| .. |
|
__init__.py
|
[Bugfix] Add __init__.py files for vllm/core/block/ and vllm/spec_decode/ (#3798)
|
2024-04-02 12:35:31 -07:00 |
|
batch_expansion.py
|
[Speculative Decoding] Fixing hidden states handling in batch expansion (#7508)
|
2024-08-19 17:58:14 -07:00 |
|
draft_model_runner.py
|
[Speculative Decoding] EAGLE Implementation with Top-1 proposer (#6830)
|
2024-08-22 02:42:24 -07:00 |
|
interfaces.py
|
[BUGFIX] Raise an error for no draft token case when draft_tp>1 (#6369)
|
2024-07-19 06:01:09 -07:00 |
|
medusa_worker.py
|
[Speculative decoding] [Multi-Step] decouple should_modify_greedy_probs_inplace (#6971)
|
2024-08-09 05:42:45 +00:00 |
|
metrics.py
|
[Core] Optimize SPMD architecture with delta + serialization optimization (#7109)
|
2024-08-18 17:57:20 -07:00 |
|
mlp_speculator_worker.py
|
[BugFix] Fix use of per-request seed with pipeline parallel (#6698)
|
2024-07-30 10:40:08 -07:00 |
|
multi_step_worker.py
|
[Speculative Decoding] EAGLE Implementation with Top-1 proposer (#6830)
|
2024-08-22 02:42:24 -07:00 |
|
ngram_worker.py
|
[BugFix] Fix use of per-request seed with pipeline parallel (#6698)
|
2024-07-30 10:40:08 -07:00 |
|
proposer_worker_base.py
|
[Speculative decoding] [Multi-Step] decouple should_modify_greedy_probs_inplace (#6971)
|
2024-08-09 05:42:45 +00:00 |
|
smaller_tp_proposer_worker.py
|
[Speculative decoding] [Multi-Step] decouple should_modify_greedy_probs_inplace (#6971)
|
2024-08-09 05:42:45 +00:00 |
|
spec_decode_worker.py
|
[Speculative Decoding] EAGLE Implementation with Top-1 proposer (#6830)
|
2024-08-22 02:42:24 -07:00 |
|
target_model_runner.py
|
[VLM] Refactor MultiModalConfig initialization and profiling (#7530)
|
2024-08-17 13:30:55 -07:00 |
|
top1_proposer.py
|
[Speculative Decoding] Fixing hidden states handling in batch expansion (#7508)
|
2024-08-19 17:58:14 -07:00 |
|
util.py
|
[Bugfix] spec decode handle None entries in topk args in create_sequence_group_output (#7232)
|
2024-08-22 09:33:48 -04:00 |
