Instructions to use ifx-pse-sys-ml/flame-27m-instruct with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ifx-pse-sys-ml/flame-27m-instruct with Transformers:

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

pipe = pipeline("text-generation", model="ifx-pse-sys-ml/flame-27m-instruct", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("ifx-pse-sys-ml/flame-27m-instruct", trust_remote_code=True, device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use ifx-pse-sys-ml/flame-27m-instruct with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ifx-pse-sys-ml/flame-27m-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": "ifx-pse-sys-ml/flame-27m-instruct",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/ifx-pse-sys-ml/flame-27m-instruct

SGLang

How to use ifx-pse-sys-ml/flame-27m-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 "ifx-pse-sys-ml/flame-27m-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": "ifx-pse-sys-ml/flame-27m-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 "ifx-pse-sys-ml/flame-27m-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": "ifx-pse-sys-ml/flame-27m-instruct",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use ifx-pse-sys-ml/flame-27m-instruct with Docker Model Runner:
```
docker model run hf.co/ifx-pse-sys-ml/flame-27m-instruct
```

lyonsn4 commited on Sep 11

Commit

df43de5

verified ·

1 Parent(s): cf4f7b7

Card: document generation parameters (repetition_penalty, no_repeat_ngram_size, temperature, top_p, max_new_tokens) + overrides

Browse files

Files changed (1) hide show

README.md +32 -0

README.md CHANGED Viewed

@@ -43,6 +43,38 @@ The decoder also accepts `inputs_embeds` (exactly one of `input_ids` / `inputs_e
 vision projector can splice visual tokens in — usable as a small VLM text backbone. A raw
 PyTorch checkpoint (`pytorch_model.pth`) is included alongside the safetensors weights.
 ## Benchmarks
 Accuracy (%) via [lm-evaluation-harness](https://github.com/EleutherAI/lm-evaluation-harness)

 vision projector can splice visual tokens in — usable as a small VLM text backbone. A raw
 PyTorch checkpoint (`pytorch_model.pth`) is included alongside the safetensors weights.
+## Generation parameters
+The repo ships sensible defaults in `generation_config.json`, so a bare
+`model.generate(ids)` already applies them — you don't need to pass anything. At **27M
+parameters** the model can loop across sentences, so the defaults enable repetition control:
+| Parameter | Default | What it does / when to change |
+|---|---|---|
+| `repetition_penalty` | `1.3` | Penalises tokens already generated. **The main lever against this model's multi-sentence looping.** Try `1.1`–`1.4`; `1.0` disables it (you'll see more repetition); >`1.5` starts to sound unnatural. |
+| `no_repeat_ngram_size` | `3` | Hard-blocks any 3-gram from repeating. Belt-and-braces with `repetition_penalty`. Set `0` to disable, `2` for stricter (can hurt fluency on short replies). |
+| `temperature` | `0.7` | Sampling randomness. Lower (`0.3`–`0.5`) = more focused/repetitive; higher (`0.9`–`1.1`) = more varied but more likely to go off-topic at this size. |
+| `top_p` | `0.9` | Nucleus sampling cutoff. Lower (`0.8`) tightens word choice; `1.0` allows the full distribution. |
+| `max_new_tokens` | `200` | Length cap. This model is strongest in **short** replies; long generations drift — keep it small (`64`–`200`) for best quality. |
+| `do_sample` | `True` | `False` = deterministic greedy decoding (reproducible; combine with `repetition_penalty` to avoid greedy loops). |
+Override any of them per call to experiment:
+```python
+out = model.generate(
+    ids,
+    max_new_tokens=120,
+    do_sample=True,
+    temperature=0.6,          # more focused
+    top_p=0.9,
+    repetition_penalty=1.2,   # tune 1.0 (off) – 1.4
+    no_repeat_ngram_size=3,   # 0 to disable
+)
+# Deterministic (greedy) — still repetition-guarded:
+out = model.generate(ids, max_new_tokens=120, do_sample=False, repetition_penalty=1.3)
+```
 ## Benchmarks
 Accuracy (%) via [lm-evaluation-harness](https://github.com/EleutherAI/lm-evaluation-harness)