legend-of-elya-n64

(★ 130)

Legend of Elya — N64 game with a real 6.36M-parameter ternary transformer on the VR4300 MIPS III CPU. Zelda-style dungeon, AI NPCs, byte-level inference at 1.23 tok/s scalar / 2.19 tok/s on the RSP overlay (measured under ares, never on silicon). Built with libdragon.

파일 탐색기

ZIP 다운로드

- - n64_llm_bridge.py
- - n64_llm_rpc.c
  - n64_llm_rpc.h
  - pak_io.c
  - pak_io.h
- - main.cpp
- build.sh
- cobalt_qube3.jpg
- firstboot.sh
- libdragon.h
- nano_gpt.c
- nano_gpt.h
- oracle_k6_run.png
- oracle_main.c
- qube-setup.conf
- README.md
- RESTORE.md
- rsp_matmul.h
- stdint.h
- personas.yaml
- training_data.json
- - consensus_tern8_int84_s1337.json
  - consensus_tern8_int84_s1337.log
  - consensus_tern8_int88_s1337.json
  - consensus_tern8_int88_s1337.log
  - dual_eval_s1337.json
  - dual_eval_s1337.log
  - dual_run1.log
  - dual_run_serial_vs_overlap.log
  - dual_run_shift0.log
  - dual_run_tern8_tern4.log
  - dual_run_terntern.log
  - dual_run_verify_shift0.log
  - host_verify_s1337_shift0.log
  - host_verify_tern8tern4_shift0.log
  - host_verify_terntern_shift0.log
  - regress_ovl_after_flush.log
  - regress_ovl_noflush.log
  - regress_ovl_split_driver.log
  - sweep.log
  - sweep_A_int84_s101.json
  - sweep_A_int84_s101.log
  - sweep_A_int84_s1337.json
  - sweep_A_int84_s1337.log
  - sweep_A_int84_s23.json
  - sweep_A_int84_s23.log
  - sweep_A_int84_s7.json
  - sweep_A_int84_s7.log
  - sweep_B_int88_s101.json
  - sweep_B_int88_s101.log
  - sweep_B_int88_s1337.json
  - sweep_B_int88_s1337.log
  - sweep_B_int88_s23.json
  - sweep_B_int88_s23.log
  - sweep_B_int88_s7.json
  - sweep_B_int88_s7.log
  - sweep_C_cross_101_1337.json
  - sweep_C_cross_101_1337.log
  - sweep_C_cross_1337_7.json
  - sweep_C_cross_1337_7.log
  - sweep_C_cross_23_101.json
  - sweep_C_cross_23_101.log
  - sweep_C_cross_7_23.json
  - sweep_C_cross_7_23.log
  - sweep_D_terntern_101_1337.json
  - sweep_D_terntern_101_1337.log
  - sweep_D_terntern_1337_7.json
  - sweep_D_terntern_1337_7.log
  - sweep_D_terntern_23_101.json
  - sweep_D_terntern_23_101.log
  - sweep_D_terntern_7_23.json
  - sweep_D_terntern_7_23.log
  - sweep_E_tern4_s1337.json
- - rsp_overlay_headless.log
  - rsp_overlay_ingame.log
  - scalar_headless.log
  - scalar_ingame.log
  - vi_rate_1s.log
  - vi_rate_20s.log
  - vi_rate_28s_wrapped.log
  - vi_rate_90s.log
- EMULATOR_ACCURACY.md
- N64_ARES_FINDINGS.md
- N64_COHERENCE_FINDINGS.md
- N64_DUAL_FINDINGS.md
- N64_MEASUREMENT_FINDINGS.md
- N64_QUANTIZATION_FINDINGS.md
- N64_RATE_FINDINGS.md
- N64_RETRAIN_FINDINGS.md
- N64_RSP_FINDINGS.md
- N64_SAMPLING_FINDINGS.md
- PERSONALITY_PACK_BOUNTY.md
- STREAMING_MOE.md
- persona_eval_20260329_0143.md
- .gitkeep
- sophia_weights.bin
- sophia_int8.bin
- sophia_weights.bin
- sophia_int8.bin
- sophia_weights.bin
- sophia_int8.bin
- sophia_weights.bin
- - n64_bridge.py
- - n64_attest.c
  - n64_attest.h
- - main.cpp
- README.md
- rsp_budget_probe.S
- trace_temp64.txt
- - rpi_format.h
  - rpi_n64.h
  - rpi_runtime.h
- - rpi_game_data.json
  - sophia_game_n64.rpi
  - vocab_map.json
  - vocab_strings.json
- - rpi_decode.c
  - rpi_model.c
  - rpi_n64.c
- n64_llm_ibm_power8.png
- n64_llm_nodes_validate.png
- n64_llm_zelda_triforce.png
- rspq_overlay_dungeon.png
- rspq_overlay_npc_answer.png
- rspq_overlay_renders_after_generation.png
- rspq_overlay_title.png
- standalone_rsp_crash.png
- ares_rate_run.py
- blob_layout.py
- evaluate_personas.py
- evaluate_personas_simple.py
- export_weights.py
- generate_weights.py
- generate_weights_simple.py
- train_personas.py
- expert_cache.c
- expert_cache.h
- expert_cache_regress.c
- expert_cache_test.c
- legend_of_elya.c
- legend_of_elya_mining.c
- nano_gpt.c
- nano_gpt.h
- - libdragon.h
- ce_probe.c
- game_sim.c
- host_eval.c
- kernel_test.c
- mag_probe.c
- pse_trace.c
- qat_finetune.py
- quantize_n64.py
- quantize_ternary_n64.py
- rdram_budget.py
- reset_trace.c
- samp_probe.c
- tf_probe.c
- compare_rom_host.py
- consensus.py
- dual_eval.py
- dual_sweep.sh
- dual_train.sh
- eval12.py
- eval_qat.py
- ptq_from_ckpt.py
- qat_npz_to_seq.py
- README.md
- report.py
- report_sweep.py
- run_eval_all.py
- score_rom.py
- test_microbatch_equiv.py
- train_mix.py
- train_sophia_v7.py
- train_sophia_v9_qat.py
- blacksmith.bin
- guard.bin
- librarian.bin
- sophia.bin
- sophia_weights.bin
- sophia_weights_large_v5fmt.bin
.gitignore
BCOS.md
COMPLETION_REPORT.md
CONTRIBUTING.md
dual_probe.c
FINDINGS.md
gen_sophia_host.c
github_issue_claim.md
github_pr_template.md
legend_of_elya.c
legend_of_elya.z64
legend_of_elya_mining.c
legend_of_elya_rpc.z64
LICENSE
Makefile
matmul_rsp2.c
matmul_rsp_drv.c
multi_npc.c
multi_npc.h
nano_gpt.c
nano_gpt.h
nano_gpt_host.c
PRESS_KIT.md
README.md
reference_cli.c
rsp_matmul.h
rsp_matmul.S
rsp_mm2.S
rsp_mm2_ovl.S
screenshot.png
test_multi_npc.py
train_sophia.py
train_sophia_v5.py
train_sophia_v7.py
train_sophia_v8.py

# CDN으로 사용하기

jsDelivr

jsDelivr는 공개 GitHub 리포지토리를 별도 설정 없이 CDN으로 즉시 서빙합니다. 버전과 파일을 고르면 웹페이지에 바로 붙일 수 있는 링크와 예시 코드가 만들어집니다.

버전

파일

링크

예시

명령어 용어집

이 문서에서 사용된 명령어를 모아봤습니다. 낯선 명령어가 있다면 펼쳐서 확인해보세요.

🔍

`python3`

설명 보기 ▼

이 명령은 `python`의 별칭입니다.

tldr python

자세한 내용은 원본 명령을 참고하세요: