 model support

docs/source/models/supported-models.md

@@ -30,6 +30,7 @@ The following is a table of supported models for the PyTorch backend:
 | `GptOssForCausalLM`                  | GPT-OSS                            | `openai/gpt-oss-20b`, `openai/gpt-oss-120b`  |
 | `HunYuanDenseForCausalLM` [^5]       | Hunyuan Dense                      | `tencent/Hunyuan-7B-Instruct`                |
 | `HunYuanMoEForCausalLM` [^5]         | Hunyuan MoE                        | `tencent/Hunyuan-A13B-Instruct`              |
+| `InklingForConditionalGeneration` [^15] | Inkling                         | `thinkingmachines/Inkling-NVFP4`, `thinkingmachines/Inkling-Small-NVFP4` |
 | `InternLM3ForCausalLM` [^5]          | InternLM3                          | `internlm/internlm3-8b-instruct`             |
 | `KimiK25ForConditionalGeneration`    | Kimi-K2.5                          | `moonshotai/Kimi-K2.5`                       |
 | `LagunaForCausalLM`                  | Laguna-XS                          | `poolside/laguna-XS.2`                       |
@@ -79,6 +80,7 @@ Note: Support for other models may vary. Features marked "N/A" are not applicabl
 | `Qwen3_5MoeForCausalLM`          | Yes               | Yes        | Yes                        | Yes                   | Yes             | Yes | No               | No                | No     | Yes           | Untested         | Yes            | N/A                      | Untested              | Untested        |
 | `Llama4ForConditionalGeneration` | Yes               | Yes        | Yes                        | Yes                   | Yes             | No  | Yes              | Yes               | No     | Yes           | Yes              | Untested       | N/A                      | Yes                   | Yes             |
 | `GptOssForCausalLM`              | Yes               | Yes        | Yes                        | Yes                   | Yes             | No  | Yes              | No                | Yes    | Yes           | Yes              | Yes            | N/A                      | Yes                   | Yes             |
+| `InklingForConditionalGeneration` [^15] | Yes       | Yes        | Yes                        | No                    | Yes             | No  | No               | No                | No     | Yes           | Untested         | Yes            | Yes                      | Yes                   | No              |
 | `Glm4MoeLiteForCausalLM` [^5]    | Yes               | Yes        | Untested                   | Untested              | Yes             | No  | No               | No                | No     | Yes           | Untested         | Untested       | N/A                      | Untested              | Untested        |
 | `NemotronHForCausalLM`           | Yes               | Yes        | Yes                        | Yes                   | Yes             | Yes | No               | No                | No     | Yes           | Yes              | Yes            | N/A                      | Untested              | Untested        |
 | `Gemma4ForConditionalGeneration` | Untested          | Yes        | Untested                   | No                    | Yes             | Yes | No               | No                | No     | Yes           | Untested         | No             | Yes                      | Untested              | Untested        |
@@ -99,6 +101,7 @@ Note: Support for other models may vary. Features marked "N/A" are not applicabl
 [^12]: Supports text, image, and video inputs over the block-sparse attention path. The published MXFP8 checkpoint is dequantized on load so the runtime sees an effectively BF16 model. The text decoder is also usable standalone (text-only) via the `MiniMaxM3SparseForCausalLM` architecture. KV cache reuse and MTP are not supported on the sparse-attention path in this release.
 [^13]: The Cosmos 3 family also supports visual generation through the VisualGen API. See [Visual Generation Models](#visual-generation-models).
 [^14]: Requires `transformers>=5.7.0`: MiniCPM-V 4.6 was upstreamed into transformers as a native model type (`minicpmv4_6`) and the checkpoint ships no remote code (`auto_map`) to fall back on. The Qwen3.5-hybrid text tower runs in BF16. Image, video, and text inputs are supported in this release (video reuses the same NaViT-packed vision path as image via `MiniCPMV4_6InputProcessor`).
+[^15]: NVFP4 checkpoint with a RoPE-free hybrid-attention text decoder and BF16 vision and audio towers; video is handled as multi-frame images. Use `--reasoning_parser inkling` for its typed-content thinking blocks. Enforced at load: `moe_expert_parallel_size` must divide 256, and `attn_backend` must stay in the `TRTLLM` family (the framework default). MTP, LoRA, function calling, constrained/guided decoding, disaggregated serving, and multimodal-hash prefix caching are not supported in this release. KV cache reuse additionally needs `kv_cache_config.mamba_state_config.periodic_snapshot_interval` set to a positive number of tokens, which snapshots the short-conv window a prefix hit has to restore; it applies to text prompts only. CUDA graph is supported, including under expert parallelism.
 
 # Encoder-Decoder Feature Support Matrix (PyTorch Backend)
 
@@ -127,6 +130,7 @@ complete encoder input must fit in the iteration token budget.
 | `Gemma4ForConditionalGeneration`     | Untested          | Yes        | Yes             | Yes           | Untested         | No             | Untested              | No                        | L + I + V + A [^9] |
 | `Gemma4UnifiedForConditionalGeneration` | Untested          | Untested   | Untested        | Yes           | Untested         | No             | Untested              | No                        | L + I + A |
 | `HCXVisionForCausalLM`               | Yes               | Yes        | No              | Yes           | Yes              | Yes            | Yes                   | No                        | L + I     |
+| `InklingForConditionalGeneration`    | Yes               | Yes        | Yes             | Yes           | Untested         | No             | Yes                   | No                        | L + I + V + A [^15] |
 | `LlavaLlamaModel (VILA)`             | Yes               | Yes        | No              | Yes           | Yes              | No             | Yes                   | No                        | L + I + V |
 | `LlavaNextForConditionalGeneration`  | Yes               | Yes        | Yes             | Yes           | Yes              | Yes            | Yes                   | Yes                       | L + I     |
 | `Llama4ForConditionalGeneration`     | Yes               | Yes        | No              | Yes           | Yes              | No             | Yes                   | No                        | L + I     |

tensorrt_llm/_torch/attention_backend/sparse/inkling/__init__.py

@@ -0,0 +1,72 @@
+# SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""Inkling attention: Triton kernels, backend, cache manager, conv-state pool.
+
+Inkling's attention is dense (full causal, sliding window on local layers, plus a
+learned relative-bias ``score_mod``), not sparse. It lives under ``sparse/``
+because that is where every model-specific attention backend already lives, and
+selection goes through ``sparse/registry.py`` like its neighbours.
+
+The attention metadata adds one thing to the base: the conv pool's per-step slot
+write, which has to run outside the CUDA-graph capture region. Everything else the
+decode kernel needs is a view of the base's own buffers (``page_table.py``).
+"""
+
+from .backend import InklingTritonAttention
+from .cache_manager import InklingHybridCacheManager
+from .conv_state import (
+    CONV_ROLE,
+    InklingConvRuntime,
+    InklingConvState,
+    InklingConvStateCache,
+    InklingRole,
+    apply_short_conv,
+)
+from .kernels import (
+    build_page_table,
+    inkling_decode_attention,
+    inkling_prefill_attention,
+    write_kv_cache_hnd,
+)
+from .metadata import InklingAttentionMetadata
+from .page_table import gen_page_table, gen_seq_lens, page_div
+from .params import (
+    InklingBackendForwardArgs,
+    InklingSparseAttentionConfig,
+    InklingSparseParams,
+    inkling_forward_args,
+)
+
+__all__ = [
+    "InklingAttentionMetadata",
+    "InklingBackendForwardArgs",
+    "InklingSparseAttentionConfig",
+    "InklingSparseParams",
+    "InklingConvRuntime",
+    "InklingConvState",
+    "InklingConvStateCache",
+    "InklingRole",
+    "InklingHybridCacheManager",
+    "InklingTritonAttention",
+    "apply_short_conv",
+    "build_page_table",
+    "gen_page_table",
+    "gen_seq_lens",
+    "inkling_forward_args",
+    "inkling_decode_attention",
+    "inkling_prefill_attention",
+    "page_div",
+    "write_kv_cache_hnd",
+]