sumadoratyper's log : 2023-11-16

  1. MSI G253PF (24.5インチワイド 液晶モニター) ドスパラ限定モデル ｜ パソコン通販のドスパラ【公式】
    24.5" FHD 380Hzが43k!? 安すぎだろ……

  2. MSI's 24.5-Inch G253PF Gaming Monitor Can Hit 380Hz | Tom's Hardware

  3. MSI Oculux NXG253R Review - RTINGS.com

    パネルの仕様は多分これとほぼ同じだと思うけど,

    The MSI Oculux NXG253R has an outstanding response time at the max refresh rate. At our recommended Response Time setting of 'Normal', the response time is a bit slow for a 360Hz monitor, but motion still looks incredible, and there's no overshoot. The 'Fast' setting has a much faster rise/fall time and achieves the 1ms response time advertised by MSI in most transitions, but there's noticeable overshoot. The 'Normal' mode has a bit more noticeable blur behind fast-moving objects. The 'Fast' mode has no noticeable blur, but there are noticeable overshoot artifacts behind some fast-moving objects. In the end, whichever one is better is up to you. Like most monitors, the 'Fastest' setting is practically unusable and provides no real benefit over 'Fast'.

    ということでオーバーシュートも"Overdrive Setting: Normal"なら問題ないみたい. Response Timeは多少落ちるけどそこは音ゲー用途的には別にいい.

  4. Dell Alienware AW2523HF Review - RTINGS.com
    一応今一度確認したけど10kほど価格差のあるAW2523HFもResponse Timeやコントラスト比は大差ない. ただこっちは"Overdrive Setting: Extreme"でもオーバーシュートしないし, Motion Blur Photoを見た感じ残像感は薄そう.

  5. いやまあG253PFでいいか. 金ないよ.

  6. Rapid IPSってAUO製なのか. G253PFにします.

  7. 安心ワイド保証を一年だけ付けておいた……

  8. Fast and Portable Llama2 Inference on the Heterogeneous Edge

  9. とりあえずCPU (7950X)で動かしてみた:

    % wasmedge --dir .:. --nn-preload default:GGML:AUTO:llama-2-7b-chat.Q5_K_M.gguf wasmedge-ggml-llama-interactive.wasm default

Question:
Please tell me some sightseeing spots in Tokyo
llama_new_context_with_model: n_ctx      = 1024
llama_new_context_with_model: freq_base  = 10000.0
llama_new_context_with_model: freq_scale = 1
llama_new_context_with_model: kv self size  =  512.00 MB
llama_new_context_with_model: compute buffer total size = 96.13 MB
Answer:
[2023-11-16 03:39:02.449] [info] [WASI-NN] GGML backend: llama_system_info: AVX = 1 | AVX2 = 1 | AVX512 = 0 | AVX512_VBMI = 0 | AVX512_VNNI = 0 | FMA = 1 | NEON = 0 | ARM_FMA = 0 | F16C = 1 | FP16_VA = 0 | WASM_SIMD = 0 | BLAS = 1 | SSE3 = 1 | SSSE3 = 1 | VSX = 0 |
Of course! Tokyo has a plethora of exciting sightseeing spots to explore. Here are some must-visit places:
1. Tokyo Skytree: At 634 meters tall, Tokyo Skytree offers breathtaking views of the city from its observation decks.
2. Shibuya Crossing: This famous intersection is a hub of activity, with neon lights, street performers, and some of the busiest streets in the world.
3. Tsukiji Fish Market: While the inner market has moved to a new location, the outer market still offers a fascinating glimpse into Tokyo's seafood culture, with fresh sushi and seafood available at every turn.
4. Meiji Shrine: Dedicated to the deified spirits of Emperor Meiji and his wife, Empress Shoken, this serene shrine is a peaceful oasis in the midst of the bustling city.
5. Tokyo National Museum: With a collection of over 110,000 objects, this museum offers a comprehensive look at Japanese art and history, from ancient times to the present day.
6. Asakusa: This historic district is home to the famous Senso-ji Temple, which offers a glimpse into old Japan, with its Nakamise shopping street and traditional architecture.
7. Roppongi Hills: This modern complex offers shopping, dining, and entertainment options, as well as an observation deck with panoramic views of the city.
8. Akihabara: Known as "Electric Town," Akihabara is the go-to destination for all things electronic, from anime and manga to video games and gadgets.
9. Harajuku: This fashionable district is known for its unique and eclectic styles, with shopping and dining options catering to all tastes.
10. Mount Fuji: While not technically in Tokyo, a day trip to the Fuji Five Lakes region offers stunning views of Japan's iconic mountain, as well as beautiful scenery and outdoor activities.
These are just a few of the many amazing sightseeing spots in Tokyo. Enjoy your trip!
llama_print_timings:        load time =   27987.98 ms
llama_print_timings:      sample time =       3.39 ms /   495 runs   (    0.01 ms per token, 145974.64 tokens per second)
llama_print_timings: prompt eval time =    3829.14 ms /    50 tokens (   76.58 ms per token,    13.06 tokens per second)
llama_print_timings:        eval time =   49234.48 ms /   494 runs   (   99.66 ms per token,    10.03 tokens per second)
llama_print_timings:       total time =   77238.24 ms

    LLM周辺全然追ってないからよくわからん. Rustで高速化がどうとか言われてるけどこれllama.cppより速いの？

  10. 聞いちゃいけないこと聞いてみるとI cannot provide advice or guidance on illegal or harmful activities から始まる決まり文句しか返してくれなくて何も面白くない. ローカルで頑張って脱獄する意味ないしそれ用のモデル使ってみるか……

  11. curl -LO 叩く度に「コミックLO!?」って脊髄反射してるのアレ.

  12. とりあえずWizard-Vicuna-7B-Uncensored使って遊んでみてるけど, こちらの欲望を尊重するような素振りは見せつつ精神科や更生機関へ行くよう執拗に勧めてくる. That's a great desire!じゃねんだわ.

  13. "what should i do?"じゃなくて"how can my dream is archived?"と尋ねるようにしたらそこそこ建設的な意見を述べてくれた. やるやん.

  14. 真面目に遊ぶならAMD GPUで動くようにしないといけないけど怠そうだなあ.

  15. How to use GPU with wasi-nn-pytorch? · Issue #2654 · WasmEdge/WasmEdge

  16. ROCm Port by SlyEcho · Pull Request #1087 · ggerganov/llama.cpp

  17. Install and run llama.cpp with ROCm 5.7 on Ubuntu 22.04 | by Housebyte Bestofnet | Nov, 2023 | Medium

  18. sudo がアレなので doas に乗り替えてみる #Linux - Qiita
    doas 知らんかった. 次にarchinstallbattleするときに乗り換えようかな.

  19. doas - ArchWiki

  20. SudoLoop doesn't work elegantly with doas · Issue #192 · Morganamilo/paru

  21. blockquoteとqの色は変えることにした. そっちの方が流し読みしやすい.

  22. abbr[title]{ text-decoration: underline dotted; } の指定はせずにUAに丸投げした方がいい. スマホ/タブレットで @title あるの分かっても読めねえんだって.

  23. 20:51, 呼気アセトン0.1mmol/L, 血中βOHB 0.3mmol/L. 最後に糖質食ってから丸二日くらい経過するのに全然上がってこねえ.