-
Notifications
You must be signed in to change notification settings - Fork 1
/
01-answer.sh
executable file
·41 lines (41 loc) · 8.91 KB
/
01-answer.sh
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
# time python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path /models/llm/shisa/ft-ablation/shisa-bad-7b-v1 --model-id shisa-bad-7b-v1 --num-gpus-total 2
# time python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path /models/llm/shisa/ft-ablation/shisa-gamma-7b-v1 --model-id shisa-gamma-7b-v1 --num-gpus-total 2
# time python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path /models/llm/shisa/zero-extra/shisa-mega-dpo-7b-v1.1 --model-id shisa-mega-dpo-7b-v1.1 --num-gpus-total 2
# time python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path /models/llm/shisa/zero-extra/shisa-mega-7b-v1.2 --model-id shisa-mega-7b-v1.2 --num-gpus-total 2
# time python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path /models/llm/shisa/augmxnt_shisa-mega-7b-v1.2-dpo --model-id shisa-mega-7b-v1.2-dpo --num-gpus-total 2
# time python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path /workspace/llm-jp-13b-instruct-full-jaster-v1.0 --model-id llm-jp-13b-instruct-full-jaster-v1.0
# time python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path /workspace/llm-jp_llm-jp-13b-instruct-full-jaster-dolly-oasst-v1.0 --model-id llm-jp-13b-instruct-full-jaster-dolly-oasst-v1.0
# time python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path /models/llm/hf/moneyforward_houou-instruction-7b-v1 --model-id houou-instruction-7b-v1 --num-gpus-total 2
# time python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path /models/llm/hf/moneyforward_houou-instruction-7b-v1 --model-id houou-instruction-7b-v1-correctedtemplate --num-gpus-total 2
# time python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path /models/llm/hf/moneyforward_houou-instruction-7b-v1 --model-id houou-instruction-7b-v1-correctedtemplate2 --num-gpus-total 2
# time python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path /models/llm/hf/tokyotech-llm_Swallow-7b-instruct-hf --model-id tokyotech-llm_Swallow-7b-instruct-hf --num-gpus-total 2 --num-gpus-per-model 1
# time python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path /models/shisa-7b-v1 --model-id shisa-7b-v1-fullprompt --num-gpus-total 2 --num-gpus-per-model 1
# time python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path tokyotech-llm/Swallow-13b-instruct-hf --model-id tokyotech-llm_Swallow-13b-instruct-hf --num-gpus-total 2 --num-gpus-per-model 2
# time python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path TheBloke/Swallow-70B-instruct-AWQ --model-id Swallow-70b-instruct-AWQ --num-gpus-total 2 --num-gpus-per-model 2
# time python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path TheBloke/Swallow-70B-instruct-GPTQ --model-id Swallow-70b-instruct-GPTQ --num-gpus-total 2 --num-gpus-per-model 2
# time CUDA_VISIBLE_DEVICES=0 python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path TheBloke/shisa-7B-v1-GPTQ --model-id shisa-7B-v1-GPTQ --num-gpus-total 2 --num-gpus-per-model 1
# time python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path TheBloke/shisa-7B-v1-AWQ --model-id shisa-7B-v1-AWQ --num-gpus-total 2 --num-gpus-per-model 2
# time python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path /models/llm/gptq/TheBloke_Swallow-70B-instruct-GPTQ --model-id Swallow-70b-instruct-GPTQ --num-gpus-total 2 --num-gpus-per-model 2
# time python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path /models/llm/gptq/TheBloke_Xwin-LM-70B-V0.1-GPTQ --model-id Xwin-LM-70B-V0.1-GPTQ --num-gpus-total 2 --num-gpus-per-model 2
# time python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path Qwen/Qwen-72b-Chat --model-id Qwen-72b-Chat --num-gpus-total 4 --num-gpus-per-model 4
# time python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path /workspace/rinna_nekomata-14b-instruction --model-id nekomata-14b-instruction-correctedprompt --num-gpus-total 1 --num-gpus-per-model 1
# time python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path /workspace/rinna_nekomata-14b-instruction --model-id nekomata-14b-instruction-correctedprompt-hf --num-gpus-total 1 --num-gpus-per-model 1
# time CUDA_VISIBLE_DEVICES=1 python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path /models/chatntq-nekomata-14b --model-id chatntq-nekomata-14b --num-gpus-total 1 --num-gpus-per-model 1
# time CUDA_VISIBLE_DEVICES=1 python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path /models/chatntq-qwen-14b-chat --model-id chatntq-qwen-14b-chat --num-gpus-total 1 --num-gpus-per-model 1
# time CUDA_VISIBLE_DEVICES=1 python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path /models/chatntq-elyza-13b-fast --model-id chatntq-elyza-13b-fast --num-gpus-total 1 --num-gpus-per-model 1
# time CUDA_VISIBLE_DEVICES=1 python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path /models/chatntq-qwen-14b-chat --model-id chatntq-qwen-14b-chat --num-gpus-total 1 --num-gpus-per-model 1
# time CUDA_VISIBLE_DEVICES=0 python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path /models/chatntq-orion-14b-chat --model-id chatntq-orion-14b-chat --num-gpus-total 1 --num-gpus-per-model 1
# time CUDA_VISIBLE_DEVICES=0 python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path OrionStarAI/Orion-14B-Chat --model-id orionstarai-orion-14b-chat --num-gpus-total 1 --num-gpus-per-model 1
# time CUDA_VISIBLE_DEVICES=0 python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path /models/shisa-base-7b-v1_sharegpt-clean-ja --model-id qlora_shisa-base-7b-v1_sharegpt-clean-ja --num-gpus-total 1 --num-gpus-per-model 1
# time CUDA_VISIBLE_DEVICES=1 python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path /models/shisa-7b-v1_sharegpt-clean-ja --model-id qlora_shisa-7b-v1_sharegpt-clean-ja --num-gpus-total 1 --num-gpus-per-model 1
# time python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path /AKA/ai/llm/gguf/miqudev_miqu-1-70b/miqu-1-70b.q4_k_m.gguf --model-id miqu-1-70b --num-gpus-total 2 --num-gpus-per-model 2
# time python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path internlm/internlm2-chat-20b --model-id internlm2-chat-20b --num-gpus-total 2 --num-gpus-per-model 2
# time python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path Qwen/Qwen1.5-14b-Chat --model-id Qwen1.5-14b-Chat --num-gpus-total 2 --num-gpus-per-model 2
# time python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path /models/shisa-12b-v1-sft.gguf --model-id shisa-12b-v1-sft --num-gpus-total 2 --num-gpus-per-model 2
# time python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path /models/llm/hf/01-ai_Yi-34B-Chat-8bits --model-id Yi-34B-Chat-8bits --num-gpus-total 2 --num-gpus-per-model 2
# time python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path /models/llm/gguf/yi-34b.Q8_0.gguf --model-id Yi-34B-Q8_0.gguf --num-gpus-total 2 --num-gpus-per-model 2
# time python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path /models/llm/hf/01-ai_Yi-34B-Chat --model-id Yi-34B-Chat --num-gpus-total 2 --num-gpus-per-model 2
# time python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path /models/llm/hf/01-ai_Yi-9B --model-id Yi-9B --num-gpus-total 2 --num-gpus-per-model 2
# time python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path /models/llm/hf/abacusai_bigyi-15b --model-id bigyi-15b --num-gpus-total 2 --num-gpus-per-model 2
# time python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path /models/llm/hf/abacusai_bigstral-12b-32k --model-id bigstral-12b-32k --num-gpus-total 2 --num-gpus-per-model 2
time python gen_model_answer.py --bench-name japanese_mt_bench --num-choices 4 --max-turns 1 --model-path DataPilot/ArrowPro-7B-KUJIRA --model-id ArrowPro-7B-KUJIRA --num-gpus-total 2 --num-gpus-per-model 1