Kaiyue/xiyansql-32b

Kaiyue/ xiyansql-32b:latest

173 Downloads Updated 6 months ago

This model uses IQ4_XS quantization and is from https://huggingface.co/mradermacher/XiYanSQL-QwenCoder-32B-2504-GGUF. Running the model requires 23.5 GB of GPU memory.

ollama run Kaiyue/xiyansql-32b

curl http://localhost:11434/api/chat \
  -d '{
    "model": "Kaiyue/xiyansql-32b",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='Kaiyue/xiyansql-32b',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'Kaiyue/xiyansql-32b',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Details

Updated 6 months ago

6 months ago

97ec39aadca4 · 18GB ·

model

archqwen2

parameters32.8B

quantizationIQ4_XS

18GB

template

{{ .Prompt }}\n--end-of-sql--

29B

params

{ "num_ctx": 32768, "num_gpu": 75, "stop": [ "--end-of-sql--" ] }

57B

Readme

The model can “translate” queries from natural language to SQL. Ex: prompt: "Only generate a single SQL query that returns each customer's total expense and number of orders, sorted by expense descending. Do not include extra examples or alternative interpretations." Generated SQL: SELECT c.name, SUM(o.total) AS total_expense, COUNT(o.id) AS number_of_orders FROM customers c JOIN orders o ON c.id = o.customer_id GROUP BY c.id ORDER BY total_expense DESC;

I would like to thank mradermacher and XGenerationLab for their contributions to the XiYanSQL-QwenCoder-32B-2504 models hosted on Hugging Face.