Instructions to use vtava/SmolLM2-135M-MemoryFusion-Sequential-R64 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use vtava/SmolLM2-135M-MemoryFusion-Sequential-R64 with Transformers:

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

pipe = pipeline("text-generation", model="vtava/SmolLM2-135M-MemoryFusion-Sequential-R64")

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("vtava/SmolLM2-135M-MemoryFusion-Sequential-R64", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use vtava/SmolLM2-135M-MemoryFusion-Sequential-R64 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "vtava/SmolLM2-135M-MemoryFusion-Sequential-R64"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "vtava/SmolLM2-135M-MemoryFusion-Sequential-R64",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/vtava/SmolLM2-135M-MemoryFusion-Sequential-R64

SGLang

How to use vtava/SmolLM2-135M-MemoryFusion-Sequential-R64 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 "vtava/SmolLM2-135M-MemoryFusion-Sequential-R64" \
    --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": "vtava/SmolLM2-135M-MemoryFusion-Sequential-R64",
		"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 "vtava/SmolLM2-135M-MemoryFusion-Sequential-R64" \
        --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": "vtava/SmolLM2-135M-MemoryFusion-Sequential-R64",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use vtava/SmolLM2-135M-MemoryFusion-Sequential-R64 with Docker Model Runner:
```
docker model run hf.co/vtava/SmolLM2-135M-MemoryFusion-Sequential-R64
```

SmolLM2-135M-MemoryFusion-Sequential-R64 / prompt_smoke_test.json

vtava

Publish accepted Memory Fusion prefix [0]

3d11298 verified 22 days ago

Raw History Blame

2.7 kB

	{
	"status": "prompt_generation_completed",
	"base_model": "HuggingFaceTB/SmolLM2-135M",
	"accepted_layers": [
	0
	],
	"memory_fusion_config": {
	"feature_dim": 32,
	"memory_rank": 64,
	"dilations": [
	1,
	2,
	4,
	8,
	16,
	32,
	64,
	128
	],
	"shifted_window": 8,
	"train_output_projection": true
	},
	"last_accepted_report": {
	"layer": 0,
	"accepted": true,
	"steps": 125,
	"nmse": 0.018665021285414696,
	"cosine": 0.9919254779815674,
	"probe_nll": 2.9270507097244263,
	"incremental_delta_nll": 0.01395869255065918,
	"cumulative_delta_nll": 0.01395869255065918,
	"round": 4
	},
	"tests": [
	{
	"prompt": "Austria is a country in Central Europe. The capital of Austria is",
	"baseline": "Austria is a country in Central Europe. The capital of Austria is Vienna.\n\nAustria is a country in Central Europe. The capital of Austria is Vienna.\n\nAustria",
	"memory_fusion": "Austria is a country in Central Europe. The capital of Austria is Vienna. The country is bordered by Germany to the north, Italy to the east, Slovenia to the south, and Hungary"
	},
	{
	"prompt": "Question: What is the capital of Austria?\nAnswer:",
	"baseline": "Question: What is the capital of Austria?\nAnswer: Vienna.\n\nQuestion: What is the capital of Austria?\nAnswer: Vienna.\n\nQuestion: What is",
	"memory_fusion": "Question: What is the capital of Austria?\nAnswer: Vienna.\n\nQuestion: What is the capital of Austria?\nAnswer: Vienna.\n\nQuestion: What is"
	},
	{
	"prompt": "Paris is the capital of France. Vienna is the capital of",
	"baseline": "Paris is the capital of France. Vienna is the capital of Austria. Paris is the capital of France.\n\nThe capital of France is Paris. The capital of Austria is Vienna",
	"memory_fusion": "Paris is the capital of France. Vienna is the capital of Austria.\n\nThe capital of the Czech Republic is Prague.\n\nThe capital of the United States is Washington,"
	},
	{
	"prompt": "The largest planet in the Solar System is",
	"baseline": "The largest planet in the Solar System is Jupiter. It is the largest planet in the Solar System and the largest planet in the Solar System. It is the largest",
	"memory_fusion": "The largest planet in the Solar System is Jupiter, which is 11 times the mass of Earth. The largest planet in the Solar System is Saturn, which"
	},
	{
	"prompt": "2 + 2 =",
	"baseline": "2 + 2 = 10\n\nThe sum of the first 10 natural numbers is 1 + 2 + 3",
	"memory_fusion": "2 + 2 = 3\n\nThe sum of the first 3 terms of the sequence is 1 + 2 + 3"
	}
	]
	}