Instructions to use alexchapin/portkit-coder-8b-grpo8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use alexchapin/portkit-coder-8b-grpo8 with Transformers:

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

pipe = pipeline("text-generation", model="alexchapin/portkit-coder-8b-grpo8")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("alexchapin/portkit-coder-8b-grpo8")
model = AutoModelForCausalLM.from_pretrained("alexchapin/portkit-coder-8b-grpo8", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use alexchapin/portkit-coder-8b-grpo8 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "alexchapin/portkit-coder-8b-grpo8"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "alexchapin/portkit-coder-8b-grpo8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/alexchapin/portkit-coder-8b-grpo8

SGLang

How to use alexchapin/portkit-coder-8b-grpo8 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 "alexchapin/portkit-coder-8b-grpo8" \
    --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": "alexchapin/portkit-coder-8b-grpo8",
		"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 "alexchapin/portkit-coder-8b-grpo8" \
        --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": "alexchapin/portkit-coder-8b-grpo8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use alexchapin/portkit-coder-8b-grpo8 with Docker Model Runner:
```
docker model run hf.co/alexchapin/portkit-coder-8b-grpo8
```

alexchapin commited on May 20

Commit

512d564

verified ·

1 Parent(s): 987e684

Update model card with training details and anti-hallucination fixes

Browse files

Files changed (1) hide show

README.md +41 -8

README.md CHANGED Viewed

@@ -12,18 +12,51 @@ tags:
 - bedrock-edition
 - minecraft-addon
 - mod-conversion
-- self-reflection
-- grpo8
 ---
 # alexchapin/portkit-coder-8b-grpo8
-This model was fine-tuned from [Qwen/Qwen3-8B](https://huggingface.co/Qwen/Qwen3-8B) using [Tinker](https://thinkingmachines.ai/tinker) and [tinker-cookbook](https://github.com/thinking-machines-lab/tinker-cookbook).
-## Model details
-- **Base model:** [Qwen/Qwen3-8B](https://huggingface.co/Qwen/Qwen3-8B)
-- **Format:** Merged model
 ## Usage
@@ -33,8 +66,8 @@ from transformers import AutoModelForCausalLM
 model = AutoModelForCausalLM.from_pretrained("alexchapin/portkit-coder-8b-grpo8")
 ```
-## Framework versions
 - tinker-cookbook: 0.4.1
 - transformers: 5.5.3
-- torch: 2.5.1+cu121

 - bedrock-edition
 - minecraft-addon
 - mod-conversion
+- anti-hallucination
 ---
 # alexchapin/portkit-coder-8b-grpo8
+Anti-hallucination focused GRPO fine-tuned model for Minecraft Java (Forge) to Bedrock Add-on conversion.
+## Model Details
+- **Base model:** Qwen/Qwen3-8B
+- **Training method:** GRPO with anti-hallucination focus
+- **Checkpoint:** GRPO8 final (50 steps, group_size=16-20)
+- **Learning rate:** 5e-7
+## Reward Components
+| Component | Weight | Description |
+|-----------|--------|-------------|
+| manifest | 0.10 | UUID v4, version array, module type validation |
+| real_api | 0.25 | Event subscription + API chain depth |
+| anti_hallucination | 0.25 | Hallucination penalties (shifted to 0.5-1.0) |
+| concise | 0.15 | Output length optimization |
+| code_bleu | 0.25 | Token overlap in code blocks |
+## Anti-Hallucination Fixes
+- **Issue #1582**: Hardened hallucination detection with comprehensive fake API patterns
+- **Issue #1583**: Real API fixes (event subscription, API chain depth)
+- **Issue #1584**: GRPO Stabilization (group_size 16, lr 5e-7, clipped surrogate)
+- **Issue #1586**: Manifest fixes (UUID v4, version array, module type)
+## Hard Hallucination Patterns (NEVER valid in Bedrock)
+- ServerPlayerAPI, ServerPlayer, PlayerAPI
+- WorldEvent, BlockEntityAPI, EntityPlayerAPI, WorldAPI
+- require('@minecraft/server')
+- registerMod(), getServer(), Server.getInstance()
+- event.level, server.getWorld
+## Real Bedrock APIs (ALWAYS valid)
+- world.afterEvents, world.beforeEvents
+- system.runInterval, system.runTimeout
+- player.sendMessage, player.getComponent
+- { world, system, player } imports from @minecraft/server
 ## Usage
 model = AutoModelForCausalLM.from_pretrained("alexchapin/portkit-coder-8b-grpo8")
 ```
+## Framework Versions
 - tinker-cookbook: 0.4.1
 - transformers: 5.5.3
+- torch: 2.11.0+rocm7.2