Instructions to use Eclipse-Senpai/KeyLM-75M-Instruct with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Eclipse-Senpai/KeyLM-75M-Instruct with Transformers:

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

pipe = pipeline("text-generation", model="Eclipse-Senpai/KeyLM-75M-Instruct", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("Eclipse-Senpai/KeyLM-75M-Instruct", trust_remote_code=True, dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use Eclipse-Senpai/KeyLM-75M-Instruct with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Eclipse-Senpai/KeyLM-75M-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": "Eclipse-Senpai/KeyLM-75M-Instruct",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Eclipse-Senpai/KeyLM-75M-Instruct

SGLang

How to use Eclipse-Senpai/KeyLM-75M-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 "Eclipse-Senpai/KeyLM-75M-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": "Eclipse-Senpai/KeyLM-75M-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 "Eclipse-Senpai/KeyLM-75M-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": "Eclipse-Senpai/KeyLM-75M-Instruct",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Eclipse-Senpai/KeyLM-75M-Instruct with Docker Model Runner:
```
docker model run hf.co/Eclipse-Senpai/KeyLM-75M-Instruct
```

Eclipse-Senpai commited on 2 days ago

Commit

7f614be

verified ·

1 Parent(s): 6ec88ae

Replace variants table with SmolLM2-style base-vs-instruct benchmark table

Browse files

Files changed (1) hide show

README.md +12 -15

README.md CHANGED Viewed

@@ -58,13 +58,6 @@ KeyLM is a compact decoder-only transformer built on the standard small-model re
 | Precision | bfloat16 |
 | Training tokens | ~18B |
-### Model variants
-| Variant | Type | Chat template | IFEval (4-metric avg) | Use for |
-|---|---|---|---|---|
-| [KeyLM-75M](https://huggingface.co/Eclipse-Senpai/KeyLM-75M) | Base (pretrained) | No | — | Fine-tuning, text completion |
-| [KeyLM-75M-Instruct](https://huggingface.co/Eclipse-Senpai/KeyLM-75M-Instruct) | Instruction-tuned | Yes | 17.85 | Chat, instruction following |
 GGUF builds for `llama.cpp`, LM Studio, and Ollama are available at [KeyLM-75M-Instruct-GGUF](https://huggingface.co/Eclipse-Senpai/KeyLM-75M-Instruct-GGUF).
 ## How to Use
@@ -108,17 +101,21 @@ This is where KeyLM is competitive. All rows are evaluated with `lm_eval` (`ifev
 KeyLM beats the original SmolLM-135M-Instruct at roughly half the size and a fraction of the training data. SmolLM2-135M-Instruct, a far more heavily trained model, remains ahead.
-### Knowledge and reasoning
-On zero-shot multiple-choice benchmarks (`lm_eval`; accuracy, with length-normalized accuracy for ARC and HellaSwag) KeyLM is modest but above random on basic commonsense, and at chance on knowledge-heavy tasks. This is expected at 75M parameters and 18B tokens.
-| Model | MMLU | ARC (avg) | HellaSwag | PIQA | WinoGrande | OpenBookQA |
-|---|---|---|---|---|---|---|
-| KeyLM-75M (base) | 23.0 | 29.9 | 29.7 | 60.0 | 48.4 | 25.0 |
-| **KeyLM-75M-Instruct** | **24.0** | **30.8** | **31.0** | **61.3** | **48.3** | **25.0** |
-| Random baseline | 25.0 | 25.0 | 25.0 | 50.0 | 50.0 | 25.0 |
-Base and instruct track each other closely, so instruction tuning leaves knowledge and reasoning roughly unchanged. PIQA and ARC-easy land clearly above chance, while MMLU sits at the random baseline.
 ## Training

 | Precision | bfloat16 |
 | Training tokens | ~18B |
 GGUF builds for `llama.cpp`, LM Studio, and Ollama are available at [KeyLM-75M-Instruct-GGUF](https://huggingface.co/Eclipse-Senpai/KeyLM-75M-Instruct-GGUF).
 ## How to Use
 KeyLM beats the original SmolLM-135M-Instruct at roughly half the size and a fraction of the training data. SmolLM2-135M-Instruct, a far more heavily trained model, remains ahead.
+### Base vs Instruct
+The base and instruction-tuned checkpoints across all benchmarks. Commonsense and knowledge tasks are zero-shot via `lm_eval` (accuracy; ARC and HellaSwag length-normalized); IFEval is the 4-metric average. Bold marks the stronger version per row.
+| Benchmark | KeyLM-75M (base) | KeyLM-75M-Instruct | Random |
+|---|---|---|---|
+| IFEval (4-metric avg) | — | **17.85** | — |
+| MMLU | 23.0 | **24.0** | 25.0 |
+| ARC (avg) | 29.9 | **30.8** | 25.0 |
+| HellaSwag | 29.7 | **31.0** | 25.0 |
+| PIQA | 60.0 | **61.3** | 50.0 |
+| WinoGrande | **48.4** | 48.3 | 50.0 |
+| OpenBookQA | 25.0 | 25.0 | 25.0 |
+Instruction tuning leaves knowledge and reasoning roughly unchanged; its real effect is the instruction-following ability IFEval captures. Both versions sit modestly above random on basic commonsense and at chance on MMLU.
 ## Training