Instructions to use anicka/nla-phi4-av-arnative-grpo with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use anicka/nla-phi4-av-arnative-grpo with PEFT:

from peft import PeftModel
from transformers import AutoModelForCausalLM

base_model = AutoModelForCausalLM.from_pretrained("microsoft/phi-4")
model = PeftModel.from_pretrained(base_model, "anicka/nla-phi4-av-arnative-grpo")

Transformers

How to use anicka/nla-phi4-av-arnative-grpo with Transformers:

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

pipe = pipeline("text-generation", model="anicka/nla-phi4-av-arnative-grpo")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("anicka/nla-phi4-av-arnative-grpo", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use anicka/nla-phi4-av-arnative-grpo with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "anicka/nla-phi4-av-arnative-grpo"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "anicka/nla-phi4-av-arnative-grpo",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/anicka/nla-phi4-av-arnative-grpo

SGLang

How to use anicka/nla-phi4-av-arnative-grpo 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 "anicka/nla-phi4-av-arnative-grpo" \
    --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": "anicka/nla-phi4-av-arnative-grpo",
		"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 "anicka/nla-phi4-av-arnative-grpo" \
        --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": "anicka/nla-phi4-av-arnative-grpo",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use anicka/nla-phi4-av-arnative-grpo with Docker Model Runner:
```
docker model run hf.co/anicka/nla-phi4-av-arnative-grpo
```

anicka commited on Jun 28

Commit

eaf8950

verified ·

1 Parent(s): 1ded9e1

Upload README.md with huggingface_hub

Browse files

Files changed (1) hide show

README.md +5 -5

README.md CHANGED Viewed

@@ -39,10 +39,10 @@ The supervised adapter scored 0.474 mean-subtracted cosine on round-trip eval
 adapter scores **0.585** — a 23% improvement that closes 77% of the gap to the
 AR ceiling (0.619).
-On 4 of 7 evaluation layers, the GRPO adapter produces descriptions that
-reconstruct **better than the ground-truth descriptions** the SL adapter was
-imitating. The AR-native reward found output patterns that frontier-LLM
-descriptions never used.
 Qualitative difference: the SL adapter produced descriptions with correct style
 but vague content ("forward-looking sentiment," "narrative setup"). The GRPO
@@ -153,7 +153,7 @@ Double-holdout round-trip eval (49 texts unseen by both AV and AR):
 | L38 (96%) | 0.437 | 0.378 | 0.604 |
 | **Mean** | **0.585** | 0.474 | 0.619 |
-On L13, L22, L25: GRPO exceeds the GT ceiling — the adapter found description
 patterns that reconstruct better than the human-written targets.
 ## Companion models

 adapter scores **0.585** — a 23% improvement that closes 77% of the gap to the
 AR ceiling (0.619).
+On 2 of 9 evaluation layers (L13, L22), the GRPO adapter produces
+descriptions that reconstruct **better than the ground-truth descriptions**
+the SL adapter was imitating. The AR-native reward found output patterns
+that frontier-LLM descriptions never used.
 Qualitative difference: the SL adapter produced descriptions with correct style
 but vague content ("forward-looking sentiment," "narrative setup"). The GRPO
 | L38 (96%) | 0.437 | 0.378 | 0.604 |
 | **Mean** | **0.585** | 0.474 | 0.619 |
+On L13 and L22: GRPO exceeds the GT ceiling — the adapter found description
 patterns that reconstruct better than the human-written targets.
 ## Companion models