Skip to content

DotBlossom/flask-AI-inference-bedrock-controller

Repository files navigation

descriptions

스크린샷 2024-12-14 083350

스크린샷 2024-12-14 083357 스크린샷 2024-12-14 083404 스크린샷 2024-12-14 083411 스크린샷 2024-12-14 083140

스크린샷 2024-12-23 131657 스크린샷 2024-12-23 131706

스크린샷 2024-12-23 131647

front fatching 전략

- bedrock custom-cart async result func : var = await func in succ
- product add : axios main async - next useRef useEffect
- make "body" object to convert any type of datastreams.
- function of propagation of actions : await syncer

P,N, A boundary Indexing

N = negative
Lx = Reg number of B
fn = features of categorize
B(u,l).fn = u,l boundary


N = T - (A - P) / Lx
B(u,l)f1 = [0.1 , Pi({x,n}, 1->k) x / Ly 

feature selector : Dominent[1] , Sub(W) = f(k-1)[0.01, 0.001 .. ]


B(u,l)f1 = [0.1 , Pi({x,n}, 1->k) x / Ly1 
B(u,l)f2 = [B(l)f1 , Pi({x,n}, 1->k) x / Ly2]
..
B(u,l)fn = [B(l)fn , Pi({x,n}, 1->k) x / Lyn]

S(B(u,l)f(k)) = key indexing Boundary values.

or Cx(B(u,l)f(k)) -> n filters. of parallel (Vector N d)

ver2.
multivar_ranker(operated by type of LLM res)
rf = dominent_selector + S{(1,argmax(k:3), w =: 0.001w} w(k) * sub_selector(k)}
re = rf * t(pre) , d = 16
v(re) = 16 innerproductor, if thrs >= k , merge types
v(re(n)) * v(re(u)) : pretrained input selector of thrs

migrate to fastAPI

https://github.com/DotBlossom/ai-pref-pipeline-fastAPI

propagations ctrl

- f_registor && usr_actions(n>=5, metadata * other usr)
- if usr action && scheduler 
- sheduler && is_var

gemini 활용법

- 직접 작성한 flask 코드를 기반으로, 동일한 양식을 프롬프트에 적절히 요청하여, 하나의 코드로직으로 여러 코드를 생성 후 바로 이식
> py 기반 code기 때문에, 생성 정확도가 매우 높다. (flask 채택 이유)
- flask --> fastAPI migrate의 경우, 완성된 flask code에, svc async 와 sync type을 구분 후, async : DB 의 type을 맞추고, 코드를 생성 후 바로 이식
> flask 기반 code기 때문에, 생성 정확도가 매우 높다 (flask 채택 이유)

About

two-tower-addon using bedrock LLM, inference Pipeline API Controller & scheduler, one shot dataset

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages