Rodha iyo LLM modhi uye tokenizer

**Kubatanidza DSPy neLLMs yeAutomated Software Testing: Iyo New Paradigm muKusimbisa Hunhu **

Introduction

Kubatanidzwa kweLarge Language Models (LLMs) uye Deep Software Programming (DSPy) kwakashandura mamiriro ekuongorora software. Nekusimudzira kugona kweLLMs uye DSPy, vagadziri vanogona kuita otomatiki maitiro ekuyedza, vachiwedzera kuvimbika uye kushanda zvakanaka muinjiniya yesoftware. Muchikamu chino, tichaongorora kuti DSPy inogona sei kugadzirisa maitiro ekuedzwa kwezvikumbiro zvakavakwa paLLMs, uye kukurukura mabhenefiti nematambudziko eiyi paradigm iri kubuda.

Chii chinonzi DSPy?

DSPy iraibhurari yePython inopa seti yezvishandiso zve automating software yekuongorora. Inobvumira vanogadzira kugadzira uye kuita bvunzo makesi, kutarisa mhinduro dzebvunzo, uye kuongorora bvunzo data. DSPy yakagadzirirwa kushanda zvisina musono neLLMs, ichigonesa kugadzirwa kweakaomesesa ebvunzo otomatiki masisitimu.

MaLLM chii?

MaLLM imhando yeartificial intelligence (AI) modhi inogona kugadzirisa uye kugadzira mutauro wakafanana nemunhu. Ivo vakadzidziswa pahuwandu hukuru hwe data data uye vanogona kudzidza kuziva mapatani, hukama, uye mamiriro. MaLLM ane akawanda maapplication muinjiniya yesoftware, kusanganisira kugadzirisa mutauro wechisikigo, kuongororwa kwemavara, uye kugadzira kodhi.

Kubatanidza DSPy neLLMs

Iko kubatanidzwa kweDSPy neLLMs kunogonesa kugadzirwa kwemaotomatiki ebvunzo makesi anogona kuita mamiriro akaomarara, kutevedzera kupindirana kwevashandisi, uye kuongorora mhinduro dzebvunzo. Kubatanidzwa uku kunopa zvikomborero zvakawanda, kusanganisira:

Kunatsiridza bvunzo yekuvhara: DSPy inogona kugadzira makesi ebvunzo anovhara akasiyana siyana ezviitiko, kuve nechokwadi chekuti mashandisirwo akanyatsoedzwa.
Kuwedzera kushanda zvakanaka: Kuedza otomatiki neDSPy neLLMs kunoderedza nguva nekuedza kunodiwa kuita nyaya dzebvunzo, kusunungura vanogadzira kuti vatarise mamwe mabasa.
** Kuvimbika kwakawedzerwa **: DSPy inogona kutarisa mhinduro dzebvunzo uye yambiro vanogadzira kune chero nyaya kana kukanganisa, kuve nechokwadi chekuti application yacho yakavimbika uye yakagadzikana.

**Muenzaniso kodhi: Kuita otomatiki Mhosva dzekuyedza neDSPy uye LLMs **

Heino muenzaniso kodhi block inoratidza mashandisiro eDSPy neLLM kuita otomatiki makesi ebvunzo:

import dspy
kubva kune vanoshandura pinza AutoModelForSeq2SeqLM, AutoTokenizer


muenzaniso = AutoModelForSeq2SeqLM.from_pretrained("t5-base")
tokenizer = AutoTokenizer.from_pretrained("t5-base")

# Gadzira DSPy bvunzo kesi
test_case = dspy.TestCase(
    zita="Test Case 1",
    tsananguro="Edza mashandiro ekupinda",
    nhanho=[
        dspy.Step("Vhura peji rekupinda"),
        dspy.Step("Isa magwaro anoshanda"),
        dspy.Step("Dzvanya bhatani rekupinda"),
    ]
)

# Ita iyo bvunzo kesi uchishandisa iyo LLM
test_results = dspy.execute_test_case(test_case, modhi, tokenizer)

# Ongorora mhinduro dzebvunzo
test_results.analyze()

Iyi kodhi yekodhi inoratidza nzira yekugadzira DSPy bvunzo kesi, kurodha LLM modhi uye tokenizer, kuita bvunzo kesi uchishandisa iyo LLM, uye ongorora mhinduro dzebvunzo.

Pamusoro 5-10 Inonyanya Kushandiswa Mirairo/Zvidimbu

Heano epamusoro 5-10 anonyanya kushandiswa mirairo/zvimedu zveDSPy neLLMs:

dspy.TestCase(): Inogadzira imwe DSPy test case.
dspy.Step(): Inowedzera danho idzva kune DSPy test kesi.
dspy.execute_test_case(): Inoita DSPy test case uchishandisa LLM.
model = AutoModelForSeq2SeqLM.from_pretrained("t5-base"): Inotakura modhi yeLLM.
tokenizer = AutoTokenizer.from_pretrained("t5-base"): Inotakura LLM tokenizer.
test_results.analyze(): Inoongorora zvabuda zve DSPy test case.
dspy.monitor_test_results(): Inotarisa zvinenge zvabuda pabvunzo yeDSPy.
dspy.alert_developer(): Inozivisa mugadziri nezvechero nyaya kana kukanganisa.
dspy.generate_test_cases(): Inogadzira makesi matsva ebvunzo uchishandisa LLM.
dspy.optimize_test_cases(): Inonatsiridza kuitwa kwemakesi ebvunzo uchishandisa LLM.

Kuona

Heino ratidziro yekubatanidzwa kweDSPy neLLM uchishandisa Chart.js:

{
  "type": "bha",
  "data": {
    "labels": ["Test Case 1", "Test Case 2", "Test Case 3"],
    "datasets": [
      {
        "label": "Test Case Execution Time",
        "data": [10, 20, 30],
        "backgroundColor": "rgba(255, 99, 132, 0.2)",
        "borderColor": "rgba(255, 99, 132, 1)",
        "borderWidth": 1
      },
      {
        "label": "Test Case Failure Rate",
        "data": [5, 10, 15],
        "backgroundColor": "rgba(54, 162, 235, 0.2)",
        "borderColor": "rgba(54, 162, 235, 1)",
        "borderWidth": 1
      }
    ]
  },
  "sarudzo": {
    "zita": {
      "kuratidza": chokwadi,
      "text": "DSPy uye LLM Kubatanidzwa"
    },
    "zvikero": {
      "yAxes": [
        {
          "zvikwekwe": {
            "beginAtZero": chokwadi
          }
        }
      ]
    }
  }
}

Ichi chitarisiko chinoratidza nguva yekuuraya uye mwero wekutadza kwemakesi matatu ebvunzo anoitwa uchishandisa DSPy neLLM.

Mhedziso

Kubatanidzwa kweDSPy neLLMs kwakashandura mamiriro ekuongorora software, zvichiita kuti kugadzirwa kwemakesi ekuedza otomatiki anogona kuita mamiriro akaoma, kutevedzera kupindirana kwevashandisi, uye kuongorora mhinduro dzebvunzo. Nekusimudzira kugona kweDSPy neLLMs, vagadziri vanogona kukwenenzvera maitiro ekuyedza, kuwedzera kuvimbika uye kushanda nesimba, uye kugadzirisa huwandu hwese hwekushandisa kwavo. Sezvo tekinoroji iyi ichiramba ichishanduka, tinogona kutarisira kuona zvakatowanda zvekushandisa zveDSPy neLLMs muinjiniya yesoftware.

Rodha iyo LLM modhi uye tokenizer

Have Questions?