Instructions to use VikramPal/kambo-v1-sql-code-DynQuant-4bit with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use VikramPal/kambo-v1-sql-code-DynQuant-4bit with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="VikramPal/kambo-v1-sql-code-DynQuant-4bit", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("VikramPal/kambo-v1-sql-code-DynQuant-4bit", trust_remote_code=True, device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use VikramPal/kambo-v1-sql-code-DynQuant-4bit with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "VikramPal/kambo-v1-sql-code-DynQuant-4bit"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "VikramPal/kambo-v1-sql-code-DynQuant-4bit",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/VikramPal/kambo-v1-sql-code-DynQuant-4bit

SGLang

How to use VikramPal/kambo-v1-sql-code-DynQuant-4bit with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "VikramPal/kambo-v1-sql-code-DynQuant-4bit" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "VikramPal/kambo-v1-sql-code-DynQuant-4bit",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "VikramPal/kambo-v1-sql-code-DynQuant-4bit" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "VikramPal/kambo-v1-sql-code-DynQuant-4bit",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use VikramPal/kambo-v1-sql-code-DynQuant-4bit with Docker Model Runner:
```
docker model run hf.co/VikramPal/kambo-v1-sql-code-DynQuant-4bit
```

kambo-v1-sql-code-DynQuant-4bit / dynquant_allocation.json

VikramPal

Kambo-v1 SQL + Code, DynQuant 4-bit (4.25 bits per weight, packed)

6de8522 3 days ago

Raw History Blame Contribute Delete

7.66 kB

	{
	"schema": "dynquant_allocation_v1",
	"dynquant_core": "0.5.3",
	"model": "runs/ft/model",
	"stats": "runs/ft/stats",
	"allocator": "sensitivity",
	"group_size": 128,
	"map_key": "4.25",
	"map": {
	"target_label": "4.25 avg bits",
	"average_bits": 4.247889911339871,
	"nbytes": 898002432,
	"group_size": 128,
	"pricing": {
	"measured_modules": 133,
	"proxied_modules": 72,
	"measured_params": 331743232,
	"proxied_params": 1358954496,
	"proxied_share": 0.803783239010776,
	"scale": 1.1908850493541702e-17
	},
	"histogram": {
	"2": 9,
	"3": 27,
	"4": 119,
	"8": 50
	},
	"violations": [],
	"bits": {
	"model.embed_tokens": 8,
	"model.layers.0.conv.in_proj": 8,
	"model.layers.0.conv.out_proj": 8,
	"model.layers.0.moe.sw1": 4,
	"model.layers.0.moe.sw2": 8,
	"model.layers.0.moe.sw3": 8,
	"model.layers.0.moe.w1": 4,
	"model.layers.0.moe.w2": 2,
	"model.layers.0.moe.w3": 4,
	"model.layers.1.conv.in_proj": 4,
	"model.layers.1.conv.out_proj": 4,
	"model.layers.1.moe.sw1": 4,
	"model.layers.1.moe.sw2": 4,
	"model.layers.1.moe.sw3": 4,
	"model.layers.1.moe.w1": 4,
	"model.layers.1.moe.w2": 2,
	"model.layers.1.moe.w3": 2,
	"model.layers.10.conv.in_proj": 4,
	"model.layers.10.conv.out_proj": 4,
	"model.layers.10.moe.sw1": 4,
	"model.layers.10.moe.sw2": 8,
	"model.layers.10.moe.sw3": 8,
	"model.layers.10.moe.w1": 4,
	"model.layers.10.moe.w2": 4,
	"model.layers.10.moe.w3": 3,
	"model.layers.11.moe.sw1": 4,
	"model.layers.11.moe.sw2": 4,
	"model.layers.11.moe.sw3": 4,
	"model.layers.11.moe.w1": 4,
	"model.layers.11.moe.w2": 4,
	"model.layers.11.moe.w3": 4,
	"model.layers.11.self_attn.k_proj": 8,
	"model.layers.11.self_attn.o_proj": 8,
	"model.layers.11.self_attn.q_proj": 8,
	"model.layers.11.self_attn.v_proj": 8,
	"model.layers.12.conv.in_proj": 4,
	"model.layers.12.conv.out_proj": 4,
	"model.layers.12.moe.sw1": 4,
	"model.layers.12.moe.sw2": 4,
	"model.layers.12.moe.sw3": 4,
	"model.layers.12.moe.w1": 4,
	"model.layers.12.moe.w2": 3,
	"model.layers.12.moe.w3": 3,
	"model.layers.13.conv.in_proj": 4,
	"model.layers.13.conv.out_proj": 8,
	"model.layers.13.moe.sw1": 4,
	"model.layers.13.moe.sw2": 4,
	"model.layers.13.moe.sw3": 4,
	"model.layers.13.moe.w1": 4,
	"model.layers.13.moe.w2": 4,
	"model.layers.13.moe.w3": 4,
	"model.layers.14.conv.in_proj": 8,
	"model.layers.14.conv.out_proj": 8,
	"model.layers.14.moe.sw1": 4,
	"model.layers.14.moe.sw2": 8,
	"model.layers.14.moe.sw3": 8,
	"model.layers.14.moe.w1": 4,
	"model.layers.14.moe.w2": 3,
	"model.layers.14.moe.w3": 4,
	"model.layers.15.moe.sw1": 4,
	"model.layers.15.moe.sw2": 4,
	"model.layers.15.moe.sw3": 4,
	"model.layers.15.moe.w1": 4,
	"model.layers.15.moe.w2": 4,
	"model.layers.15.moe.w3": 3,
	"model.layers.15.self_attn.k_proj": 8,
	"model.layers.15.self_attn.o_proj": 8,
	"model.layers.15.self_attn.q_proj": 8,
	"model.layers.15.self_attn.v_proj": 8,
	"model.layers.16.conv.in_proj": 8,
	"model.layers.16.conv.out_proj": 8,
	"model.layers.16.moe.sw1": 4,
	"model.layers.16.moe.sw2": 4,
	"model.layers.16.moe.sw3": 4,
	"model.layers.16.moe.w1": 4,
	"model.layers.16.moe.w2": 3,
	"model.layers.16.moe.w3": 3,
	"model.layers.17.conv.in_proj": 4,
	"model.layers.17.conv.out_proj": 4,
	"model.layers.17.moe.sw1": 4,
	"model.layers.17.moe.sw2": 4,
	"model.layers.17.moe.sw3": 4,
	"model.layers.17.moe.w1": 4,
	"model.layers.17.moe.w2": 3,
	"model.layers.17.moe.w3": 3,
	"model.layers.18.conv.in_proj": 8,
	"model.layers.18.conv.out_proj": 8,
	"model.layers.18.moe.sw1": 4,
	"model.layers.18.moe.sw2": 8,
	"model.layers.18.moe.sw3": 8,
	"model.layers.18.moe.w1": 4,
	"model.layers.18.moe.w2": 3,
	"model.layers.18.moe.w3": 3,
	"model.layers.19.moe.sw1": 4,
	"model.layers.19.moe.sw2": 8,
	"model.layers.19.moe.sw3": 4,
	"model.layers.19.moe.w1": 4,
	"model.layers.19.moe.w2": 3,
	"model.layers.19.moe.w3": 3,
	"model.layers.19.self_attn.k_proj": 8,
	"model.layers.19.self_attn.o_proj": 8,
	"model.layers.19.self_attn.q_proj": 8,
	"model.layers.19.self_attn.v_proj": 8,
	"model.layers.2.conv.in_proj": 4,
	"model.layers.2.conv.out_proj": 4,
	"model.layers.2.moe.sw1": 4,
	"model.layers.2.moe.sw2": 4,
	"model.layers.2.moe.sw3": 4,
	"model.layers.2.moe.w1": 4,
	"model.layers.2.moe.w2": 2,
	"model.layers.2.moe.w3": 2,
	"model.layers.20.conv.in_proj": 4,
	"model.layers.20.conv.out_proj": 4,
	"model.layers.20.moe.sw1": 4,
	"model.layers.20.moe.sw2": 4,
	"model.layers.20.moe.sw3": 4,
	"model.layers.20.moe.w1": 4,
	"model.layers.20.moe.w2": 2,
	"model.layers.20.moe.w3": 2,
	"model.layers.21.conv.in_proj": 4,
	"model.layers.21.conv.out_proj": 4,
	"model.layers.21.moe.sw1": 4,
	"model.layers.21.moe.sw2": 4,
	"model.layers.21.moe.sw3": 4,
	"model.layers.21.moe.w1": 4,
	"model.layers.21.moe.w2": 3,
	"model.layers.21.moe.w3": 3,
	"model.layers.22.conv.in_proj": 4,
	"model.layers.22.conv.out_proj": 4,
	"model.layers.22.moe.sw1": 4,
	"model.layers.22.moe.sw2": 8,
	"model.layers.22.moe.sw3": 4,
	"model.layers.22.moe.w1": 4,
	"model.layers.22.moe.w2": 3,
	"model.layers.22.moe.w3": 3,
	"model.layers.23.moe.sw1": 8,
	"model.layers.23.moe.sw2": 8,
	"model.layers.23.moe.sw3": 8,
	"model.layers.23.moe.w1": 4,
	"model.layers.23.moe.w2": 3,
	"model.layers.23.moe.w3": 3,
	"model.layers.23.self_attn.k_proj": 8,
	"model.layers.23.self_attn.o_proj": 4,
	"model.layers.23.self_attn.q_proj": 8,
	"model.layers.23.self_attn.v_proj": 8,
	"model.layers.3.moe.sw1": 4,
	"model.layers.3.moe.sw2": 4,
	"model.layers.3.moe.sw3": 4,
	"model.layers.3.moe.w1": 4,
	"model.layers.3.moe.w2": 3,
	"model.layers.3.moe.w3": 2,
	"model.layers.3.self_attn.k_proj": 8,
	"model.layers.3.self_attn.o_proj": 8,
	"model.layers.3.self_attn.q_proj": 8,
	"model.layers.3.self_attn.v_proj": 8,
	"model.layers.4.conv.in_proj": 4,
	"model.layers.4.conv.out_proj": 4,
	"model.layers.4.moe.sw1": 4,
	"model.layers.4.moe.sw2": 4,
	"model.layers.4.moe.sw3": 4,
	"model.layers.4.moe.w1": 4,
	"model.layers.4.moe.w2": 3,
	"model.layers.4.moe.w3": 3,
	"model.layers.5.conv.in_proj": 4,
	"model.layers.5.conv.out_proj": 4,
	"model.layers.5.moe.sw1": 4,
	"model.layers.5.moe.sw2": 4,
	"model.layers.5.moe.sw3": 4,
	"model.layers.5.moe.w1": 4,
	"model.layers.5.moe.w2": 3,
	"model.layers.5.moe.w3": 3,
	"model.layers.6.conv.in_proj": 8,
	"model.layers.6.conv.out_proj": 8,
	"model.layers.6.moe.sw1": 4,
	"model.layers.6.moe.sw2": 8,
	"model.layers.6.moe.sw3": 4,
	"model.layers.6.moe.w1": 4,
	"model.layers.6.moe.w2": 2,
	"model.layers.6.moe.w3": 3,
	"model.layers.7.moe.sw1": 4,
	"model.layers.7.moe.sw2": 8,
	"model.layers.7.moe.sw3": 4,
	"model.layers.7.moe.w1": 4,
	"model.layers.7.moe.w2": 3,
	"model.layers.7.moe.w3": 3,
	"model.layers.7.self_attn.k_proj": 8,
	"model.layers.7.self_attn.o_proj": 8,
	"model.layers.7.self_attn.q_proj": 8,
	"model.layers.7.self_attn.v_proj": 8,
	"model.layers.8.conv.in_proj": 4,
	"model.layers.8.conv.out_proj": 4,
	"model.layers.8.moe.sw1": 4,
	"model.layers.8.moe.sw2": 4,
	"model.layers.8.moe.sw3": 4,
	"model.layers.8.moe.w1": 4,
	"model.layers.8.moe.w2": 4,
	"model.layers.8.moe.w3": 4,
	"model.layers.9.conv.in_proj": 4,
	"model.layers.9.conv.out_proj": 4,
	"model.layers.9.moe.sw1": 4,
	"model.layers.9.moe.sw2": 4,
	"model.layers.9.moe.sw3": 4,
	"model.layers.9.moe.w1": 4,
	"model.layers.9.moe.w2": 4,
	"model.layers.9.moe.w3": 4
	}
	}
	}