Instructions to use BSC-LT/salamandra-2b-instruct with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use BSC-LT/salamandra-2b-instruct with Transformers:

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

pipe = pipeline("text-generation", model="BSC-LT/salamandra-2b-instruct")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("BSC-LT/salamandra-2b-instruct")
model = AutoModelForCausalLM.from_pretrained("BSC-LT/salamandra-2b-instruct", 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=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use BSC-LT/salamandra-2b-instruct with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "BSC-LT/salamandra-2b-instruct"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "BSC-LT/salamandra-2b-instruct",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/BSC-LT/salamandra-2b-instruct

SGLang

How to use BSC-LT/salamandra-2b-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 "BSC-LT/salamandra-2b-instruct" \
    --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": "BSC-LT/salamandra-2b-instruct",
		"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 "BSC-LT/salamandra-2b-instruct" \
        --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": "BSC-LT/salamandra-2b-instruct",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use BSC-LT/salamandra-2b-instruct with Docker Model Runner:
```
docker model run hf.co/BSC-LT/salamandra-2b-instruct
```

Scandeval benchmarking error

by sarthu - opened Oct 3, 2024

Discussion

sarthu

Oct 3, 2024

I tried benchmarking the model on the dutch language benchmarks using scandeval library and I get the following error:
Benchmarking library: https://github.com/ScandEval/ScandEval

ferran-espuna

Oct 14, 2024

•

edited Oct 14, 2024

Hello,

I'm trying to work out exactly what's the problem so I need to reproduce the error. What command are you running exactly? I tried running

 scandeval -m path/to/salamandra-2b-instruct/

But I get the following error:

Traceback (most recent call last):
  File "/venv/bin/scandeval", line 8, in <module>
    sys.exit(benchmark())

  File "python3.11/site-packages/click/core.py", line 1157, in __call__
    return self.main(*args, **kwargs)

  File "python3.11/site-packages/click/core.py", line 1078, in main
    rv = self.invoke(ctx)

  File "python3.11/site-packages/click/core.py", line 1434, in invoke
    return ctx.invoke(self.callback, **ctx.params)

  File "python3.11/site-packages/click/core.py", line 783, in invoke
    return __callback(*args, **kwargs)

  File "venv/lib/python3.11/site-packages/scandeval/cli.py", line 342, in benchmark
    benchmarker(model=models)
  File "venv/lib/python3.11/site-packages/scandeval/benchmarker.py", line 781, in __call__
    return self.benchmark(*args, **kwargs)

  File "venv/lib/python3.11/site-packages/scandeval/benchmarker.py", line 597, in benchmark
    benchmark_output = self._benchmark_single(

  File "venv/lib/python3.11/site-packages/scandeval/benchmarker.py", line 730, in _benchmark_single
    dataset = dataset_factory.build_dataset(dataset_config)

  File "venv/lib/python3.11/site-packages/scandeval/dataset_factory.py", line 57, in build_dataset
    raise ValueError(
ValueError: Could not find a benchmark class for any of the following potential names: swerec, sentiment-classification, sequence-classification.

sarthu

Oct 14, 2024

•

edited Oct 14, 2024

The command I used was

scandeval --model <model-id>  --language nl

Also the error happens in the outlines library but the odd part is all others models are fine, only happens with this model

ferran-espuna

Oct 14, 2024

Hello again,

I'm trying this but now I get a similar error in the same place:

ValueError: Could not find a benchmark class for any of the following potential names: dutch-social, sentiment-classification, sequence-classification.

Do you know what could be the problem?

Meanwhile, I found this discussion on a similar issue:
https://github.com/dottxt-ai/outlines/issues/820

Can you try doing what the solution there suggests on your end and see what happens?