TScale
No description available.
파일 탐색기
최종 버전 다운로드 (.zip)- data_basic.cfg
- data_fineweb.cfg
- data_owt.cfg
- fed_fineweb.cfg
- index_fineweb.cfg
- train_basic.cfg
- train_enwik8.cfg
- train_fineweb.cfg
- train_owt125M.cfg
- train_perf_test.cfg
- att.cpp
- att.h
- nodes_batch.cpp
- nodes_batch.h
- rope.cpp
- rope.h
- sliding_window.cpp
- sliding_window.h
- stdafx.cpp
- stdafx.h
- z.cfg
- cfg_precision.h
- gpt_att.cu
- gpt_att_fp16.cuh
- gpt_att_fp8.cuh
- gpt_combiner.cuh
- gpt_cpu.cpp
- gpt_cpu.h
- gpt_cuda.cu
- gpt_cuda.cuh
- gpt_embedding.cuh
- gpt_final.cuh
- gpt_layer_norm.cuh
- gpt_rope.cuh
- matmul_fwdbwd.cu
- matmul_fwdbwd.cuh
- model.cpp
- model.h
- par_delta.cpp
- par_delta.h
- par_matrix.cpp
- par_matrix.h
- par_matrix_cuda.cu
- par_matrix_cuda.cuh
- row_scale.cu
- row_scale.cuh
- row_tile_scale.cuh
- stdafx.cpp
- stdafx.h
- z.cfg
- bpe.cpp
- bpe.h
- data.cpp
- data.h
- dataset.cpp
- dataset.h
- dataset_builder.cpp
- dataset_builder.h
- fragment_gen.cpp
- fragment_gen.h
- net_data.cpp
- net_data.h
- ppm_lmatch.cpp
- ppm_lmatch.h
- ppm_window.cpp
- ppm_window.h
- stdafx.cpp
- stdafx.h
- text_saveload.cpp
- text_saveload.h
- z.cfg
- data_config.cpp
- data_config.h
- stdafx.cpp
- stdafx.h
- z.cfg
- data_server.cpp
- stdafx.cpp
- stdafx.h
- z.cfg
- fed_center.cpp
- stdafx.cpp
- stdafx.h
- z.cfg
- fed_lib.cpp
- fed_lib.h
- stdafx.cpp
- stdafx.h
- z.cfg
- fed_worker.cpp
- stdafx.cpp
- stdafx.h
- z.cfg
- infer.cpp
- sample_model.cpp
- sample_model.h
- stdafx.cpp
- stdafx.h
- z.cfg
- mlm_data_server.cpp
- stdafx.cpp
- stdafx.h
- z.cfg
- fed_model.cpp
- fed_model.h
- model_dim.cpp
- model_dim.h
- model_matrix.cpp
- model_matrix.h
- model_params.cpp
- model_params.h
- sse_utils.cpp
- sse_utils.h
- stdafx.cpp
- stdafx.h
- z.cfg
- net_test.cpp
- stdafx.cpp
- stdafx.h
- z.cfg
- stdafx.cpp
- stdafx.h
- xrng.cpp
- xrng.h
- z.cfg
- stdafx.cpp
- stdafx.h
- tokenizer.cpp
- z.cfg
- cpu_infer.cpp
- cpu_infer.h
- fed_sim.cpp
- fed_sim.h
- main_gpt.cpp
- mmlu_score.cpp
- mmlu_score.h
- net_train.cpp
- net_train.h
- network.cpp
- network.h
- stdafx.cpp
- stdafx.h
- z.cfg
- stdafx.cpp
- stdafx.h
- train_config.cpp
- train_config.h
- train_step.cpp
- train_step.h
- z.cfg
- backprop.cpp
- backprop.h
- batch_config.cpp
- batch_config.h
- stdafx.cpp
- stdafx.h
- train_ctx.cpp
- train_ctx.h
- z.cfg
- cfg_file.cpp
- cfg_file.h
- config.cpp
- config.h
- stdafx.cpp
- stdafx.h
- z.cfg
- cuda_arrays.cpp
- cuda_arrays.h
- cuda_fp16.cu
- cuda_fp16.cuh
- cuda_fp8.cu
- cuda_fp8.cuh
- cuda_graph.cu
- cuda_graph.cuh
- cuda_i8.cu
- cuda_i8.cuh
- cuda_init.cpp
- cuda_init.h
- cuda_matmul.cu
- cuda_matmul.cuh
- cuda_memory.cpp
- cuda_memory.h
- cuda_mma.cu
- cuda_mma.cuh
- cuda_sort.cu
- cuda_sort.cuh
- cuda_util.cu
- cuda_util.cuh
- stdafx.cpp
- stdafx.h
- vec_util.cu
- vec_util.cuh
- z.cfg
- doc_info.cpp
- doc_info.h
- make_bin_features.cpp
- make_bin_features.h
- stdafx.cpp
- stdafx.h
- z.cfg
- dir.cpp
- dir.h
- fmt_reader.cpp
- fmt_reader.h
- stdafx.cpp
- stdafx.h
- z.cfg
- citymurmur.cpp
- citymurmur.h
- guid.cpp
- guid.h
- stdafx.cpp
- stdafx.h
- z.cfg
- hp_timer.cpp
- hp_timer.h
- stdafx.cpp
- stdafx.h
- z.cfg
- json.cpp
- json.h
- stdafx.cpp
- stdafx.h
- z.cfg
- log.cpp
- log.h
- stdafx.cpp
- stdafx.h
- z.cfg
- eigen.cpp
- eigen.h
- matrix.cpp
- matrix.h
- matrix_utils.cpp
- matrix_utils.h
- stdafx.cpp
- stdafx.h
- z.cfg
- html_compose.cpp
- html_compose.h
- http_client.cpp
- http_client.h
- http_header.cpp
- http_header.h
- http_request.cpp
- http_request.h
- http_server.cpp
- http_server.h
- ip_address.cpp
- ip_address.h
- net_init.cpp
- net_util.cpp
- net_util.h
- poller.cpp
- poller.h
- self_connection.cpp
- self_connection.h
- stdafx.cpp
- stdafx.h
- tcp_cmds.cpp
- tcp_cmds.h
- tcp_net.cpp
- tcp_net.h
- z.cfg
- mersenne.h
- poisson.cpp
- poisson.h
- rand_utils.cpp
- rand_utils.h
- random250.cpp
- random250.h
- stdafx.cpp
- stdafx.h
- z.cfg
- 2Darray.h
- atomic.h
- bin_saver.h
- eden_core.h
- event.cpp
- event.h
- fast_io.cpp
- fast_io.h
- fp8.cpp
- fp8.h
- hu_alloc.cpp
- hu_alloc.h
- mem_io.cpp
- mem_io.h
- nalgobase.h
- nhash_fun.h
- nhash_map.h
- nhash_table.h
- nlist.h
- npair.h
- nstring.h
- nuninitialized.h
- nvector.h
- radix_sort.cpp
- radix_sort.h
- stdafx.cpp
- stdafx.h
- string.cpp
- string.h
- thread.cpp
- thread.h
- tools.cpp
- tools.h
- ysafeptr.cpp
- ysafeptr.h
- z.cfg
- 1.5B_model.md
- 125M_model.md
- 1T_model.md
- data_script.md
- fed.md
- fo.md
- lm_search.md
- model.md
- precision.md
- tokenizer.md
- train_script.md
- fo.cpp
- fed_hellaswag.png
- fed_iteration.png
- fed_loss.png
- tl125M.png
- tl15b.png
- tl1T.png
- 1.5B_fineweb.log
- 125m_1T_fineweb.log
- 125m_fineweb.log
- 125m_owt.log
- fed.log
- hell_import.py
- import.py
- ds.py
- .gitignore
- Dockerfile
- LICENSE
- README.md
- test.cfg
# 설치 가이드
1. 코드 내려받기
git clone https://github.com/Foreseerr/TScale
깃허브에서 프로젝트 코드 전체를 내 컴퓨터로 내려받습니다.
cd TScale
방금 내려받은 프로젝트 폴더 안으로 이동합니다.
2. Docker
쉬움 추천사전 준비물
- Git GitHub에서 프로젝트 코드를 내려받으려면 필요합니다.
- Docker Desktop 컨테이너를 빌드하고 실행하려면 필요합니다. 설치 후 실행해서 백그라운드에 켜두세요.
docker build -t tscale .
Dockerfile을 기반으로 실행 가능한 이미지를 빌드합니다.
docker run -p 8080:80 tscale
빌드된 이미지를 실제 컨테이너로 실행합니다.
터미널에 docker compose ps 를 입력해 컨테이너들이 Up 상태인지 확인하세요. README에 포트 번호가 적혀있다면 브라우저에서 http://localhost:포트번호 로 접속해보세요.
// repository documentation
Was this content helpful?
(0 ratings)