Search results

[Bug]: SamplingParams accepts NaN for top_p, min_p and temperature

Decoding/Sampling<NV>Token sampling algorithms in TRTLLM for text gen (top-k, top-p, beam).<NV>Token sampling algorithms in TRTLLM for text gen (top-k, top-p, beam).

Status: Open.

#17158 In NVIDIA/TensorRT-LLM;

· Yigtwxx opened on Aug 1, 2026

[Bug]: Streaming reasoning parser drops text buffered at the end of the stream

Status: Open.

#17156 In NVIDIA/TensorRT-LLM;

· Yigtwxx opened on Aug 1, 2026

[Bug]: Poolside's Laguna models' reasoning parser is broken: the </think> block leaks into the model's answer

bugSomething isn't workingSomething isn't working

Status: Open.

#17146 In NVIDIA/TensorRT-LLM;

· 1MrazorT1 opened on Jul 31, 2026

[Bug]: SM121 (GB10) excluded from FP8 block-scales SM120 Python paths while C++ getSMVersion() aliases 121→120

Customized kernels<NV>Specialized/modified CUDA kernels in TRTLLM for LLM ops, beyond standard TRT. Dev & perf.<NV>Specialized/modified CUDA kernels in TRTLLM for LLM ops, beyond standard TRT. Dev & perf.

Status: Open.

#17126 In NVIDIA/TensorRT-LLM;

· Smallfu666 opened on Jul 31, 2026

[Bug] DSpark: disaggregated gen-only benchmark deadlocks under attention-DP with asymmetric (gen-only) speculative decoding

Disaggregated serving<NV>Deploying with separated, distributed components (params, kv-cache, compute). Arch & perf.<NV>Deploying with separated, distributed components (params, kv-cache, compute). Arch & perf.

Speculative Decoding<NV>MTP/Eagle/Medusa/Lookahead/Prompt-Lookup-Decoding/Draft-Target-Model/ReDrafter<NV>MTP/Eagle/Medusa/Lookahead/Prompt-Lookup-Decoding/Draft-Target-Model/ReDrafter

Status: Open.

#17095 In NVIDIA/TensorRT-LLM;

· lishicheng1996-nv opened on Jul 31, 2026

[Bug]: GDN FlashInfer decode: initial_state_indices.int() is a no-op for int32 and leaves misaligned index slices (Misaligned Tensor data, alignment=32 bytes)

Status: Open.

#17027 In NVIDIA/TensorRT-LLM;

· Navjot10 opened on Jul 29, 2026

DeepSeek-V4 + MTP + attention DP: CUDA_ERROR_ILLEGAL_ADDRESS in CUDAGraph::replay() during executor init

Status: Open.

#17025 In NVIDIA/TensorRT-LLM;

· thorjohnsen opened on Jul 29, 2026

DeepSeek-V4 + MTP: AttributeError: '_num_tables' in DeepseekV4CacheManager.copy_batch_block_offsets during warmup

Status: Open.

#17024 In NVIDIA/TensorRT-LLM;

· thorjohnsen opened on Jul 29, 2026

trtllm-eval --max_output_length is silently discarded on text tasks whose yaml sets max_gen_toks

Status: Open.

#17022 In NVIDIA/TensorRT-LLM;

· thorjohnsen opened on Jul 29, 2026

DeepSeek-V4: OpenAI tools field reorders the system prompt after the tool schemas

Status: Open.

#17021 In NVIDIA/TensorRT-LLM;

· thorjohnsen opened on Jul 29, 2026

trtllm-eval aime25/aime26 silently disable thinking on DeepSeek-V4, halving reported accuracy

Status: Open.

#17020 In NVIDIA/TensorRT-LLM;

· thorjohnsen opened on Jul 29, 2026

[RFC]: Add openengine gRPC server support to trtllm-serve

RFC

Status: Open.

#17016 In NVIDIA/TensorRT-LLM;

· tanmayv25 opened on Jul 29, 2026