Instructions to use primce/tiny-aya-hinglish-merged with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use primce/tiny-aya-hinglish-merged with Transformers:

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

pipe = pipeline("text-generation", model="primce/tiny-aya-hinglish-merged")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

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

tokenizer = AutoTokenizer.from_pretrained("primce/tiny-aya-hinglish-merged")
model = AutoModelForCausalLM.from_pretrained("primce/tiny-aya-hinglish-merged", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use primce/tiny-aya-hinglish-merged with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/primce/tiny-aya-hinglish-merged

SGLang

How to use primce/tiny-aya-hinglish-merged 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 "primce/tiny-aya-hinglish-merged" \
    --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": "primce/tiny-aya-hinglish-merged",
		"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 "primce/tiny-aya-hinglish-merged" \
        --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": "primce/tiny-aya-hinglish-merged",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use primce/tiny-aya-hinglish-merged with Docker Model Runner:
```
docker model run hf.co/primce/tiny-aya-hinglish-merged
```

tiny-aya-hinglish-merged

CohereLabs/tiny-aya-global with the shipped LoRA adapter primce/tiny-aya-hinglish-lora merged into the base weights. Same network, no PEFT:

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("primce/tiny-aya-hinglish-merged", dtype="float16")
tokenizer = AutoTokenizer.from_pretrained("primce/tiny-aya-hinglish-merged")

Everything that describes behaviour, data, safety numbers and limitations lives in the adapter card and the write-up; this repository exists so the same weights load without installing peft.

Adapter card: https://huggingface.co/primce/tiny-aya-hinglish-lora
Write-up: https://huggingface.co/spaces/primce/hinglish-track3-article
Run logs behind every number: https://huggingface.co/datasets/primce/lexsi-runs
Demo: https://huggingface.co/spaces/primce/hinglish-expense-agent

This is not a second model and not a second result. It is the published adapter with merge_and_unload() applied, and the run log merged_parity in primce/lexsi-runs checks that the merged weights reproduce the adapter's per-task verdicts on the same three tasks.

Licence: the adapter and these merged weights are CC BY-NC 4.0, inherited from CohereLabs/tiny-aya-global. The evaluation code in the GitHub repository is MIT.

Downloads last month: -

Safetensors

Model size

3B params

Tensor type

F16

Model tree for primce/tiny-aya-hinglish-merged

Base model

CohereLabs/tiny-aya-base

Finetuned

CohereLabs/tiny-aya-global

Adapter

(257)

this model

primce
/

tiny-aya-hinglish-merged

tiny-aya-hinglish-merged

Model tree for primce/tiny-aya-hinglish-merged

Dataset used to train primce/tiny-aya-hinglish-merged