From 8f12069337fc3b2b0a00ed858d9540203679f4ef Mon Sep 17 00:00:00 2001 From: moosecrap Date: Fri, 17 Jul 2026 18:45:14 -0700 Subject: [PATCH] Script update, restore model list --- bench-llms.sh | 3 ++- models.txt | 31 ++++++++++++++++++++++++------- readme.txt | 1 + 3 files changed, 27 insertions(+), 8 deletions(-) diff --git a/bench-llms.sh b/bench-llms.sh index f1b89a4..8885100 100755 --- a/bench-llms.sh +++ b/bench-llms.sh @@ -4,6 +4,7 @@ PREFILL_REQ='{model: $model, chat_template_kwargs: {"enable_thinking": false}, m ENDPOINT='http://localhost:8080/v1/chat/completions' WORDLIST='/etc/dictionaries-common/words' +WORDLIST_GREP='^[A-Z][a-z]+$' # Proper nouns only get_response() { model=$1 @@ -25,7 +26,7 @@ get_response() { } sort --random-sort models.txt | while read model; do - random_name=$(grep -Po '^[A-Z][a-z]+$' $WORDLIST | sort --random-sort | head -n1) + random_name=$(grep -Po $WORDLIST_GREP $WORDLIST | sort --random-sort | head -n1) random_name=${random_name,} echo $random_name : $model >> key.txt for prompt_file in prompts/*.txt; do diff --git a/models.txt b/models.txt index 191471b..824eb50 100644 --- a/models.txt +++ b/models.txt @@ -1,8 +1,25 @@ -NVIDIA-Nemotron-3-Super-120B-A12B-MXFP4_MOE -gemma-4-31B-it-BF16 -GLM-4.5-Air-UD-Q5_K_XL -c4ai-command-r-plus-08-2024-Q6_K -Devstral-2-123B-Instruct-2512-UD-Q5_K_XL -mistral-nemo_12b-instruct-2407-fp16 -Ministral-3-14B-Instruct-2512-BF16 Mistral-Large-Instruct-2411-Q5_K_M +GLM-4.5-Air-UD-Q5_K_XL +NVIDIA-Nemotron-3-Super-120B-A12B-MXFP4_MOE +c4ai-command-r-plus-08-2024-Q6_K +gemma-4-31B-it-BF16 +Devstral-2-123B-Instruct-2512-UD-Q5_K_XL +Qwen3.6-27B-BF16 +Qwen3.5-122B-A10B-MXFP4_MOE +Olmo-3.1-32B-Instruct-BF16 +granite-4.0-h-small-BF16 +Mistral-Small-3.2-24B-Instruct-2506-BF16 +Llama-4-Scout-17B-16E-Instruct-UD-Q5_K_XL +Qwen3-32B-BF16 +google_gemma-3-27b-it-bf16 +miqu-1-70b.q5_K_M +Ministral-3-14B-Instruct-2512-BF16 +c4ai-command-a-03-2025-Q5_K_M +EXAONE-4.0-32B-BF16 +mistral-nemo_12b-instruct-2407-fp16 +c4ai-command-r-08-2024-f16 +phi-4-bf16 +Mistral-Small-4-119B-2603-MXFP4_MOE +Llama-3.3-70B-Instruct-UD-Q8_K_XL +gpt-oss-120b-mxfp4 +GLM-4.7-Flash-BF16 diff --git a/readme.txt b/readme.txt index a06a03a..d252495 100644 --- a/readme.txt +++ b/readme.txt @@ -4,4 +4,5 @@ Requirements: * curl * jq +Run llama.cpp server with a command line like `./llama-server --models-preset ~/models/preset.ini --models-dir ~/models/ --models-max 1` \ No newline at end of file