Instructions to use fraunhofer-iis/elmod-2.7b-it with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use fraunhofer-iis/elmod-2.7b-it with Transformers:

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

pipe = pipeline("text-generation", model="fraunhofer-iis/elmod-2.7b-it")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("fraunhofer-iis/elmod-2.7b-it")
model = AutoModelForCausalLM.from_pretrained("fraunhofer-iis/elmod-2.7b-it", 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 fraunhofer-iis/elmod-2.7b-it with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "fraunhofer-iis/elmod-2.7b-it"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "fraunhofer-iis/elmod-2.7b-it",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/fraunhofer-iis/elmod-2.7b-it

SGLang

How to use fraunhofer-iis/elmod-2.7b-it 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 "fraunhofer-iis/elmod-2.7b-it" \
    --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": "fraunhofer-iis/elmod-2.7b-it",
		"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 "fraunhofer-iis/elmod-2.7b-it" \
        --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": "fraunhofer-iis/elmod-2.7b-it",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use fraunhofer-iis/elmod-2.7b-it with Docker Model Runner:
```
docker model run hf.co/fraunhofer-iis/elmod-2.7b-it
```

A problem with pre-tokenizer regex

by sszymczyk - opened 11 days ago

Discussion

sszymczyk

11 days ago

•

edited 11 days ago

In tokenizer.json we have:

    "pretokenizers": [
      {
        "type": "Split",
        "pattern": {
          "Regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?\\p{L}+|\\p{N}{1,3}| ?[^\\s\\p{L}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|s+(?!\\S)|\\s+"
        },
        "behavior": "Isolated",
        "invert": false
      },

Note the '|s' near the end, shouldn't it be '|\\s' like in the original llama pre-tokenizer regex below?

      {
        "type": "Split",
        "pattern": {
          "Regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?\\p{L}+|\\p{N}{1,3}| ?[^\\s\\p{L}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+"
        },
        "behavior": "Isolated",
        "invert": false
      },

schwirjo

Fraunhofer-Institut für Integrierte Schaltungen IIS org 9 days ago

Hey, thanks for noticing! I don't think the double escape will make any difference while for AutoTokenizer, but we can test it.

LuzianHahn

Fraunhofer-Institut für Integrierte Schaltungen IIS org 9 days ago

•

edited 8 days ago

Alright I have been checking this out. The Tokenizer causes a different behavior for this scenario:
"Hello World" (so anything with more than one whitespace char between two words.
current version: ['Hello', ' ', 'World']
version with "\s"-fix: ['Hello', ' ', ' World']

I think ordinary, the model should benefit from the fix, since it learns to connect the convention space-prefixed version of other tokens to situations, where there are more spaces involved. Right now it learned, that multi-spaces are connected to tokens, which are ordinary used as first tokens in sentences.

However, I did some perplexity check with a longer text using multi spaces instead of white spaces and the model was far off, using the fix. I therefore suggest to stay with this "broken" version, since we'd need to perform a retraining to really make use of it.

(Maybe there are some other situations affected by this bug, but I did not find them so far.)

Edit: remark, the formatting broke in this comment. Imagine multiple spaces in the above mentioned example

sszymczyk

8 days ago

I compared perplexity in llama.cpp on CPP code. Used ggml/src/ggml-opencl/ggml-opencl.cpp from llama.cpp since it's very long.

With the original "broken" regex I have: Final estimate: PPL = 3.7883 +/- 0.02268
With the "corrected" regex I have: Final estimate: PPL = 4.3889 +/- 0.02808

So I guess it's a good decision to keep the original form used for training the model.

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