Search is not available for this dataset
source_url string | chunk_id string | text string |
|---|---|---|
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_0 | Sunflower API Documentation Version 1.0 Date September 2025 Developed by Sunbird AI Table of Contents Introduction Authentication Rate Limiting API Endpoints Error Handling Code Examples Best Practices Troubleshooting Introduction The Sunflower API provides access to Sunbird AIs multilingual language model, specificall... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_1 | ons in Ugandan languages (Luganda, Acholi, Ateso, etc.) Cross-lingual translations and explanations Cultural context understanding Educational content in local languages Key Features Automatic retry with exponential backoff Context-aware responses Usage tracking and monitoring Support for custom system messages Message... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_10 | 25-09-26T103000Z Common Error Messages Model Loading (503) detail The AI model is currently loading. This usually takes 2-3 minutes. Please try again shortly. Timeout (504) detail The request timed out. Please try again with a shorter prompt or check your network connection. Invalid Request (400) detail Messag... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_11 | ion import requests import json def chatwithsunflower ( messages , apikey , baseurl httpsapi.sunbird.ai ) Send a chat completion request to Sunflower API url f baseurl taskssunflowerinference headers Authorization f Bearer apikey , Content-Type applicationjson payload messages messages , modeltype qw... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_12 | aders headers , json payload , timeout 30 ) response . raiseforstatus () return response . json () except requests . exceptions . RequestException as e print ( f Request failed e ) return None Example usage apikey yourapikeyhere messages role system , content You are Sunflower, a multilingual assistant fo... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_13 | morning to Luganda result chatwithsunflower ( messages , apikey ) if result print ( f Response result content ) print ( f Tokens used result usage totaltokens ) 2. Simple Inference import requests def simpleinference ( instruction , apikey , baseurl httpsapi.sunbird.ai ) Send a simple inference request... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_14 | n f Bearer apikey data instruction instruction , modeltype qwen , temperature 0.3 try response requests . post ( url , headers headers , data data , timeout 30 ) response . raiseforstatus () return response . json () except requests . exceptions . RequestException as e print ( f Request failed e ) r... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_15 | g of Ubuntu in Ugandan context result simpleinference ( instruction , apikey ) if result print ( f Response result response ) 3. Conversation Management class SunflowerConversation def init ( self , apikey , baseurl httpsapi.sunbird.ai ) self . apikey apikey self . baseurl baseurl self . messages def setsy... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_16 | system message self . messages msg for msg in self . messages if msg role ! system Add new system message at the beginning self . messages . insert ( 0 , role system , content message ) def addusermessage ( self , content ) Add a user message to the conversation self . messages . append ( role user , content... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_17 | response and add it to conversation url f self . baseurl taskssunflowerinference headers Authorization f Bearer self . apikey , Content-Type applicationjson payload messages self . messages , modeltype modeltype , temperature temperature try response requests . post ( url , headers headers , json ... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_18 | Add assistant response to conversation self . messages . append ( role assistant , content result content ) return result except requests . exceptions . RequestException as e print ( f Request failed e ) return None def clearconversation ( self ) Clear conversation history but keep system message systemmessag... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_19 | emmessages Example usage conversation SunflowerConversation ( yourapikeyhere ) Set system message conversation . setsystemmessage ( You are Sunflower, a helpful assistant specializing in Ugandan languages and culture. ) Have a conversation conversation . addusermessage ( Hello, can you greet me in Luganda? ) respon... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_2 | SunbirdSunflower-14B-FP8 Supported Languages Luganda, Acholi, Ateso, English, and other Ugandan languages Model Types Qwen Authentication All API requests require authentication using a valid API key. Include your API key in the request headers Authorization Bearer YOURAPIKEY Obtaining API Keys If you dont already have... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_20 | ation . addusermessage ( How do I say thank you in Acholi? ) response2 conversation . getresponse () print ( f AI response2 content ) 4. Error Handling with Retry Logic import time import random from typing import Optional def sunflowerrequestwithretry ( messages , apikey , maxretries 3 , basedelay 2.0 , baseur... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_21 | exponential backoff retry url f baseurl taskssunflowerinference headers Authorization f Bearer apikey , Content-Type applicationjson payload messages messages , modeltype qwen , temperature 0.3 for attempt in range ( maxretries 1 ) try response requests . post ( url , headers headers , json pay... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_22 | if response . statuscode 503 Model loading - retry with longer delay if attempt maxretries delay basedelay ( 2 attempt ) random . uniform ( 0 , 1 ) print ( f Model loading, retrying in delay .1f seconds... ) time . sleep ( delay ) continue elif response . statuscode 429 Rate limited - retry with exponent... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_23 | ate limited, retrying in delay .1f seconds... ) time . sleep ( delay ) continue else print ( f Request failed with status response . statuscode response . text ) return None except requests . exceptions . Timeout if attempt maxretries delay basedelay ( 2 attempt ) print ( f Request timed out, retrying i... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_24 | t timed out after all retries ) return None except requests . exceptions . RequestException as e print ( f Request failed e ) return None print ( All retry attempts exhausted ) return None Example usage messages role user , content What is the capital of Uganda? result sunflowerrequestwithretry ( messages ... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_25 | tNode.js Examples 1. Basic Chat Completion const axios require ( axios ) async function chatWithSunflower ( messages , apiKey , baseUrl httpsapi.sunbird.ai ) const url baseUrl taskssunflowerinference const headers Authorization Bearer apiKey , Content-Type applicationjson const payload messages messa... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_26 | wait axios . post ( url , payload , headers headers , timeout 30000 ) return response . data catch ( error ) console . error ( Request failed , error . message ) if ( error . response ) console . error ( Response data , error . response . data ) return null Example usage const apiKey yourapikeyhere const ... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_27 | for Ugandan languages. , role user , content How do you say welcome in Luganda? chatWithSunflower ( messages , apiKey ) . then ( result if ( result ) console . log ( Response , result . content ) console . log ( Tokens used , result . usage . totaltokens ) ) 2. Simple Inference const FormData require ( form-d... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_28 | n , apiKey , baseUrl httpsapi.sunbird.ai ) const url baseUrl taskssunflowersimple const formData new FormData () formData . append ( instruction , instruction ) formData . append ( modeltype , qwen ) formData . append ( temperature , 0.3 ) const headers Authorization Bearer apiKey , ... formData . getHead... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_29 | ders , timeout 30000 ) return response . data catch ( error ) console . error ( Request failed , error . message ) return null Example usage const apiKey yourapikeyhere const instruction Translate How are you? to Ateso simpleInference ( instruction , apiKey ) . then ( result if ( result ) console . log (... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_3 | ns page to get your access token which youll use to authenticate Rate Limiting The API implements rate limiting based on account types Free Tier Limited requests per hour Professional Higher rate limits Enterprise Custom rate limits API Endpoints 1. Chat Completions Endpoint Endpoint POST taskssunflowerinference Profes... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_30 | httpsapi.sunbird.aitaskssunflowerinference -H Authorization Bearer YOURAPIKEY -H Content-Type applicationjson -d messages role system, content You are Sunflower, a multilingual assistant for Ugandan languages made by Sunbird AI. , role user, content Translate Good evening to Luganda , modeltype qwen, temperatu... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_31 | -H Authorization Bearer YOURAPIKEY -F instructionWhat are some traditional Ugandan foods? -F modeltypeqwen -F temperature0.3 Best Practices 1. Message Management Always include a system message to provide context and improve response quality Keep conversation history relevant - trim old messages to stay within tok... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_32 | ment retry logic for transient errors (503, 504, 429) Handle model loading delays - allow 2-3 minutes for cold starts Validate input before sending requests 3. Performance Optimization Use appropriate temperature settings 0.0-0.3 More deterministic, good for translations 0.4-0.7 Balanced creativity 0.8-1.0 More creati... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_33 | nt caching for repeated requests 4. Rate Limiting Monitor rate limit headers in responses Implement exponential backoff for rate limit exceeded errors Consider request batching where appropriate 5. Security Never expose API keys in client-side code Use environment variables to store credentials Implement proper authent... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_34 | l Loading Errors (503) Problem The AI model is currently loading Solutions - Wait 2-3 minutes before retrying - Implement exponential backoff retry logic - Use longer timeout values during cold starts 2. Empty Responses (502) Problem The model returned an empty response Solutions - Rephrase your request to be more spec... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_35 | rature setting 3. Timeout Errors (504) Problem Request times out Solutions - Reduce prompt length - Use simpler queries - Increase timeout values in your code - Check network connectivity 4. Rate Limiting (429) Problem Too many requests Solutions - Implement request queuing - Use exponential backoff - Consider upgradin... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_36 | rrors Solutions - Ensure all messages have role and content fields - Check that content is not empty - Validate role values (system, user, assistant) Debugging Tips Enable detailed logging to track requestresponse cycles Test with simple queries first before complex conversations Use the simple endpoint to isolate issu... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_37 | pected costs Support and Resources Getting Help Email Support infosunbird.ai Documentation httpssalt.sunbird.aidocs 2025 Sunbird AI. All rights reserved. This documentation is subject to updates and improvements. Please check the official documentation website for the latest version. |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_4 | ent. Request Format messages role systemuserassistant , content Message content , modeltype qwen , temperature 0.3 , stream false , systemmessage Optional custom system message Parameters Parameter Type Required Default Description messages Array Yes - List of conversation messages modeltype String No qwen ... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_5 | oolean No false Whether to stream the response systemmessage String No null Custom system message Message Object Field Type Required Description role String Yes Message role system, user, or assistant content String Yes Message content (cannot be empty) Response Format content AI response content , modeltype qwen , ... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_6 | gtime 2.35 , inferencetime 1.85 , messagecount 3 Response Fields Field Type Description content String The AIs response modeltype String Model used for inference usage Object Token usage statistics processingtime Float Total processing time in seconds inferencetime Float Model inference time in seconds messagecount... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_7 | taskssunflowersimple Simplified interface for single instructionresponse interactions. Request Format (Form Data) instruction Translate Good morning to Luganda , modeltype qwen , temperature 0.3 , systemmessage You are Sunflower, a multilingual assistant for Ugandan languages made by Sunbird AI. Parameter Type R... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_8 | modeltype String No qwen Model type qwen temperature Float No 0.3 Sampling temperature (0.0-2.0) systemmessage String No null Custom system message Response Format response AI response content , modeltype qwen , processingtime 1.85 , usage completiontokens 120 , prompttokens 80 , totaltokens 200 , success tr... |
https://salt.sunbird.ai/sunflower/sunflower/#__codelineno-3-1 | 100_9 | error messages. HTTP Status Codes Code Description 200 Success 400 Bad Request - Invalid input 401 Unauthorized - Invalid API key 429 Too Many Requests - Rate limit exceeded 500 Internal Server Error 502 Bad Gateway - Empty model response 503 Service Unavailable - Model loading 504 Gateway Timeout - Request timeout Er... |
https://salt.sunbird.ai/sunflower/overview/ | 10_0 | Sunflower Quantized Inference Overview The Sunflower models are available in 14B and 32B sizes and support 8-bit and 4-bit quantized inference for efficient performance on GPUs with limited memory. Quantization reduces memory requirements while keeping inference quality high, enabling large models to run on consumer-gr... |
https://salt.sunbird.ai/sunflower/overview/ | 10_1 | (16GB for 14B) Lower (10GB for 14B) Speed Fast Faster Accuracy Very Good Slightly Lower VRAM Efficiency Moderate High Important Do not set both 8-bit and 4-bit modes at the same time. Sunflower 14B Models 14B 8-bit Balanced memory and accuracy, suitable for most GPUs. 14B 4-bit Optimized for memory-limited GPUs and... |
https://salt.sunbird.ai/sunflower/overview/ | 10_2 | bit High accuracy, requires more GPU memory. 32B 4-bit Reduced memory usage, faster inference, slightly lower accuracy. The usage process is identical for 14B and 32B only model size and quantization type differ. Tips Best Practices Use 4-bit models when GPU memory is limited or faster inference is needed. 8-bit model... |
https://salt.sunbird.ai/sunflower/overview/ | 10_3 | or 4-bit for a model. For large inputs or batch processing, monitor GPU memory to avoid out-of-memory errors. Adjust inference parameters (like sequence length or tokens) for optimal performance based on your hardware. |
https://salt.sunbird.ai/sunflower/inference/ | 11_0 | Sunbird Sunflower Quantized Model Usage Installation Installation Install the required Python packages bash !pip install torch transformers accelerate bitsandbytes torch PyTorch library for model computation. transformers Hugging Face Transformers library for model handling. accelerate Optimizes model loading acros... |
https://salt.sunbird.ai/sunflower/inference/ | 11_1 | mport the necessary libraries for loading and running the model. import torch import transformers from transformers import AutoModelForCausalLM , AutoTokenizer , BitsAndBytesConfig import os from getpass import getpass 3. Authentication Set your Hugging Face token to access models os . environ HFTOKEN getpass ( HFT... |
https://salt.sunbird.ai/sunflower/inference/ | 11_2 | u want to use, e.g., SunbirdSunflower-14B-4bit-nf4-bnb or SunbirdSunflower-14B-8bit-bnb . Load tokenizer tokenizer AutoTokenizer . frompretrained ( MODELNAME ) Configure quantization (choose one) quantizationconfig BitsAndBytesConfig ( For 4-bit models loadin4bit True , bnb4bitquanttype nf4 , bnb4bitcomputedtype... |
https://salt.sunbird.ai/sunflower/inference/ | 11_3 | t above and use loadin8bitTrue, llmint8threshold6.0 ) Load model model AutoModelForCausalLM . frompretrained ( MODELNAME , quantizationconfig quantizationconfig , devicemap auto , or cuda0 for single GPU trustremotecode True ) Note Only change MODELNAME and adjust quantization parameters. Everything else remain... |
https://salt.sunbird.ai/sunflower/inference/ | 11_4 | nt role) SYSTEMMESSAGE You are Sunflower, a multilingual assistant for Ugandan languages made by Sunbird AI. You specialise in accurate translations, explanations, summaries, and other cross-lingual tasks. Example user prompt prompttext Sunbird AI is a non-profit research organization in Kampala, Uganda. We build an... |
https://salt.sunbird.ai/sunflower/inference/ | 11_5 | ompt f Translate to Luganda prompttext Structure messages messages role system , content SYSTEMMESSAGE , role user , content userprompt Prepare model input prompt tokenizer . applychattemplate ( messages , tokenize False , addgenerationprompt True ) inputs tokenizer ( prompt , returntensors pt ) . ... |
https://salt.sunbird.ai/sunflower/inference/ | 11_6 | eal-time textstreamer transformers . TextStreamer ( tokenizer , skipprompt True , skipspecialtokens True ) Generation parameters numbeams 5 Adjust for higher quality generationconfig maxnewtokens 512 , Maximum tokens to generate temperature 0.3 , Lower more deterministic dosample True , Enable sampling n... |
https://salt.sunbird.ai/sunflower/inference/ | 11_7 | width Beam search improves output quality but slows generation. TextStreamer works only when numbeams 1 . 7. Generate Output Generate model output outputs model . generate ( inputs , generationconfig , streamer textstreamer if numbeams 1 else None , ) Decode if using multi-beam search if numbeams 1 response ... |
https://salt.sunbird.ai/sunflower/inference/ | 11_8 | True ) print ( response ) Single-beam real-time streaming. Multi-beam output decoded after generation. 8. Notes for Future Models |
https://salt.sunbird.ai/sunflower/quantization/ | 12_0 | Converting Sunflower LoRA Fine-tuned Models to GGUF Quantizations In this guide, provide a tutorial for converting LoRA fine-tuned Sunflower models to GGUF format with multiple quantization levels, including experimental ultra-low bit quantizations. Table of Contents Prerequisites Hardware Requirements RAM Minimum 32G... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_1 | iles GPU Optional but recommended for faster processing Software Requirements LinuxmacOS (WSL2 for Windows) Python 3.9 Git and Git LFS CUDA toolkit (optional, for GPU acceleration) Environment Setup 1. Install Dependencies Install Python packages pip install torch transformers accelerate pip install sentencepiece pro... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_10 | ce matrix (this takes 30-60 minutes) .llama.cppbuildbinllama-imatrix -m ggufoutputsmodel-merged-f16.gguf -f wiki.test.raw --chunk 512 -o ggufoutputsmodel-imatrix.dat -ngl 32 --verbose Note Adjust -ngl based on your GPU memory (0 for CPU-only) Note Understanding Importance Matrix (imatrix) The importance matrix i... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_11 | gnificantly to output quality. During quantization, weights deemed important by the matrix receive higher precision allocation, while less critical weights can be more aggressively compressed. This selective approach significantly improves quantized model quality compared to uniform quantization. The imatrix is generat... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_12 | terns. While general text datasets (like WikiText) work well for most models, using domain-specific calibration data (e.g., translation examples for the Sunflower model) can provide marginal quality improvements. The process adds 30-60 minutes to quantization time but is highly recommended for production models, especi... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_13 | zations Create quantized models with different qualitysize trade-offs Q80 Near-lossless quality (15GB for 14B model) .llama.cppbuildbinllama-quantize --imatrix ggufoutputsmodel-imatrix.dat ggufoutputsmodel-merged-f16.gguf ggufoutputsmodel-q80.gguf Q80 Q6K High quality (12GB for 14B model) .llama.cppbuildbinllama-... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_14 | uf ggufoutputsmodel-q6k.gguf Q6K Q5KM Balanced qualitysize (10GB for 14B model) .llama.cppbuildbinllama-quantize --imatrix ggufoutputsmodel-imatrix.dat ggufoutputsmodel-merged-f16.gguf ggufoutputsmodel-q5km.gguf Q5KM Q4KM Recommended for most users (8GB for 14B model) .llama.cppbuildbinllama-quantize --imatrix... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_15 | l-q4km.gguf Q4KM 3. Quantization Options Reference Quantization Bits per Weight Quality Use Case Q80 8.0 Highest Production, quality critical Q6K 6.6 High Production, balanced Q5KM 5.5 Good Most users Q4KM 4.3 Acceptable Resource constrained Experimental Quantizations Warning These quantizations achieve extreme compr... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_16 | IQ2XXS Extreme compression (4GB for 14B model) .llama.cppbuildbinllama-quantize --imatrix ggufoutputsmodel-imatrix.dat ggufoutputsmodel-merged-f16.gguf ggufoutputsmodel-iq2xxs.gguf IQ2XXS TQ10 Ternary quantization (3.7GB for 14B model) .llama.cppbuildbinllama-quantize --imatrix ggufoutputsmodel-imatrix.dat ggufo... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_17 | ression (3.4GB for 14B model) .llama.cppbuildbinllama-quantize --imatrix ggufoutputsmodel-imatrix.dat ggufoutputsmodel-merged-f16.gguf ggufoutputsmodel-iq1s.gguf IQ1S Experimental Quantization Reference Quantization Bits per Weight Compression Warning Level IQ2XXS 2.06 85 smaller Moderate quality loss TQ10 1.69 87 ... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_18 | ng 1. Quick Functionality Test Test standard quantization .llama.cppbuildbinllama-cli -m ggufoutputsmodel-q4km.gguf -p Your test prompt here -n 100 --verbose Test experimental quantization .llama.cppbuildbinllama-cli -m ggufoutputsmodel-iq1s.gguf -p Your test prompt here -n 100 --verbose 2. Perplexity Evaluat... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_19 | ikitextwikitext-2-raw-v1.zip unzip wikitext-2-raw-v1.zip Test perplexity (lower is better) .llama.cppbuildbinllama-perplexity -m ggufoutputsmodel-q4km.gguf -f wikitext-2-rawwiki.test.raw -ngl 32 3. Size Verification Check all quantization sizes ls -lh ggufoutputs.gguf Expected output (14B model) 28G model-merged-f... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_2 | tem dependencies (UbuntuDebian) sudo apt update sudo apt install -y cmake build-essential git git-lfs 2. Clone and Build llama.cpp Clone llama.cpp git clone httpsgithub.comggml-orgllama.cpp cd llama.cpp Build with CUDA support (if available) cmake -B build -DGGMLCUDA ON cmake --build build --config Release OR build... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_20 | m.gguf 4.1G model-iq2xxs.gguf 3.7G model-tq10.gguf 3.4G model-iq1s.gguf Ollama Integration Ollama provides an easy way to run your quantized models locally with a simple API interface. Installation and Setup Install Ollama (LinuxmacOS) curl -fsSL httpsollama.aiinstall.sh sh Or download from httpsollama.ai for Window... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_21 | r Different Quantizations Q4KM (Recommended) - Modelfile cat Modelfile.q4 EOF FROM .ggufoutputsmodel-q4km.gguf System prompt for your specific use case SYSTEM You are a linguist and translator specializing in Ugandan languages, made by Sunbird AI. Chat template (adjust for your base model architecture) TEMPLATE ims... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_22 | imend Stop tokens PARAMETER stop imstart PARAMETER stop imend Generation parameters PARAMETER temperature 0.3 PARAMETER topp 0.95 PARAMETER topk 40 PARAMETER repeatpenalty 1.1 PARAMETER numctx 4096 PARAMETER numpredict 500 EOF Experimental IQ1S - Modelfile cat Modelfile.iq1s EOF FROM .ggufoutputsmodel-iq1s.gguf SY... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_23 | tra-compressed model - quality may be limited. Same template and parameters as above TEMPLATE imstartsystem .System imend imstartuser .Prompt imend imstartassistant .Response imend PARAMETER stop imstart PARAMETER stop imend PARAMETER temperature 0.3 PARAMETER topp 0.95 PARAMETER numctx 2048 Smaller context for ex... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_24 | ollama create sunflower-14bq4 -f Modelfile.q4 Import experimental IQ1S model ollama create sunflower-14biq1s -f Modelfile.iq1s Import other quantizations ollama create sunflower-14bq5 -f Modelfile.q5 ollama create sunflower-14bq6 -f Modelfile.q6 Verify models are imported ollama list Expected output NAME ID SIZE MO... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_25 | 3.4GB 1 minute ago Using Ollama Models Interactive Chat Start interactive session with Q4 model ollama run sunflower-14bq4 Example conversation Translate to Luganda Hello, how are you today? Give a dictionary definition of the Samia term ovulwaye in English bye (to exit) Start with experimental model ollama r... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_26 | lama run sunflower-14bq4 Translate to Luganda People in villages rarely accept new technologies. Test experimental model ollama run sunflower-14biq1s Translate to Luganda Good morning Dictionary definition ollama run sunflower-14bq4 Give a dictionary definition of the Samia term ovulwaye in English Ollama API Usage S... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_27 | endpoint curl httplocalhost11434apiversion Python API Client import requests import json def translatewithollama ( text , targetlang Luganda , model sunflower-14bq4 ) url httplocalhost11434apigenerate payload model model , prompt f Translate to targetlang text , stream False response requests . post ( ... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_28 | lt translatewithollama ( Hello, how are you? ) print ( result ) Test experimental model resultexperimental translatewithollama ( Good morning , model sunflower-14biq1s ) print ( Experimental model , resultexperimental ) curl API Examples Basic translation curl -X POST httplocalhost11434apigenerate -H Content-Type... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_29 | you today?, stream false Streaming response curl -X POST httplocalhost11434apigenerate -H Content-Type applicationjson -d model sunflower-14bq4, prompt Translate to Luganda People in villages rarely accept new technologies., stream true Model Management List all models ollama list Show model details ollama sho... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_3 | utputs cd .. Model Preparation 1. Download Models Download base model huggingface-cli download jqsunflower-14b-bs64-lr1e-4 --local-dir modelsbasemodel Download LoRA adapter huggingface-cli download jqqwen3-14b-sunflower-20250915 --local-dir modelsloramodel Download merged model (LoRA already merged) huggingface-cli ... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_30 | ew name ollama cp sunflower-14bq4 sunflower-translator Pullpush to Ollama registry (if you publish there) ollama push sunflower-14bq4 Performance Comparison Script Create testmodels.py import time import requests models sunflower-14bq4 , sunflower-14biq1s testprompt Translate to Luganda Hello, how are you today?... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_31 | ts . post ( httplocalhost11434apigenerate , json model modelname , prompt prompt , stream False ) endtime time . time () result response . json () return model modelname , response result response , time endtime - starttime , tokens len ( result response . split ()) Test all models for model in models ... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_32 | int ( f Response result response ) print ( f Time result time .2f s ) print ( f Tokens result tokens ) print ( - 50 ) Production Deployment Create production Modelfile with optimized settings cat Modelfile.production EOF FROM .ggufoutputsmodel-q4km.gguf SYSTEM You are a professional translator for Ug... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_33 | translations. TEMPLATE imstartsystem .System imend imstartuser .Prompt imend imstartassistant .Response imend PARAMETER stop imstart PARAMETER stop imend Production-optimized parameters PARAMETER temperature 0.1 Lower for consistency PARAMETER topp 0.9 Slightly more focused PARAMETER repeatpenalty 1.05 Minimal r... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_34 | Reasonable response length EOF Create production model ollama create sunflower-translatorproduction -f Modelfile.production Distribution 1. Hugging Face Upload Create upload script uploadmodels.py !usrbinenv python3 from huggingfacehub import HfApi , login , createrepo def uploadggufmodels () repoid your-orgyour-mod... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_35 | () api HfApi () createrepo ( repoid , existok True , repotype model ) Upload all files api . uploadfolder ( folderpath localfolder , repoid repoid , allowpatterns .gguf , .dat , commitmessage Add GGUF quantized models ) print ( f Upload complete httpshuggingface.co repoid ) if name main uploadggufmodels ... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_36 | py 2. Ollama Integration (Complete Guide) Installation and Setup Install Ollama (LinuxmacOS) curl -fsSL httpsollama.aiinstall.sh sh Or download from httpsollama.ai for Windows Start Ollama service (runs in background) ollama serve Creating Modelfiles for Different Quantizations Q4KM (Recommended) - Modelfile cat M... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_37 | ic use case SYSTEM You are a linguist and translator specializing in Ugandan languages, made by Sunbird AI. Chat template (adjust for your base model architecture) TEMPLATE imstartsystem .System imend imstartuser .Prompt imend imstartassistant .Response imend Stop tokens PARAMETER stop imstart PARAMETER stop imend... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_38 | topk 40 PARAMETER repeatpenalty 1.1 PARAMETER numctx 4096 PARAMETER numpredict 500 EOF Experimental IQ1S - Modelfile cat Modelfile.iq1s EOF FROM .ggufoutputsmodel-iq1s.gguf SYSTEM You are a translator for Ugandan languages. Note This is an experimental ultra-compressed model - quality may be limited. Same template a... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_39 | t imend imstartassistant .Response imend PARAMETER stop imstart PARAMETER stop imend PARAMETER temperature 0.3 PARAMETER topp 0.95 PARAMETER numctx 2048 Smaller context for experimental model EOF Importing Models to Ollama Import Q4KM model (recommended) ollama create sunflower-14bq4 -f Modelfile.q4 Import experime... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_4 | huggingface-cli download jqsunflower-14b-bs64-lr1e-4 --local-dir modelsbasemodel huggingface-cli download jqqwen3-14b-sunflower-20250915 --local-dir modelsloramodel Download merged model (LoRA already merged) huggingface-cli download Sunbirdqwen3-14b-sunflower-merged --local-dir modelsmergedmodel LoRA Merging Create ... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_40 | quantizations ollama create sunflower-14bq5 -f Modelfile.q5 ollama create sunflower-14bq6 -f Modelfile.q6 Verify models are imported ollama list Expected output NAME ID SIZE MODIFIED sunflower-14bq4 abc123def 8.4GB 2 minutes ago sunflower-14biq1s def456ghi 3.4GB 1 minute ago Using Ollama Models Interactive Chat Start... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_41 | ion Translate to Luganda Hello, how are you today? Give a dictionary definition of the Samia term ovulwaye in English bye (to exit) Start with experimental model ollama run sunflower-14biq1s Single Prompt Inference Quick translation with Q4 model ollama run sunflower-14bq4 Translate to Luganda People in village... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_42 | 14biq1s Translate to Luganda Good morning Dictionary definition ollama run sunflower-14bq4 Give a dictionary definition of the Samia term ovulwaye in English Ollama API Usage Start API Server Ollama automatically serves API on httplocalhost11434 Test API endpoint curl httplocalhost11434apiversion Python API Client i... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_43 | , model sunflower-14bq4 ) url httplocalhost11434apigenerate payload model model , prompt f Translate to targetlang text , stream False response requests . post ( url , json payload ) return response . json () response Test translation result translatewithollama ( Hello, how are you? ) print ( result ... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_44 | g , model sunflower-14biq1s ) print ( Experimental model , resultexperimental ) curl API Examples Basic translation curl -X POST httplocalhost11434apigenerate -H Content-Type applicationjson -d model sunflower-14bq4, prompt Translate to Luganda How are you today?, stream false Streaming response curl -X POST htt... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_45 | -14bq4, prompt Translate to Luganda People in villages rarely accept new technologies., stream true Model Management List all models ollama list Show model details ollama show sunflower-14bq4 Remove a model ollama rm sunflower-14biq1s Copy model with new name ollama cp sunflower-14bq4 sunflower-translator Pullpus... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_46 | ance Comparison Script Create testmodels.py import time import requests models sunflower-14bq4 , sunflower-14biq1s testprompt Translate to Luganda Hello, how are you today? def testmodel ( modelname , prompt ) starttime time . time () response requests . post ( httplocalhost11434apigenerate , json model model... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_47 | . json () return model modelname , response result response , time endtime - starttime , tokens len ( result response . split ()) Test all models for model in models result testmodel ( model , testprompt ) print ( f Model result model ) print ( f Response result response ) print ( f Time result ... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_48 | ooting Ollama Common Issues Model fails to load Check model file exists ls -la ggufoutputsmodel-q4km.gguf Verify Modelfile syntax ollama create sunflower-14bq4 -f Modelfile.q4 --verbose Out of memory Use smaller quantization ollama run sunflower-14biq1s Uses only 3.4GB Or reduce context size in Modelfile PARAMETER... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_49 | unflower-14bq4 Your test prompt ollama run sunflower-14biq1s Your test prompt Expected IQ1S may have degraded quality Ollama service not running Start service ollama serve Or check if running ps aux grep ollama Production Deployment Create production Modelfile with optimized settings cat Modelfile.production EOF... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_5 | 3-14b-sunflower-merged, skip this section and go directly to GGUF Conversion. Create mergelora.py only if using separate base model and LoRA adapter Create mergelora.py !usrbinenv python3 Merge LoRA adapter with base model from transformers import AutoModelForCausalLM , AutoTokenizer from peft import PeftModel impor... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_50 | gandan languages, made by Sunbird AI. Provide accurate, contextually appropriate translations. TEMPLATE imstartsystem .System imend imstartuser .Prompt imend imstartassistant .Response imend PARAMETER stop imstart PARAMETER stop imend Production-optimized parameters PARAMETER temperature 0.1 Lower for consistency ... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_51 | repetition penalty PARAMETER numctx 4096 Full context PARAMETER numpredict 200 Reasonable response length EOF Create production model ollama create sunflower-translatorproduction -f Modelfile.production Troubleshooting Common Issues 1. Out of Memory During Merge Use smaller precision or CPU offloading export PYTORC... |
https://salt.sunbird.ai/sunflower/quantization/ | 12_52 | ture compatibility python3 llama.cppconverthftogguf.py modelsmergedmodel --outfile test.gguf --dry-run 3. Quantization Too Slow Use fewer threads or disable imatrix for testing .llama.cppbuildbinllama-quantize ggufoutputsmodel-merged-f16.gguf ggufoutputsmodel-q4km.gguf Q4KM 4 Use 4 threads 4. Experimental Quantiz... |
End of preview. Expand in Data Studio
COCIS WEB INFO
Dataset Summary
This dataset contains text chucks scraped from its official website and corresponding websites. The dataset consists of JSON chunks, designed for high-performance streaming and parallel processing. Each chunk represents a discrete unit of data structured for machine learning tasks.
By sharding the data into chuck files, this repository supports the datasets library's streaming mode, allowing users to train models without downloading the entire dataset into RAM—a critical feature for resource-constrained environments or high-concurrency CI/CD pipelines.
Repository Structure
The data is organized into a chunks/ directory to maintain a clean root level:
.
├── README.md # This file
└── chunks/ # Directory containing JSON files
├── chunk_1.json
├── chunk_2.json
└── ...
Usage
You can load this dataset directly using the Hugging Face datasets library:
from datasets import load_dataset
# Standard loading
dataset = load_dataset("jimjunior/sunbird_salt_docs")
# Streaming mode (Recommended for many shards)
streamed_dataset = load_dataset("jimjunior/sunbird_salt_docs", streaming=True)
print(next(iter(streamed_dataset["train"])))
Maintenance and Contributions
This dataset was created as part of the Sunbird AI Internship 2026. Its actively mantained by Beingana Jim Junior.
- Downloads last month
- 10