Quantong Qiu 8ebe963269 Add dsv4_pro6000_sglang_tp_dp_matrix experiment for RTX 6000D ...

- New experiment directory with adaptive concurrency benchmark for SGLang
- Platform config for 8x NVIDIA RTX 6000D 84GB
- Adapted from dsv4_h200_sglang_tp_dp_matrix with updated paths:
  - Model path: /data/hf_models/DeepSeek-V4-Flash
  - Venv path: /root/.miniconda3/envs/sglang
  - Dataset path: /data/yy/sskj/datasets/

2026-07-13 02:18:34 +00:00

..

dsv4_h200_64k_sglang_vs_vllm

Add num_prompts = 5×concurrency convention and update all experiments

2026-07-09 09:26:41 +00:00

dsv4_h200_256k_4k_probe

Add num_prompts = 5×concurrency convention and update all experiments

2026-07-09 09:26:41 +00:00

dsv4_h200_dspark

feat: add H200 SGLang baseline experiment and fix health check

2026-07-08 06:56:45 +00:00

dsv4_h200_long_context_matrix

Add num_prompts = 5×concurrency convention and update all experiments

2026-07-09 09:26:41 +00:00

dsv4_h200_max_context_length

add dp_proxy for vLLM DP, write kv_cache_analysis.md, fix DP bench port

2026-07-09 10:19:30 +00:00

dsv4_h200_sglang

Add num_prompts = 5×concurrency convention and update all experiments

2026-07-09 09:26:41 +00:00

dsv4_h200_sglang_tp_dp_matrix

bench: add adaptive concurrency search

2026-07-11 06:11:40 +00:00

dsv4_h200_sglang_vs_vllm

Add num_prompts = 5×concurrency convention and update all experiments

2026-07-09 09:26:41 +00:00

dsv4_h200_vllm

Add num_prompts = 5×concurrency convention and update all experiments

2026-07-09 09:26:41 +00:00

dsv4_h200_vllm_dspark_vs_default

Add num_prompts = 5×concurrency convention and update all experiments

2026-07-09 09:26:41 +00:00

dsv4_h200_vllm_mtp_vs_default

Add num_prompts = 5×concurrency convention and update all experiments

2026-07-09 09:26:41 +00:00

dsv4_h200_vllm_tp2_custom_bench

Add custom benchmark setup for vLLM TP=8 on dsv4 H200

2026-07-10 09:40:10 +00:00

dsv4_h200_vllm_tp2_custom_bench_no_fp8

Add custom benchmark setup for vLLM TP=8 on dsv4 H200

2026-07-10 09:40:10 +00:00

dsv4_h200_vllm_tp4_custom_bench

Add custom benchmark setup for vLLM TP=8 on dsv4 H200

2026-07-10 09:40:10 +00:00

dsv4_h200_vllm_tp4_custom_bench_no_fp8

Add custom benchmark setup for vLLM TP=8 on dsv4 H200

2026-07-10 09:40:10 +00:00

dsv4_h200_vllm_tp8_custom_bench

Add custom benchmark setup for vLLM TP=8 on dsv4 H200

2026-07-10 09:40:10 +00:00

dsv4_h200_vllm_tp8_custom_bench_no_fp8

Add custom benchmark setup for vLLM TP=8 on dsv4 H200

2026-07-10 09:40:10 +00:00

dsv4_h200_vllm_tp_dp_matrix

bench: add adaptive concurrency search

2026-07-11 06:11:40 +00:00

dsv4_p800_long_context_matrix

Scale P800 long-context matrix to concurrency=1 for stability

2026-07-09 05:59:28 +00:00

dsv4_p800_max_context_length

results(p800): max context length exploration run 20260709-033134

2026-07-09 04:08:40 +00:00

dsv4_p800_sglang

Add num_prompts = 5×concurrency convention and update all experiments

2026-07-09 09:26:41 +00:00

dsv4_pro6000_sglang_tp_dp_matrix

Add dsv4_pro6000_sglang_tp_dp_matrix experiment for RTX 6000D

2026-07-13 02:18:34 +00:00

legacy_bench_results

refactor: move final legacy result dspark_grid into experiments/legacy_bench_results/

2026-07-08 06:40:39 +00:00

TEMPLATE

Add num_prompts = 5×concurrency convention and update all experiments

2026-07-09 09:26:41 +00:00

ADAPTIVE_CONCURRENCY_USAGE.md

bench: add adaptive concurrency search

2026-07-11 06:11:40 +00:00