Instructions to use AlekseyCalvin/LYRICAL_ru2en_SFT_RL_ver4_Spark4B_FP8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use AlekseyCalvin/LYRICAL_ru2en_SFT_RL_ver4_Spark4B_FP8 with Transformers:

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

pipe = pipeline("text-generation", model="AlekseyCalvin/LYRICAL_ru2en_SFT_RL_ver4_Spark4B_FP8", 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("AlekseyCalvin/LYRICAL_ru2en_SFT_RL_ver4_Spark4B_FP8", trust_remote_code=True, device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use AlekseyCalvin/LYRICAL_ru2en_SFT_RL_ver4_Spark4B_FP8 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "AlekseyCalvin/LYRICAL_ru2en_SFT_RL_ver4_Spark4B_FP8"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "AlekseyCalvin/LYRICAL_ru2en_SFT_RL_ver4_Spark4B_FP8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/AlekseyCalvin/LYRICAL_ru2en_SFT_RL_ver4_Spark4B_FP8

SGLang

How to use AlekseyCalvin/LYRICAL_ru2en_SFT_RL_ver4_Spark4B_FP8 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 "AlekseyCalvin/LYRICAL_ru2en_SFT_RL_ver4_Spark4B_FP8" \
    --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": "AlekseyCalvin/LYRICAL_ru2en_SFT_RL_ver4_Spark4B_FP8",
		"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 "AlekseyCalvin/LYRICAL_ru2en_SFT_RL_ver4_Spark4B_FP8" \
        --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": "AlekseyCalvin/LYRICAL_ru2en_SFT_RL_ver4_Spark4B_FP8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use AlekseyCalvin/LYRICAL_ru2en_SFT_RL_ver4_Spark4B_FP8 with Docker Model Runner:
```
docker model run hf.co/AlekseyCalvin/LYRICAL_ru2en_SFT_RL_ver4_Spark4B_FP8
```

LYRICAL_Spark4B_FP8 · `v4-sft`

LYRICAL Russian→English verse translator: XHToken/Spark-X2.5-4B with the LoRA chain below merged in (fp32 sum of deltas), then FP8-quantized (180 decoder projections as float8_e4m3fn with float32 scales per 128×128 block, 110 other tensors unchanged; 4.5 GB).

Merged adapters, in order:

AlekseyCalvin/LYRICAL_Spark4B_V4_SFT

v4 SFT: continued from GRPO-80 on the 360-poem roster with per-meter acceptance (ternary 39%), Calliope form spec in the prompt, no think.

Prompt style: spec (see lyrical.json).

from transformers import AutoModelForCausalLM, AutoTokenizer
repo, rev = "AlekseyCalvin/LYRICAL_Spark4B_FP8", "v4-sft"
tok = AutoTokenizer.from_pretrained(repo, revision=rev, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(repo, revision=rev, trust_remote_code=True, device_map="cuda")

FP8 kernels need a GPU with compute capability ≥ 8.9 (L4, 4090, H100/H200) and pip install kernels (transformers ≥ 5); elsewhere transformers dequantizes to bf16 on load.

Downloads last month: 90

Safetensors

Model size

4B params

Tensor type

BF16

F8_E4M3

Model tree for AlekseyCalvin/LYRICAL_ru2en_SFT_RL_ver4_Spark4B_FP8

Base model

XHToken/Spark-X2.5-4B-Base

Finetuned

XHToken/Spark-X2.5-4B

Quantized

(55)

this model

AlekseyCalvin
/

LYRICAL_ru2en_SFT_RL_ver4_Spark4B_FP8

LYRICAL_Spark4B_FP8 · `v4-sft`

Model tree for AlekseyCalvin/LYRICAL_ru2en_SFT_RL_ver4_Spark4B_FP8

Spaces using AlekseyCalvin/LYRICAL_ru2en_SFT_RL_ver4_Spark4B_FP8 2

LYRICAL_Spark4B_FP8 · v4-sft

Model tree for AlekseyCalvin/LYRICAL_ru2en_SFT_RL_ver4_Spark4B_FP8

Spaces using AlekseyCalvin/LYRICAL_ru2en_SFT_RL_ver4_Spark4B_FP8 2

LYRICAL_Spark4B_FP8 · `v4-sft`