William Zhang a6a88985cf [TRTLLM-9409][feat] Pass MRoPE tensors for EPD disagg (#9758) ... * Why? Certain VLMs like the Qwen family need more than just the multimodal embeddings in the language model, and need MRoPE position IDs and deltas. Prior to this commit, only the embeddings could be communicated from the encoder worker to the prefill worker. * What? This commit extends the `DisaggregatedParams` to include the MRoPE information. It also adjusts several pieces of code required to communicate that between E, P and D workers. Closes TRTLLM-9409. Signed-off-by: William Zhang <133824995+2ez4bz@users.noreply.github.com>		2025-12-22 06:32:49 -05:00
..
attention_backend	[None][feat] Cudagraph updates for helix parallelism (#10141)	2025-12-21 15:21:52 -05:00
auto_deploy	[TRTLLM-7736][feat] Incrementally update the inputs of target and draft models (#9708)	2025-12-19 15:11:25 +08:00
compilation	[TRTLLM-9819][perf] Reuse alltoall workspace for CuteDSL MoE output (#9840)	2025-12-18 10:36:38 -08:00
configs	[TRTLLM-8535][feat] Support DeepSeek V3.2 with FP8 + BF16 KV cache/NVFP4 + BF16 KV cache (#8405)	2025-10-24 13:40:41 -04:00
custom_ops	[TRTLLM-9992][perf] Enable PDL for CuteDSL kernels and overlap MoeOutputMemset (#10043)	2025-12-20 03:12:41 -05:00
cute_dsl_kernels	[TRTLLM-9992][perf] Enable PDL for CuteDSL kernels and overlap MoeOutputMemset (#10043)	2025-12-20 03:12:41 -05:00
debug	Add debug hook to support dump tensor data and add new debug functions easily (#5182)	2025-06-24 17:45:28 +08:00
distributed	[None][feat] Async pp send for PPCommTorch. (#9976)	2025-12-15 14:03:46 +08:00
models	[TRTLLM-9409][feat] Pass MRoPE tensors for EPD disagg (#9758)	2025-12-22 06:32:49 -05:00
modules	[None][fix] NVFP4 linear method's weight and weight_scale padding (#10148)	2025-12-22 15:00:31 +08:00
peft	[TRTLLM-7346][fix] Improve performance of PyTorchModelEngine._get_lora_params_from_requests (#7033)	2025-08-25 10:37:40 +03:00
pyexecutor	[TRTLLM-9409][feat] Pass MRoPE tensors for EPD disagg (#9758)	2025-12-22 06:32:49 -05:00
shared_tensor	[1/N][TRTLLM-5195][feat] Share PyTorch tensor between processes (#5396)	2025-07-10 05:12:53 +09:00
speculative	[None][feat] Support Eagle3 on Mistral Large3 (#9971)	2025-12-21 10:25:45 -05:00
__init__.py	[TRTLLM-9212][chore] move MoeLoadBalancerConfig to llm_args.py (#9002)	2025-11-13 10:47:35 +08:00
async_llm.py	[TRTLLM-9736][feat] AsyncLLM and verl integ (#9353)	2025-12-11 09:33:25 -08:00
autotuner.py	[None][chore] Remove unnecessary warning log for tuning. (#10077)	2025-12-17 01:51:17 -08:00
cublaslt_utils.py	[https://nvbugs/5451205][feat] Add cuBLASLt NVFP4 GEMM backend support (#7943)	2025-10-23 15:55:10 +08:00
cute_dsl_utils.py	[None][chore] polish error message in cute_dsl_utils.py (#7852)	2025-09-19 12:05:11 +08:00
device_mesh.py	[None][feat] Async pp send for PPCommTorch. (#9976)	2025-12-15 14:03:46 +08:00
expert_statistic.py	[TRTLLM-8831][feat] Enable early exit with overlap scheduler (#8587)	2025-11-17 18:07:13 +01:00
flashinfer_utils.py	[TRTC-1943][feat] Env vars override support in LLM API (#9104)	2025-12-01 10:04:49 -08:00
hostfunc.py	[TRTLLM-7028][feat] Enable guided decoding with speculative decoding (part 2: one-model engine) (#6948)	2025-09-03 15:16:11 -07:00
llm.py	[TRTLLM-5208][BREAKING CHANGE] chore: make pytorch LLM the default (#5312)	2025-06-20 03:01:10 +08:00
memory_buffer_utils.py	[https://nvbugs/5629833][fix] Don't fill tensors with 0 (#9296)	2025-11-21 20:50:05 +08:00
metadata.py	[None][feat] Use Separate QKV Input Layout for Context MLA (#6538)	2025-08-19 22:04:48 +08:00
model_config.py	[TRTLLM-9416][feat] Skip DS-v3.2 indexer MQA and Top-K for short sequences. (#9524)	2025-12-15 12:42:25 +08:00
utils.py	[TRTLLM-9992][perf] Enable PDL for CuteDSL kernels and overlap MoeOutputMemset (#10043)	2025-12-20 03:12:41 -05:00
virtual_memory.py	[TRTLLM-9736][feat] AsyncLLM and verl integ (#9353)	2025-12-11 09:33:25 -08:00