Instructions to use lowdown-labs/fela-autocomplete with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use lowdown-labs/fela-autocomplete with Transformers:

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

pipe = pipeline("text-generation", model="lowdown-labs/fela-autocomplete", trust_remote_code=True)

# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("lowdown-labs/fela-autocomplete", trust_remote_code=True, dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use lowdown-labs/fela-autocomplete with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/lowdown-labs/fela-autocomplete

SGLang

How to use lowdown-labs/fela-autocomplete 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 "lowdown-labs/fela-autocomplete" \
    --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": "lowdown-labs/fela-autocomplete",
		"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 "lowdown-labs/fela-autocomplete" \
        --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": "lowdown-labs/fela-autocomplete",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use lowdown-labs/fela-autocomplete with Docker Model Runner:
```
docker model run hf.co/lowdown-labs/fela-autocomplete
```

fela-autocomplete / config.json

itstheraj

initial commit

309d916 3 days ago

Raw

History Blame Contribute Delete

1.78 kB

	{
	"model_type": "fela",
	"library_name": "pytorch",
	"arch": "CPUGPT (FELA LM)",
	"edition": "FELA LLM 1.5 (code, 1.5B class)",
	"note": "FELA code language model for fill in the middle autocomplete. 1.79B parameters. 28 layers of FNO spectral token mixers plus Gated-DeltaNet recall plus Landmark global routing (SSSL pattern). Final fill in the middle SFT model.",
	"quant": "bf16",
	"vocab_size": 151936,
	"n_layer": 28,
	"n_embd": 1536,
	"n_head": 12,
	"ffn_hidden": 8960,
	"fno_modes": 512,
	"gla_chunk": 256,
	"layer_pattern": "SSSL",
	"gla_delta": true,
	"landmark_layer_every": 7,
	"landmark_chunk": 32,
	"landmark_max": 64,
	"attn_layer_every": 0,
	"seq_len": 2048,
	"n_params": 1788379536,
	"tokenizer": "Qwen2.5-Coder (tokenizer.json, vocab 151936, carries the FIM tokens <\|fim_prefix\|> / <\|fim_suffix\|> / <\|fim_middle\|>)",
	"fim_tokens": {
	"prefix": "<\|fim_prefix\|>",
	"suffix": "<\|fim_suffix\|>",
	"middle": "<\|fim_middle\|>"
	},
	"input_desc": "token ids from the Qwen2.5-Coder tokenizer. For fill in the middle: <\|fim_prefix\|> P <\|fim_suffix\|> S <\|fim_middle\|>, and the model completes the middle. For plain continuation: the prompt token ids.",
	"serving": "Runs on a plain CPU with no GPU via the FELA server (int8, constant memory streaming decode, no KV cache). The bf16 model.safetensors is the reference; model_int8.safetensors is the int8 serving export.",
	"int8_variant": "model_int8.safetensors",
	"int8_scheme": "weight only int8 per channel symmetric on linear layers; scales fp32; kept tensors bf16; keys prefixed int8 / scale / keep",
	"architectures": [
	"FelaForCausalLM"
	],
	"auto_map": {
	"AutoConfig": "configuration_fela.FelaConfig",
	"AutoModelForCausalLM": "modeling_fela.FelaForCausalLM"
	}
	}