Instructions to use vtava/SmolLM2-135M-MemoryFusion-r64 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use vtava/SmolLM2-135M-MemoryFusion-r64 with Transformers:

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

pipe = pipeline("text-generation", model="vtava/SmolLM2-135M-MemoryFusion-r64")

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("vtava/SmolLM2-135M-MemoryFusion-r64", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use vtava/SmolLM2-135M-MemoryFusion-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-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-r64",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/vtava/SmolLM2-135M-MemoryFusion-r64

SGLang

How to use vtava/SmolLM2-135M-MemoryFusion-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-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-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-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-r64",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use vtava/SmolLM2-135M-MemoryFusion-r64 with Docker Model Runner:
```
docker model run hf.co/vtava/SmolLM2-135M-MemoryFusion-r64
```

SmolLM2-135M-MemoryFusion-r64

Research artifact from TinyCeNN-LM. Architecture: smollm2-memory-fusion.

Architecture

Architecture/run type: smollm2-memory-fusion
Base model: HuggingFaceTB/SmolLM2-135M
Dataset: HuggingFaceFW/fineweb-edu (sample-10BT)
Source code: https://github.com/vtavakkoli/TinyCeNN-LM

Latest saved results

Metric	Value
`status`	trained
`stop_reason`	runtime_budget
`context_length`	128
`feature_dim`	32
`trainable`	19,760,130
`trainable_percent`	13.6917
`last_training_ce`	6.13222
`last_distillation_kl`	3.6898
`elapsed_minutes`	60.0118
`peak_vram_gib`	3.75348

The Hugging Face repository keeps timestamped run artifacts under runs/. This preserves training reports, configs and run metadata independently of the temporary Colab filesystem.

Saved experiment files

.hf_live_redacted/smollm2_memory_fusion_config.json
.hf_live_redacted/smollm2_memory_fusion_training_report.json
.hf_live_redacted/tokenizer_config.json
.hf_run_archive/r64-20260915T003900Z/.hf_live_redacted/smollm2_memory_fusion_config.json
.hf_run_archive/r64-20260915T003900Z/.hf_live_redacted/smollm2_memory_fusion_training_report.json
.hf_run_archive/r64-20260915T003900Z/.hf_live_redacted/tokenizer_config.json
.hf_run_archive/r64-20260915T003900Z/smollm2_memory_fusion_config.json
.hf_run_archive/r64-20260915T003900Z/smollm2_memory_fusion_training_report.json
.hf_run_archive/r64-20260915T003900Z/tokenizer_config.json
.hf_run_archive/r64-20260915T003910Z/.hf_live_redacted/smollm2_memory_fusion_config.json
.hf_run_archive/r64-20260915T003910Z/.hf_live_redacted/smollm2_memory_fusion_training_report.json
.hf_run_archive/r64-20260915T003910Z/.hf_live_redacted/tokenizer_config.json
.hf_run_archive/r64-20260915T003910Z/smollm2_memory_fusion_config.json
.hf_run_archive/r64-20260915T003910Z/smollm2_memory_fusion_training_report.json
.hf_run_archive/r64-20260915T003910Z/tokenizer_config.json
smollm2_memory_fusion_config.json
smollm2_memory_fusion_training_report.json
tokenizer_config.json

Reproducibility

Run the matching notebook from the TinyCeNN-LM repository. Colab notebooks use a Hugging Face write token from the HF_TOKEN Colab Secret; tokens should never be pasted into notebook source.

Limitations

This is a research checkpoint. Metrics saved here are the metrics produced by the corresponding training notebook/script; unless explicitly marked as held-out evaluation, they should not be treated as publication-grade benchmark results. Generation quality can differ substantially from the base model.

Citation

If you use this experimental checkpoint, cite the TinyCeNN-LM repository and the upstream base model.

Downloads last month: -; Downloads are not tracked for this model. How to track

Model tree for vtava/SmolLM2-135M-MemoryFusion-r64

Base model

HuggingFaceTB/SmolLM2-135M

Finetuned

(942)

this model

vtava
/

SmolLM2-135M-MemoryFusion-r64