Instructions to use nayeshdaggula/sutradhar-3b-moe-instruct-v0.1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use nayeshdaggula/sutradhar-3b-moe-instruct-v0.1 with Transformers:

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

pipe = pipeline("text-generation", model="nayeshdaggula/sutradhar-3b-moe-instruct-v0.1")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("nayeshdaggula/sutradhar-3b-moe-instruct-v0.1")
model = AutoModelForCausalLM.from_pretrained("nayeshdaggula/sutradhar-3b-moe-instruct-v0.1")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use nayeshdaggula/sutradhar-3b-moe-instruct-v0.1 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "nayeshdaggula/sutradhar-3b-moe-instruct-v0.1"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "nayeshdaggula/sutradhar-3b-moe-instruct-v0.1",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/nayeshdaggula/sutradhar-3b-moe-instruct-v0.1

SGLang

How to use nayeshdaggula/sutradhar-3b-moe-instruct-v0.1 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 "nayeshdaggula/sutradhar-3b-moe-instruct-v0.1" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "nayeshdaggula/sutradhar-3b-moe-instruct-v0.1",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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 "nayeshdaggula/sutradhar-3b-moe-instruct-v0.1" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "nayeshdaggula/sutradhar-3b-moe-instruct-v0.1",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use nayeshdaggula/sutradhar-3b-moe-instruct-v0.1 with Docker Model Runner:
```
docker model run hf.co/nayeshdaggula/sutradhar-3b-moe-instruct-v0.1
```

Sutradhar 3B MoE Instruct v0.1

This is the first experimental instruction-tuned checkpoint of Sutradhar 3B MoE.

Base Model

nayeshdaggula/sutradhar-3b-moe-base-telugu-30min

Training Summary

Training type: instruction tuning / supervised fine-tuning
Language focus: Telugu + English
Dataset: small custom instruction dataset
Max steps: 300
Max length: 512
Runtime: around 19 minutes on A100 80GB
Final train loss: around 2.937

Focus Areas

Telugu assistant responses
English assistant responses
Indian business use cases
Coding explanations
Practical assistant behavior
Safe uncertainty handling

Important

This is an experimental checkpoint, not production-ready.

The current instruction dataset is very small, so this model is mainly useful for validating the instruction-tuning pipeline.

Downloads last month: 119

Safetensors

Model size

3B params

Tensor type

BF16

Model tree for nayeshdaggula/sutradhar-3b-moe-instruct-v0.1

Base model

nayeshdaggula/sutradhar-3b-moe-base

Finetuned

nayeshdaggula/sutradhar-3b-moe-base-telugu-30min