TensorRT-LLMs/tensorrt_llm/_torch/speculative
Netanel Haber 6ee94c7ac8
Reintroduce with perf fixes: feature: unify new_tokens format sample state to trtllm samper tokens format (#5513)
58a8a8f - these changes were previously merged to main here.
6aef149 - the changes were temporarily reverted in main, due to a significant perf regression in models using the TorchSampler (observed by @byshiue).
This PR is meant to re-merge these changes along with a fix to prevent the regression.

The first commit of this PR is actually just the reverted revert - filter it out of the changes to see previously unmerged changes.

Signed-off-by: Netanel Haber <nhaber@nvidia.com>
2025-06-30 11:58:59 -07:00
..
__init__.py [TRTLLM-5000][feat] NGrams V2 (#4569) 2025-06-27 23:00:17 +08:00
draft_target.py Speculation: Draft Target in new FW (#4558) 2025-06-17 02:26:08 +08:00
drafter.py [TRTLLM-5000][feat] NGrams V2 (#4569) 2025-06-27 23:00:17 +08:00
eagle3.py Reintroduce with perf fixes: feature: unify new_tokens format sample state to trtllm samper tokens format (#5513) 2025-06-30 11:58:59 -07:00
interface.py [TRTLLM-5000][feat] NGrams V2 (#4569) 2025-06-27 23:00:17 +08:00
mtp.py Reintroduce with perf fixes: feature: unify new_tokens format sample state to trtllm samper tokens format (#5513) 2025-06-30 11:58:59 -07:00
ngram.py [TRTLLM-5000][feat] NGrams V2 (#4569) 2025-06-27 23:00:17 +08:00
utils.py Reintroduce with perf fixes: feature: unify new_tokens format sample state to trtllm samper tokens format (#5513) 2025-06-30 11:58:59 -07:00