Qalisa i-NVIDIA Nemotron 3 Nano njengemodeli ephethwe ngokugcwele engenaseva ku-Amazon Bedrock

nimda March 9, 2026

0 4 7 minutes read

Qalisa i-NVIDIA Nemotron 3 Nano njengemodeli ephethwe ngokugcwele engenaseva ku-Amazon Bedrock

Lokhu okuthunyelwe kubhalwe no-Abdullahi Olaoye, uCurtice Lockhart, u-Nirmal Kumar Juluru wase-NVIDIA.

Siyajabula ukumemezela ukuthi i-Nemotron 3 Nano ye-NVIDIA isiyatholakala njengemodeli ephethwe ngokugcwele futhi engenaseva e-Amazon Bedrock. Lokhu kulandela isimemezelo sethu sangaphambilini ku-AWS re:Invent esekela imodeli ye-NVIDIA Nemotron 2 Nano 9B kanye ne-NVIDIA Nemotron 2 Nano VL 12B.

Ngamamodeli avuliwe e-NVIDIA Nemotron ku-Amazon Bedrock, ungasheshisa ukusungula izinto ezintsha futhi ulethe inani lebhizinisi elibambekayo ngaphandle kokuphatha ubunzima bengqalasizinda. Unganika amandla izinhlelo zakho zokusebenza ze-AI ezikhiqizayo ngamakhono e-Nemotron ngokusebenzisa amandla okucabanga e-Amazon Bedrock futhi usebenzise inzuzo yezici zayo ezibanzi kanye namathuluzi.

Lokhu okuthunyelwe kuhlola izici zobuchwepheshe zemodeli ye-NVIDIA Nemotron 3 Nano futhi kudingida izimo ezingaba khona zokusebenzisa uhlelo lokusebenza. Ukwengeza, inikeza isiqondiso sobuchwepheshe ukukusiza ukuthi uqalise ukusebenzisa le modeli yezinhlelo zakho zokusebenza ezikhiqizayo ze-AI ngaphakathi kwemvelo ye-Amazon Bedrock.

Mayelana ne-Nemotron 3 Nano

I-NVIDIA Nemotron 3 Nano imodeli yolimi encane (SLM) ene-hybrid Mixture-of-Experts (MoE) yezakhiwo eletha ukusebenza kahle kwekhompyutha okuphakeme nokunemba onjiniyela abangakusebenzisa ukuze bakhe amasistimu e-AI akhethekile. Imodeli ivuleke ngokugcwele ngezisindo ezivulekile, amasethi edatha, nezindlela zokupheka ezisiza ukubonakala nokuzethemba konjiniyela namabhizinisi. Uma kuqhathaniswa namanye amamodeli anosayizi ofanayo, i-Nemotron 3 Nano ihamba phambili ekubhaleni amakhodi nemisebenzi yokucabanga, ihola kumabhentshimakhi afana ne-SWE Bench Verified, i-AIME 2025, i-Arena Hard v2, ne-IFBench.

Uhlolojikelele lwemodeli:

I-Architecture:
- Ingxube-Yochwepheshe (i-MoE) ne-Hybrid Transformer-Mamba Architecture
- Isekela Ibhajethi Yethokheni yokuhlinzeka ngokunemba ngenkathi igwema ukucabanga ngokweqile
Ukunemba:
- Ukunemba okuholayo ekubhaleni amakhodi, ukucabanga kwesayensi, izibalo, ukushaya kwamathuluzi, ukulandela imiyalelo, kanye nengxoxo
- I-Nemotron 3 Nano ihamba phambili ngamabhentshimakhi afana ne-SWE Bench, i-AIME 2025, i-Humanity Last Exam, i-IFBench, i-RULER, ne-Arena Hard (uma kuqhathaniswa nezinye izinhlobo zezilimi ezivulekile ezine-MoE eyizigidi eziyizinkulungwane ezingu-30 noma ngaphansi)
Usayizi wemodeli: 30 B namapharamitha asebenzayo angu-3 B
Ubude bomongo: 256K
Okokufaka kwemodeli: Umbhalo
Okukhipha imodeli: Umbhalo

I-Nemotron 3 Nano ihlanganisa izendlalelo ze-Mamba, i-Transformer, ne-Mixture-of-Experts ibe umgogodla owodwa ukusiza ukulinganisela ukusebenza kahle, ukunemba kokucabanga, nesikali. I-Mamba inika amandla ukumodeliswa kokulandelana kwebanga elide ngememori ephansi, kuyilapho izendlalelo ze-Transformer zisiza ukwengeza ukunaka okunembile kwemisebenzi yokucabanga ehlelekile njengekhodi, izibalo, nokuhlela. Umzila we-MoE uqhubeka nokukhulisa ukuqina ngokwenza kusebenze isethi encane yochwepheshe ngethokheni ngayinye, okusiza ukuthuthukisa ukubambezeleka nokusebenza. Lokhu kwenza i-Nemotron 3 Nano ifaneleke kahle ikakhulukazi amaqoqo e-ejenti asebenzisa ukuhamba komsebenzi okuningi ngasikhathi sinye, okungasindi.

Ukuze ufunde kabanzi mayelana nezakhiwo ze-Nemotron 3 Nano nokuthi iqeqeshwa kanjani, bona Ngaphakathi kwe-NVIDIA Nemotron 3: Amasu, Amathuluzi, kanye Nedatha Eyenza Isebenze Kahle Futhi Inembile.

Amabhentshimakhi angamamodeli

Isithombe esilandelayo sibonisa ukuthi i-Nemotron 3 Nano ihola ngequadrant ekhanga kakhulu ku-Artificial Analysis Openness Index vs. Intelligence Index. Kungani ukuvuleleka kubalulekile: Kwakha ukwethembana ngokwenza izinto obala. Onjiniyela namabhizinisi bangakha ngokuzethemba ku-Nemotron ngokubonakala okucacile kumodeli, ipayipi ledatha, nezici zedatha, okuvumela ukuhlola okuqondile kanye nokuphatha.

Isihloko: Ishadi elibonisa i-Nemotron 3 Nano kuquadrant ekhanga kakhulu ku-Artificial Analysis Openness vs Intelligence Index (Umthombo: Ukuhlaziya Okwenziwayo)

Njengoba kuboniswe esithombeni esilandelayo, i-Nemotron 3 Nano ihlinzeka ngokunemba okuholayo ngokusebenza kahle kakhulu phakathi kwamamodeli avulekile futhi ithola amaphuzu amangalisayo angu-52, ukweqa okubalulekile phezu kwemodeli yangaphambilini ye-Nemotron 2 Nano. Isidingo samathokheni siyanda ngenxa ye-AI ye-agent, ngakho-ke ikhono 'lokucabanga ngokushesha' (ukufika empendulweni efanele ngokushesha kuyilapho usebenzisa amathokheni ambalwa) libalulekile. I-Nemotron 3 Nano iletha ukusebenza okuphezulu nge-Hybrid Transformer-Mamba nezakhiwo zayo ze-MoE.

Isihloko: I-NVIDIA Nemotron 3 Nano ihlinzeka ngokusebenza kahle kakhulu ngokunemba okuholayo phakathi kwamamodeli avuliwe anamaphuzu ahlaba umxhwele angu-52 ku-Artificial Analysis Intelligence vs. Output Speed Index. (Umthombo: Ukuhlaziya okwenziwayo)

I-NVIDIA Nemotron 3 Nano amacala okusebenzisa

I-Nemotron 3 Nano isiza amandla amacala okusetshenziswa ahlukahlukene ezimboni ezahlukene. Ezinye zezimo zokusetshenziswa zihlanganisa

Ezezimali – Sheshisa ukucutshungulwa kwemalimboleko ngokukhipha idatha, ukuhlaziya amaphethini wemali engenayo, ukuthola imisebenzi yokukhwabanisa, ukunciphisa izikhathi zokujikeleza, kanye nobungozi.
I-Cybersecurity – Hlola ngokuzenzakalelayo ubungozi, yenza ukuhlaziya okujulile kwe-malware, futhi uzingele ngokuqhubekayo izinsongo zokuphepha.
Ukuthuthukiswa kwesoftware – Siza ngemisebenzi efana nokufingqwa kwekhodi.
Ukuthengisa – Lungiselela ukuphathwa kwempahla futhi usize ukuthuthukisa isevisi yasesitolo ngesikhathi sangempela, izincomo zomkhiqizo womuntu siqu kanye nokwesekwa.

Qalisa nge-NVIDIA Nemotron 3 Nano e-Amazon Bedrock

Ukuze uhlole i-NVIDIA Nemotron 3 Nano e-Amazon Bedrock, qedela lezi zinyathelo ezilandelayo:

Zulazulela ku- I-Amazon Bedrock console bese ukhetha Inkundla yokudlala yengxoxo/yombhalo kusuka kumenyu engakwesokunxele (ngaphansi kwe- Hlola ingxenye).
Khetha Khetha imodeli ekhoneni eliphezulu kwesokunxele lenkundla yokudlala.
Khetha I-NVIDIA ohlwini lwezigaba, bese ukhetha I-NVIDIA Nemotron 3 Nano.
Khetha Faka isicelo ukulayisha imodeli.

Ngemuva kokukhetha, ungahlola imodeli ngokushesha. Masisebenzise umyalo olandelayo ukwenza ukuhlolwa kweyunithi kukhodi yePython sisebenzisa i pytest uhlaka:

Write a pytest unit test suite for a Python function called calculate_mortgage(principal, rate, years). Include test cases for: 1) A standard 30-year fixed loan 2) An edge case with 0% interest 3) Error handling for negative input values.

Imisebenzi eyinkimbinkimbi efana nalesi saziso ingazuza ochungechungeni lwendlela yokucabanga ukuze isize ukukhiqiza umphumela onembayo osuselwe emandleni okucabanga akhiwe ngokwendalo kumodeli.

Ukusebenzisa i-AWS CLI nama-SDK

Ungakwazi ukufinyelela imodeli ngokohlelo usebenzisa i-ID yemodeli nvidia.nemotron-nano-3-30b. Imodeli isekela kokubili InvokeModel futhi Converse Ama-API nge-AWS Command Line Interface (AWS CLI) kanye ne-AWS SDK nge nvidia.nemotron-nano-3-30b njenge-ID yemodeli. Ngaphezu kwalokho, isekela i-Amazon Bedrock OpenAI SDK API ehambisanayo.

Qalisa umyalo olandelayo ukuze ucele imodeli ngokuqondile kutheminali yakho usebenzisa i I-AWS Command Line Interface (AWS CLI) kanye ne-InvokeModel API:

aws bedrock-runtime invoke-model  
 --model-id nvidia.nemotron-nano-3-30b  
 --region us-west-2  
 --body '{"messages": [{"role": "user", "content": "Type_Your_Prompt_Here"}], "max_tokens": 512, "temperature": 0.5, "top_p": 0.9}'  
 --cli-binary-format raw-in-base64-out  
invoke-model-output.txt

Ukuze ucele imodeli nge-AWS SDK yePython (boto3), sebenzisa umbhalo olandelayo ukuze uthumele ukwaziswa kumodeli, kulokhu ngokusebenzisa i-Converse API:

import boto3 
from botocore.exceptions import ClientError 

# Create a Bedrock Runtime client in the AWS Region you want to use. 
client = boto3.client("bedrock-runtime", region_name="us-west-2") 

# Set the model ID
model_id = "nvidia.nemotron-nano-3-30b" 

# Start a conversation with the user message. 

user_message = "Type_Your_Prompt_Here" 
conversation = [ 
   { 
       "role": "user", 

       "content": [{"text": user_message}], 
   } 
]  

try: 
   # Send the message to the model using a basic inference configuration. 
   response = client.converse( 
        modelId=model_id, 

       messages=conversation, 
        inferenceConfig={"maxTokens": 512, "temperature": 0.5, "topP": 0.9}, 
   ) 
 
   # Extract and print the response text. 
    response_text = response["output"]["message"]["content"][0]["text"] 
   print(response_text)

except (ClientError, Exception) as e: 
    print(f"ERROR: Can't invoke '{model_id}'. Reason: {e}") 
    exit(1)

Ukunxenxa imodeli nge-Amazon Bedrock OpenAI-ehambisanayo ChatCompletions ekugcineni, ungakwenza lokho ngokusebenzisa i-OpenAI SDK:

# Import OpenAI SDK
from openai import OpenAI

# Set environment variables
os.environ["OPENAI_API_KEY"] = ""
os.environ["OPENAI_BASE_URL"] = "https://bedrock-runtime..amazon.com/openai/v1"

# Set the model ID
model_id = "nvidia.nemotron-nano-3-30b"

# Set prompts
system_prompt = “Type_Your_System_Prompt_Here”
user_message = "Type_Your_User_Prompt_Here"


# Use ChatCompletionsAPI
response = client.chat.completions.create(
    model= model _ID,                 
    messages=[
        {"role": "system", "content": system_prompt},
        {"role": "user",   "content": user_message}
    ],
    temperature=0,
    max_completion_tokens=1000
)
 
# Extract and print the response text
print(response.choices[0].message.content)

Sebenzisa i-NVIDIA Nemotron 3 Nano enezici ze-Amazon Bedrock

Ungathuthukisa izinhlelo zakho ezikhiqizayo ze-AI ngokuhlanganisa i-Nemotron 3 Nano namathuluzi aphethwe yi-Amazon Bedrock. Sebenzisa i-Amazon Bedrock Guardrails ukuze usebenzise izivikelo kanye Nezisekelo Zolwazi ze-Amazon ukuze udale ukuhamba komsebenzi okuqinile kokubuyiswa kwe-Augmented Generation (RAG).

I-Amazon Bedrock guardrails

I-Guardrails isendlalelo sokuphepha esiphethwe esiza ukusebenzisa i-AI enomthwalo wemfanelo ngokuhlunga okuqukethwe okuyingozi, ukuhlela kabusha ulwazi olubucayi (PII), nokuvimbela izihloko ezithile kuzo zonke iziyalezo nezimpendulo. Isebenza kuwo wonke amamodeli amaningi ukusiza ukuthola ukuhlaselwa komjovo osheshayo kanye nokubona izinto ezingekho.

Isibonelo sokusetshenziswa: Uma wakha umsizi we-mortgage, ungasiza ukukuvimbela ekunikezeni iseluleko esivamile sokutshala izimali. Ngokulungiselela isihlungi segama elithi “stocks”, ukwaziswa komsebenzisi okuqukethe lelo gama kungavinjelwa ngokushesha futhi zithole umlayezo wangokwezifiso.

Ukuze usethe i-guardrail, gcwalisa lezi zinyathelo ezilandelayo:

Kwe I-Amazon Bedrock consolezulazula uye ku Yakha kwesokunxele bese ukhetha Abaqaphi.
Dala i-guardrail entsha futhi ulungiselele izihlungi ezidingekayo zecala lakho lokusebenzisa.

Ngemva kokuyilungisa, hlola i-Guardrail ngezaziso ezihlukahlukene ukuze uqinisekise ukusebenza kwayo. Ungakwazi ke ukushuna izilungiselelo, njengezihloko ezinqatshiwe, izihlungi zamagama, nokuhlenga kabusha kwe-PII, ukuze kufane nezidingo zakho ezithile zokuphepha. Ukuze uthole ukujula okujulile, bona okuthi Dala i-Guardrail yakho.

Izisekelo Zolwazi ze-Amazon Bedrock

I-Amazon Bedrock Knowledge Bases yenza ngokuzenzakalelayo ukuhamba komsebenzi okuphelele kwe-RAG. Iphatha ukungeniswa kokuqukethwe okuvela emithonjeni yedatha yakho, ikuhlanganise kube izingxenye eziseshekayo, iziguqule zibe ukushumeka kwe-vector, futhi ikugcine kusizindalwazi se-vector. Bese, lapho umsebenzisi ehambisa umbuzo, isistimu ifanisa okokufaka ngokumelene namavekhtha agciniwe ukuze kutholwe okuqukethwe okufana nesemantiki, okubese kusetshenziselwa ukukhulisa ukwaziswa okuthunyelwa kumodeli yesisekelo.

Kulesi sibonelo, silayishe ama-PDF (ngokwesibonelo, Ukuthenga Ikhaya Elisha, Ikhithi yamathuluzi yokubolekwa kwekhaya, Ukuthenga Imali Endayo) ku-Amazon Simple Storage Service (Amazon S3) futhi sakhetha i-Amazon OpenSearch Serverless njengesitolo se-vector. Ikhodi elandelayo ibonisa indlela yokubuza lesi sisekelo solwazi usebenzisa i-RetrieveAndGenerate API, kuyilapho kusiza ngokuzenzakalelayo ukuqondana nokuthobela ukuphepha nge-ID ethile ye-Guardrail.

import boto3
bedrock_agent_runtime_client = boto3.client('bedrock-agent-runtime')
response = bedrock_agent_runtime_client.retrieve_and_generate(
    input={
        'text': 'I am interested in purchasing a home. What steps should I take to make sure I am prepared to take on a mortgage?'
    },
    retrieveAndGenerateConfiguration={
        'knowledgeBaseConfiguration': {
            'generationConfiguration': {
                'guardrailConfiguration': {
                    'guardrailId': '',
                    'guardrailVersion': '1'
                }
            },
            'knowledgeBaseId': '',
            'modelArn': 'arn:aws:bedrock:us-east-1::foundation-model/nvidia.nemotron-nano-3-30b',
            "generationConfiguration": {
                "promptTemplate": {
                    "textPromptTemplate": (
                        "You are a helpful assistant that answers questions about mortgages"
                        "search results.nn"
                        "Search results:n$search_results$nn"
                        "User query:n$query$nn"
                        "Answer clearly and concisely."
                    )
                },
            },
            "orchestrationConfiguration": {
                "promptTemplate": {
                    "textPromptTemplate": (
                        "You are very knowledgeable on mortgages"
                        "Conversation so far:n$conversation_history$nn"
                        "User query:n$query$nn"
                        "$output_format_instructions$"
                    )
                }
            }
        },
        'type': 'KNOWLEDGE_BASE'
    }
)
print(response)

Iqondisa imodeli ye-NVIDIA Nemotron 3 Nano ukuze ihlanganise amadokhumenti abuyisiwe abe yimpendulo ecacile, enesisekelo usebenzisa isifanekiso sakho sokwaziswa ngokwezifiso. Ukuze uzibekele owakho umzila, buyekeza uhambo olugcwele ku-Amazon Bedrock User Guide.

Isiphetho

Kulokhu okuthunyelwe, sikubonise ukuthi ungaqala kanjani nge-NVIDIA Nemotron 3 Nano ku-Amazon Bedrock ukuze uthole inference ephethwe ngokuphelele engenasiphakeli. Siphinde sakubonisa ukuthi ungasebenzisa kanjani imodeli nge-Amazon Bedrock Knowledge Bases kanye ne-Amazon Bedrock Guardrails. Imodeli isiyatholakala e-US East (N. Virginia), US East (Ohio), US West (Oregon), Asia Pacific (Tokyo), Asia Pacific (Mumbai), South America (Sao Paulo), Europe (London), and Europe (Milan) AWS Regions. Hlola uhlu oluphelele Lwesifunda ukuze uthole izibuyekezo ezizayo. Ukuze ufunde kabanzi, hlola i-NVIDIA Nemotron bese uzama i-NVIDIA Nemotron 3 Nano kukhonsoli ye-Amazon Bedrock namuhla.