Gal Hubara-Agam b2095aa074 [#4674][bugfix] AutoDeploy Fix memory leak in fuse_moe (#7844) ... Delete the unstacked weights immediately to save GPU memory, cleanup occurs automatically after the transformation, but for large models we'll run out of memory during the transformation itself. Signed-off-by: Gal Hubara Agam <96368689+galagam@users.noreply.github.com>		2025-09-29 11:01:07 +03:00
..
compile	[#7675][feat] CapturedGraph to support max_batch_size > max(cuda_graph_batch_sizes) (#7888)	2025-09-24 10:11:44 -04:00
config	[#4674][bugfix] AutoDeploy Fix memory leak in fuse_moe (#7844)	2025-09-29 11:01:07 +03:00
custom_ops	[#5860][autodeploy] GPT-OSS MXFP4 support (#7451)	2025-09-26 15:36:06 -07:00
distributed	[#7308] [feat] AutoDeploy: graph-less transformers mode for HF (#7635)	2025-09-18 10:44:24 +08:00
export	[None][chore] Upgrade transformers to 4.56.0 (#7523)	2025-09-22 22:20:16 +08:00
models	[#5860][autodeploy] GPT-OSS MXFP4 support (#7451)	2025-09-26 15:36:06 -07:00
shim	[TRTLLM-6393][feat] add static tree sampling and verification (#7161)	2025-09-26 13:16:16 -04:00
transform	[#4674][bugfix] AutoDeploy Fix memory leak in fuse_moe (#7844)	2025-09-29 11:01:07 +03:00
transformations	[#7308] [feat] AutoDeploy: graph-less transformers mode for HF (#7635)	2025-09-18 10:44:24 +08:00
utils	[#5860][autodeploy] GPT-OSS MXFP4 support (#7451)	2025-09-26 15:36:06 -07:00
__init__.py	[AutoDeploy] merge feat/ad-2025-07-07 (#6196)	2025-07-23 05:11:04 +08:00
llm_args.py	[TRTLLM-7728][feat] batched sampling by strategy (supersedes enable_mixed_sampler, cf. TRTLLM-7156) (#7294)	2025-09-23 16:05:05 -07:00
llm.py	[None][fix] Read eos_token_id from generation_config for kimi_k2 (#7120)	2025-09-23 10:47:03 +08:00