Instructions to use logicwahid/Kotha-S1-Instruct with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use logicwahid/Kotha-S1-Instruct with Transformers:

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

pipe = pipeline("text-generation", model="logicwahid/Kotha-S1-Instruct")

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("logicwahid/Kotha-S1-Instruct")
model = AutoModelForCausalLM.from_pretrained("logicwahid/Kotha-S1-Instruct", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use logicwahid/Kotha-S1-Instruct with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "logicwahid/Kotha-S1-Instruct"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "logicwahid/Kotha-S1-Instruct",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/logicwahid/Kotha-S1-Instruct

SGLang

How to use logicwahid/Kotha-S1-Instruct 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 "logicwahid/Kotha-S1-Instruct" \
    --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": "logicwahid/Kotha-S1-Instruct",
		"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 "logicwahid/Kotha-S1-Instruct" \
        --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": "logicwahid/Kotha-S1-Instruct",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use logicwahid/Kotha-S1-Instruct with Docker Model Runner:
```
docker model run hf.co/logicwahid/Kotha-S1-Instruct
```

Kotha-S1-Instruct (~100M)

Md. Wahid · Kotha Lab

Kotha-S1-Instruct is a 99.9M parameter bilingual (Bengali and English) instruction-tuned language model based on logicwahid/Kotha-S1. Built on a LLaMA-style architecture, it is designed to run on standard CPUs, edge devices, and low-resource environments.

Model Specifications

Architecture: LLaMA (RMSNorm, RoPE, SwiGLU)
Parameters: 99,943,040 (~100M)
Tokenizer: Custom bilingual tokenizer (32k vocab)
Languages: Bengali (বাংলা) & English
Prompt Format: [INST] prompt [/INST]
License: MIT-0

Usage

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

device = "cuda" if torch.cuda.is_available() else "cpu"
dtype = torch.float16 if torch.cuda.is_available() else torch.float32

repo_id = "logicwahid/Kotha-S1-Instruct"
tok = AutoTokenizer.from_pretrained(repo_id)
model = AutoModelForCausalLM.from_pretrained(repo_id, torch_dtype=dtype).to(device)

prompt = "[INST] বাংলাদেশের রাজধানী কী? [/INST] "
inputs = tok(prompt, return_tensors="pt").to(device)

out = model.generate(
    **inputs,
    max_new_tokens=80,
    do_sample=True,
    top_p=0.85,
    temperature=0.3,
    repetition_penalty=1.15
)
print(tok.decode(out[0], skip_special_tokens=True))

Downloads last month: -

Safetensors

Model size

99.9M params

Tensor type

F16

logicwahid
/

Kotha-S1-Instruct

Kotha-S1-Instruct (~100M)

Model Specifications

Usage

Datasets used to train logicwahid/Kotha-S1-Instruct