20231088/vllm

History

[Core] Deprecating block manager v1 and make block manager v2 default (#8704 )

Removing the block manager v1. This is the initial piece of prefix-caching-centric design. In order to achieve prefix-caching-centric design, we need to simplify the code path so that we only use v2 block manager (which has much higher performance on prefix caching).

2024-10-17 11:38:15 -05:00

cutlass_benchmarks

[Kernel] Add per-tensor and per-token AZP epilogues (#5941 )

2024-08-06 18:17:08 +00:00

kernels

[Misc] Standardize RoPE handling for Qwen2-VL (#9250 )

2024-10-16 13:56:17 +08:00

overheads

[Core] Deprecating block manager v1 and make block manager v2 default (#8704 )

2024-10-17 11:38:15 -05:00

backend_request_func.py

[core] remove beam search from the core (#9105 )

2024-10-07 05:47:04 +00:00

benchmark_latency.py

[Core] Deprecating block manager v1 and make block manager v2 default (#8704 )

2024-10-17 11:38:15 -05:00

benchmark_prefix_caching.py

[Core] Deprecating block manager v1 and make block manager v2 default (#8704 )

2024-10-17 11:38:15 -05:00

benchmark_prioritization.py

[core] remove beam search from the core (#9105 )

2024-10-07 05:47:04 +00:00

benchmark_serving.py

pass ignore_eos parameter to all benchmark_serving calls (#9349 )

2024-10-15 13:30:44 -07:00

benchmark_throughput.py

[Core] Deprecating block manager v1 and make block manager v2 default (#8704 )

2024-10-17 11:38:15 -05:00

launch_tgi_server.sh

[benchmark] Update TGI version (#7917 )

2024-08-27 15:07:53 -07:00

README.md

Change the name to vLLM (#150 )

2023-06-17 03:07:40 -07:00

sonnet.txt

feat(benchmarks): Add Prefix Caching Benchmark to Serving Benchmark (#3277 )

2024-03-27 13:39:26 -07:00

README.md

Benchmarking vLLM

Downloading the ShareGPT dataset

You can download the dataset by running:

wget https://huggingface.co/datasets/anon8231489123/ShareGPT_Vicuna_unfiltered/resolve/main/ShareGPT_V3_unfiltered_cleaned_split.json