Shengliang Xu 470a97944b Address review: exporter fallback, Nemotron VL detection, changelog ...

- export_tensorrt_llm_checkpoint: when decoder_type cannot be detected,
  raise a clear ValueError unless the model has config.architectures for
  the generic decoder path. A Megatron-Core caller that omits decoder_type
  otherwise failed late with KeyError: 'unknown:GPTModel'.
- requantize_resmooth_fused_llm_layers: detect Nemotron VL from
  config.architectures, like is_multimodal_model and is_nemotron_vl, since
  model_type is free-form for these remote-code models.
- Drop the stale "(t5/bart/whisper)" from the layerwise-export refusal
  label; is_enc_dec also covers mt5, umt5 and mbart.
- Shorten the CHANGELOG entry to two sentences.

Signed-off-by: Shengliang Xu <shengliangx@nvidia.com>

2026-09-30 18:48:21 +00:00

..

_test_utils

[5/6] Add the IQ1_M codec (#2513)

2026-09-30 09:45:48 -07:00

examples

Merge remote-tracking branch 'origin-oss/main' into shengliangx/deprecating-trtllm-model_type

2026-09-30 18:44:20 +00:00

gpu

[4/5] Add the IQ2_S CUDA encoder and register the format (#2565)

2026-09-29 20:35:38 +00:00

gpu_megatron

Fix grouped expert quantizer checkpoint replicas (#2500)

2026-09-24 20:01:56 +00:00

gpu_trtllm

Fix 2-GPU test_model_load_utils hang; test import and fixture cleanup (#2079)

2026-08-05 21:00:42 +05:30

gpu_vllm

Fix vLLM compilation guard for models without marker (#2518)

2026-09-29 14:28:32 -07:00

regression/torch/speculative

Speed up slow unit/gpu/example tests (#1616)

2026-06-04 17:13:10 +00:00

unit

Address review: exporter fallback, Nemotron VL detection, changelog

2026-09-30 18:48:21 +00:00

conftest.py

[1/2] One MLflow tracking core behind a Tool record (#2544)

2026-09-28 19:57:03 +00:00