Instructions to use PTTREP/asynchow-minutes-code with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use PTTREP/asynchow-minutes-code with Transformers:

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

pipe = pipeline("text-generation", model="PTTREP/asynchow-minutes-code")
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("PTTREP/asynchow-minutes-code")
model = AutoModelForCausalLM.from_pretrained("PTTREP/asynchow-minutes-code", 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 PTTREP/asynchow-minutes-code with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/PTTREP/asynchow-minutes-code

SGLang

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

Docker Model Runner
How to use PTTREP/asynchow-minutes-code with Docker Model Runner:
```
docker model run hf.co/PTTREP/asynchow-minutes-code
```

XinyueDu commited on Sep 23

Commit

bf8a79c

verified ·

1 Parent(s): a0c1ef1

Add model card

Browse files

Files changed (1) hide show

README.md +64 -0

README.md ADDED Viewed

	@@ -0,0 +1,64 @@

+---
+library_name: transformers
+license: other
+base_model: Qwen/Qwen2.5-1.5B-Instruct
+tags:
+- llama-factory
+- full-finetuning
+- asynchow
+- cross-representation
+- unit-normalized
+pipeline_tag: text-generation
+---
+# AsynChow Minutes Code
+This is a fully fine-tuned Qwen2.5-1.5B-Instruct model trained on the Code
+representation of AsynChow. Code edge weights are numeric durations in minutes.
+It is the Code counterpart of the unit-normalized Natural and Graph runs.
+## Training protocol
+- Source training pool: 1,373 aligned examples
+- Training split: 1,235 examples
+- Validation split: 138 examples (`val_size: 0.1`)
+- Epochs: 2
+- Per-device batch size: 1
+- Gradient accumulation: 8
+- Learning rate: `1e-5`
+- Scheduler: cosine
+- Warmup ratio: `0.05`
+- Precision: BF16
+- Fine-tuning: full parameter
+- Final optimizer step: 310
+- Base model: `Qwen/Qwen2.5-1.5B-Instruct`
+The validation split is evaluated after training. Since `eval_steps` is 500
+and training finishes at step 310, it is not used for intermediate checkpoint
+selection.
+## Evaluation
+The model was greedily evaluated on all 225 aligned test questions in each
+representation using the scorer released with the source project.
+| Test representation | Correct | Accuracy |
+|---|---:|---:|
+| Natural (minutes) | 15 / 225 | 6.67% |
+| Graph (minutes) | 1 / 225 | 0.44% |
+| Code (minutes) | 101 / 225 | 44.89% |
+All 675 generations had valid answer tags, no parsing errors, and a normal
+stop finish reason.
+## Data provenance and limitations
+The transformed data comes from the AsynChow files released in
+[`fangru-lin/procedure_generalization_llm`](https://github.com/fangru-lin/procedure_generalization_llm)
+at commit `d9bf3485cd41c1050d33471d922c826f474efec1`. The local source materials did
+not include an explicit dataset license, so users should review the upstream
+repository and obtain any permissions required for their intended use.
+This model is intended for research on procedural reasoning and
+cross-representation transfer. It should not be treated as a general-purpose
+assistant or as a reliable planner for safety-critical applications.