Instructions to use BSC-LT/salamandra-7b-instruct with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use BSC-LT/salamandra-7b-instruct with Transformers:

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

pipe = pipeline("text-generation", model="BSC-LT/salamandra-7b-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-7b-instruct")
model = AutoModelForCausalLM.from_pretrained("BSC-LT/salamandra-7b-instruct")
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

vLLM

How to use BSC-LT/salamandra-7b-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-7b-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-7b-instruct",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/BSC-LT/salamandra-7b-instruct

SGLang

How to use BSC-LT/salamandra-7b-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-7b-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-7b-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-7b-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-7b-instruct",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use BSC-LT/salamandra-7b-instruct with Docker Model Runner:
```
docker model run hf.co/BSC-LT/salamandra-7b-instruct
```

Recommended ollama modelfile to install?

by arivero - opened Dec 16, 2024

Discussion

arivero

Dec 16, 2024

•

edited Dec 16, 2024

The model works fine with a standard template. I download the repository into the hf cache and then I create the Modelfile

FROM /home/kampal/.cache/huggingface/hub/models--BSC-LT--salamandra-7b-instruct/snapshots/e6c39f6092eb496f7feac065cf7a645bc792bbc6
TEMPLATE """
{{- range .Messages }}<|im_start|>{{ .Role }}
{{ .Content }}<|im_end|>
{{ end }}<|im_start|>assistant
""""

and it goes well:

(venv) kampal@kampal4090:~$ ollama create salamandra7b -f salamandra-7b-instruct-Modelfile 
transferring model data 100% 
using cached layer sha256:8d113310ba85d754a212e2f15a9de3c7d24fb3c0620809e95c2474ade5b1724b 
using existing layer sha256:8d113310ba85d754a212e2f15a9de3c7d24fb3c0620809e95c2474ade5b1724b 
creating new layer sha256:b7b5f247c4e3106fa129fa5f1c2a50ac79987ca766340fb66cf860d7d8ff61d3 
creating new layer sha256:6e6b543a72c4b5f811a818f1818069b98eb5ac6433e328867da7ba962d523884 
writing manifest 
success 
(venv) kampal@kampal4090:~$ ollama run salamandra7b:latest
>>> hla
Hello there!" - that is the English greeting "Hi". It is an informal way to start a conversation. If 
you are looking for something more formal, consider using "Good day" or "Good morning". 
Would you like me to assist you in any other matter?

>>> me puedes decir a que instituto de unizar corresponde el BIFI?
BIFI" es una sigla que pertenece al Instituto de Biocomputación y Física de Sistemas Complejos, 
ubicado en la Universidad de Zaragoza. ¿Desea saber más sobre este Instituto?

But I note you recommend a more complex prompt, including date and general roles, and on the other hand I am not sure if the default temperature sampling etc are adequate. Can someone recommend a more complete Modelfile along the lines of https://github.com/ollama/ollama/blob/main/docs/modelfile.md Note that ollama uses Go templates instead of Jinja

hdnh2006

Jan 23, 2025

I have quantized and uploaded the models to ollama, they can be run with ease with a single line of code:

ollama run hdnh2006/salamandra-7b-instruct

ollama run hdnh2006/salamandra-2b-instruct

I have quantized them from 2 to 8 bits and they can also be downloaded from HuggingFace:
https://huggingface.co/hdnh2006/BSC-LT-salamandra-2b-instruct-gguf
https://huggingface.co/hdnh2006/BSC-LT-salamandra-7b-instruct-gguf

Ollama links:
https://ollama.com/hdnh2006/salamandra-2b-instruct
https://ollama.com/hdnh2006/salamandra-7b-instruct

jgrivolla

Language Technologies Laboratory @ Barcelona Supercomputing Center org Jan 23, 2025

Thanks @hdnh2006 . Just to make sure, when did you download the weights? We updated them 2 days ago, so if you got them before then you may have an outdated version.

hdnh2006

Jan 24, 2025

Thanks @hdnh2006 . Just to make sure, when did you download the weights? We updated them 2 days ago, so if you got them before then you may have an outdated version.

I did it yesterday, so they should be updated.

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment