Kuvhura Inotsanangurwa AI mu Enterprise Architecture neLangChain uye DSPy

Nhanganyaya

Inotsanangurwa AI (XAI) yave chinhu chakakosha chezvivakwa zvemazuva ano zvebhizinesi, zvichiita kuti masangano avimbe uye anzwisise sarudzo dzakaitwa neAI-inotyairwa masisitimu. Nekudaro, kuita XAI mune yakaoma software masisitimu inogona kuve ibasa rinotyisa, kunyanya kana uchibata neakakura-mabhizinesi ekuvaka. Muchikamu chino, tichaongorora kuti LangChain neDSPy zvingashandiswa sei kusunungura simba reXAI mukugadzira bhizinesi, tichitarisa pakuita kweAI-inotungamirirwa neAI.

Chii Chinotsanangurwa AI (XAI)?

Inotsanangurwa AI inoreva kugona kweAI masisitimu kupa pachena uye kududzira tsananguro dzesarudzo dzavo uye kufanotaura. Izvi zvinonyanya kukosha munzvimbo dzepamusoro-soro dzakadai sehutano, mari, uye zvekufambisa, uko sarudzo dzinotungamirirwa neAI dzinogona kuva nemigumisiro inokosha.

LangChain: Chimiro Chekuvaka Inotsanangurwa AI Systems

LangChain inzvimbo yakavhurika-sosi yekuvaka inotsanangurwa AI masisitimu. Inopa seti yezvishandiso uye maraibhurari ekugadzira mamodheru anogona kupa zvakajeka uye zvinodudzirwa tsananguro dzezvisarudzo zvavo. LangChain yakavakirwa pamusoro peanozivikanwa akadzama masisitimu ekudzidza akadai seTensorFlow nePyTorch, zvichiita kuti zvive nyore kubatanidza nemasisitimu eAI aripo.

DSPy: Raibhurari Yekuvaka Inotsanangurwa AI Models

DSPy raibhurari yekuvaka inotsanangurwa AI modhi uchishandisa iyo PyTorch chimiro. Inopa seti yeakafanovakwa modhi uye maturusi ekugadzira anotsanangurika AI masisitimu, anosanganisira mamodheru ekuiswa kwemifananidzo, mitauro yechisikigo kugadzirisa, nezvimwe.

Kuita Inotsanangurwa AI muBhizinesi Architecture neLangChain uye DSPy

Kuita XAI mukuvakwa kwebhizinesi kunoda kunzwisisa kwakadzama kweiyo iri pasi peAI masisitimu uye kugona kubatanidza zvikamu zveXAI mumasisitimu aripo. Pano, tichaongorora muenzaniso wakapfava wekuita XAI mune yakaoma software system uchishandisa LangChain uye DSPy.

Muenzaniso: Kuvaka Inotsanangurika AI Modhi yeKuronga Kwemifananidzo

Mumuenzaniso uyu, tichavaka inotsanangurwa AI modhi yekuisa mufananidzo uchishandisa LangChain uye DSPy. Isu tichashandisa yakapusa convolutional neural network (CNN) dhizaini uye tinobatanidza XAI zvikamu tichishandisa LangChain.

import torch
import torch.nn as nn
import langchain
kubva ku langchain import LLAMA

# Tsanangura iyo CNN dhizaini
kirasi CNN(nn.Module):
    def __init__(iwe):
        super(CNN, self).__init__()
        self.conv1 = nn.Conv2d(1, 10, kernel_size=5)
        self.conv2 = nn.Conv2d(10, 20, kernel_size=5)
        self.fc1 = nn.Linear(320, 50)
        self.fc2 = nn.Linear(50, 10)

    def kumberi( pachedu, x):
        x = torch.relu(self.conv1(x))
        x = torch.relu(self.conv2(x))
        x = x.view(-1, 320)
        x = torch.relu(self.fc1(x))
        x = self.fc2(x)
        kudzoka x

# Tanga iyo CNN modhi
modhi = CNN ()

# Tanga iyo LangChain LLAMA modhi
llama = LLAMA(modhi)

# Tsanangura iyo XAI chikamu
kirasi XAIComponent:
    def __init__(pachako, modhi):
        self.model = muenzaniso

    def tsanangura (pachako, kuisa_data):
        # Shandisa LangChain kugadzira tsananguro yedata rekuisa
        tsananguro = llama.generate_explanation(input_data)
        dzorera tsananguro

# Tanga iyo XAI chikamu
xai_component = XAIComponent(modhi)

# Edza iyo XAI chikamu
input_data = torch.randn(1, 1, 28, 28)
tsananguro = xai_component.explain(input_data)
dhinda (tsanangudzo)

Mumuenzaniso uyu, tinotsanangura yakapfava CNN architecture tichishandisa PyTorch uye tinobatanidza XAI zvikamu uchishandisa LangChain. Isu tinozotsanangura chikamu cheXAI chinoshandisa LangChain kugadzira tsananguro yedata rekuisa.

Yepamusoro 5-10 Yakanyanya Kushandiswa Mirairo / Zvimedu

Heano epamusoro 5-10 anonyanya kushandiswa mirairo/zvimedu zveLangChain uye DSPy:

LangChain

langchain.LLAMA(model): Inotangisa LangChain LLAMA modhi.
llama.generate_explanation(input_data): Inoburitsa tsananguro yedata rekuisa.
langchain.Explanation(model, input_data): Inogadzira tsananguro yedata rekuisa.
llama.tsanangura(input_data): Inotsanangura data rekuisa uchishandisa modhi yeLLAMA.
langchain.ModelWrapper(model): Inopeta modhi nekapeta yemhando yeLangChain.

DSPy

dsp.DSPy(model): Inotangisa modhi yeDSPy.
dsp.tsanangura(input_data): Inotsanangura data rekuisa uchishandisa DSPy modhi.
dsp.ModelWrapper(modhi): Inopeta modhi nekapeta yemuenzaniso weDSPy.
dsp.Explanation(model, input_data): Inogadzira tsananguro yedata rekuisa.
dsp.generate_explanation(input_data): Inoburitsa tsananguro yedata rekuisa.

Kuona

Heino kuonesa kuri nyore kwechikamu cheXAI uchishandisa Chart.js:

{
  "type": "bha",
  "data": {
    "labels": ["Input Data", "Tsanangudzo"],
    "datasets": [
      {
        "label": "XAI Component",
        "data": [10, 20],
        "backgroundColor": [
          "rgba(255, 99, 132, 0.2)",
          "rgba(54, 162, 235, 0.2)"
        ],
        "borderColor": [
          "rgba(255, 99, 132, 1)",
          "rgba(54, 162, 235, 1)"
        ],
        "borderWidth": 1
      }
    ]
  },
  "sarudzo": {
    "zvikero": {
      "yAxes": [
        {
          "zvikwekwe": {
            "beginAtZero": chokwadi
          }
        }
      ]
    }
  }
}

Kuona uku kunoratidza data rekuisa uye tsananguro inogadzirwa nechikamu cheXAI.

Mhedziso

Kuita XAI mukuvakwa kwebhizinesi kunoda kunzwisisa kwakadzama kweiyo iri pasi peAI masisitimu uye kugona kubatanidza zvikamu zveXAI mumasisitimu aripo. LangChain uye DSPy inopa seti yezvishandiso uye maraibhurari ekuvaka anotsanangurika AI masisitimu, zvichiita kuti zvive nyore kuita XAI mune yakaoma software masisitimu. Nekutevera muenzaniso uye kuona kwakapihwa muchinyorwa chino, vagadziri vanogona kuvhura simba reXAI mune yavo bhizinesi rekuvaka uye kuvaka akajeka uye anodudzira AI-inotyairwa masisitimu.

Basa Ramangwana

Basa remangwana panyaya iyi rinogona kusanganisira:

Kugadzira zvimwe zvepamberi XAI zvikamu uchishandisa LangChain uye DSPy.
Kubatanidza zvinhu zveXAI nemamwe maAI masimusi uye maraibhurari.
Kugadzira mamwe maturusi ekuona maturusi eXAI zvikamu.
Kushandisa XAI kune yakaoma AI masisitimu uye madomasi.