Instructions to use ConflictAI/Conflict-Risk-9B-Full with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ConflictAI/Conflict-Risk-9B-Full with Transformers:

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

pipe = pipeline("text-generation", model="ConflictAI/Conflict-Risk-9B-Full")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("ConflictAI/Conflict-Risk-9B-Full")
model = AutoModelForCausalLM.from_pretrained("ConflictAI/Conflict-Risk-9B-Full", 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=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use ConflictAI/Conflict-Risk-9B-Full with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ConflictAI/Conflict-Risk-9B-Full"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ConflictAI/Conflict-Risk-9B-Full",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/ConflictAI/Conflict-Risk-9B-Full

SGLang

How to use ConflictAI/Conflict-Risk-9B-Full 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 "ConflictAI/Conflict-Risk-9B-Full" \
    --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": "ConflictAI/Conflict-Risk-9B-Full",
		"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 "ConflictAI/Conflict-Risk-9B-Full" \
        --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": "ConflictAI/Conflict-Risk-9B-Full",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use ConflictAI/Conflict-Risk-9B-Full with Docker Model Runner:
```
docker model run hf.co/ConflictAI/Conflict-Risk-9B-Full
```

Conflict-Risk-9B-Full / README.md

Firemedic15

Update README.md

7880cad verified 8 days ago

preview code

Raw

History Blame Contribute Delete

7.76 kB

	---
	library_name: transformers
	tags:
	- TravelRisk
	- Conflict
	- Finetune
	- ConflictAI
	license: mit
	datasets:
	- Firemedic15/travel-risk-conflict-combined
	language:
	- en
	metrics:
	- accuracy
	base_model:
	- Qwen/Qwen3.5-9B
	---

	# Model Card for Firemedic15/Conflict-Risk-9B-Full

	## Model Details

	### Model Description

	Conflict-Risk-9B-Full is a fine-tuned language model based on Qwen/Qwen3.5-9B. It was trained on approximately 50,000 records from a combined conflict and travel risk dataset to support risk-aware conversational analysis. The Qwen3.5-9B base model is a 9B-parameter model with native 262,144-token context support, published by Qwen on Hugging Face [web:27].

	The model is intended to help users explore travel risk, conflict context, regional instability, and related analytical questions in a conversational format.

	- Developed by: Firemedic15
	- Model type: Causal language model
	- Language(s): English
	- License: MIT
	- Base model: Qwen/Qwen3.5-9B
	- Dataset: Firemedic15/travel-risk-conflict-combined

	### Model Sources

	- Base model: [Qwen/Qwen3.5-9B](https://huggingface.co/Qwen/Qwen3.5-9B)
	- Dataset: [Firemedic15/travel-risk-conflict-combined](https://huggingface.co/datasets/Firemedic15/travel-risk-conflict-combined)
	- Model repository: [Firemedic15/Conflict-Risk-9B-Full](https://huggingface.co/Firemedic15/Conflict-Risk-9B-Full)

	## Uses

	This model is designed to support risk-aware dialogue related to travel and conflict. The training objective was to improve performance on prompts involving regional instability, conflict dynamics, and travel-related threat awareness.

	### Direct Use

	Potential direct uses include:

	- Conversational analysis of regional and global conflict conditions
	- Travel risk awareness and qualitative threat discussion
	- Summarization of conflict-related patterns and drivers
	- Research support for risk-focused exploratory tasks

	### Downstream Use

	Potential downstream uses include:

	- Risk intelligence assistants
	- Travel advisory research prototypes
	- Conflict monitoring and summarization tools
	- Retrieval-augmented applications that combine model output with current external data

	### Out-of-Scope Use

	This model should not be used for:

	- Real-time emergency response or evacuation decisions
	- Sole-source operational security decisions
	- Military targeting, surveillance, or harmful applications
	- Legal, medical, or life-safety advice
	- Any use case requiring guaranteed real-time factual accuracy without external verification

	## Bias, Risks, and Limitations

	This model may reflect biases present in the training data, including reporting bias, uneven regional coverage, and differing source quality across conflict and travel datasets.

	Because the model was trained to be risk aware, it may overemphasize threat, instability, or negative framing in some contexts. It may also hallucinate specific facts, events, timelines, or causal explanations when asked for highly detailed or current information.

	The model is not a substitute for verified reporting, official advisories, or expert human judgment.

	### Recommendations

	Users should treat this model as a supporting analytical tool rather than a source of truth.

	Recommended safeguards include:

	- Use current verified data sources for time-sensitive decisions
	- Apply human review before acting on outputs
	- Pair responses with retrieval or external validation where possible
	- Avoid use in high-stakes environments without additional controls

	## How to Get Started with the Model

	Use the example below to load the model with Transformers:

	```python
	from transformers import AutoTokenizer, AutoModelForCausalLM
	import torch

	model_id = "Firemedic15/Conflict-Risk-9B-Full"

	tokenizer = AutoTokenizer.from_pretrained(model_id)
	model = AutoModelForCausalLM.from_pretrained(
	model_id,
	torch_dtype=torch.bfloat16,
	device_map="auto"
	)

	prompt = "What are the main risk factors affecting travel in a conflict-affected region?"
	inputs = tokenizer(prompt, return_tensors="pt").to(model.device)

	outputs = model.generate(
	**inputs,
	max_new_tokens=200,
	temperature=0.7,
	do_sample=True
	)

	print(tokenizer.decode(outputs, skip_special_tokens=True))
	```

	## Training Details

	### Training Data

	The model was fine-tuned on approximately 50,000 records from the dataset `Firemedic15/travel-risk-conflict-combined`.

	The dataset combines conflict-oriented records and travel risk data to help the model produce more context-aware and risk-sensitive responses.

	### Training Procedure

	The model was fine-tuned from Qwen/Qwen3.5-9B and later merged into a standalone checkpoint for deployment.

	#### Preprocessing

	Preprocessing likely included:

	- Formatting examples into instruction or conversational pairs
	- Normalizing text fields
	- Combining conflict and travel-risk records into a unified training structure

	#### Training Hyperparameters

	- Training regime: bf16 mixed precision

	Additional hyperparameters such as learning rate, batch size, LoRA rank, number of epochs, and optimizer settings can be added here if available.

	#### Speeds, Sizes, Times

	- Training time: 26 hours
	- Training environment: Local machine

	## Evaluation

	### Testing Data, Factors, and Metrics

	#### Testing Data

	Evaluation data has not yet been fully documented. A recommended evaluation setup would include held-out prompts covering conflict analysis, regional instability, and travel-risk reasoning.

	#### Factors

	Relevant evaluation factors may include:

	- Geographic region
	- Conflict severity
	- Prompt wording and complexity
	- Time sensitivity
	- Degree of uncertainty in the prompt

	#### Metrics

	- Accuracy

	### Results

	Formal benchmark results are not yet documented in this card.

	The model is intended to improve risk-aware reasoning over the base model for conflict and travel-risk related prompts, but outputs should still be validated by human review and external sources where reliability matters.

	## Environmental Impact

	Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute).

	- Hardware Type: NVIDIA GeForce RTX 5070 Ti Blackwell
	- Hours used: 26
	- Cloud Provider: Local machine
	- Compute Region: Local
	- Carbon Emitted: Not estimated

	## Technical Specifications

	### Model Architecture and Objective

	This model is a fine-tuned causal language model based on Qwen/Qwen3.5-9B. According to the base model page, Qwen/Qwen3.5-9B is a 9B-parameter model with 32 layers and a native context length of 262,144 tokens [web:27]. Its objective in this fine-tuned form is to provide improved performance on prompts related to conflict awareness, travel risk, and risk-informed dialogue.

	### Compute Infrastructure

	The model was trained locally on consumer GPU hardware for approximately 26 hours.

	#### Hardware

	- NVIDIA GeForce RTX 5070 Ti Blackwell

	#### Software

	- Transformers
	- PyTorch
	- PEFT
	- Accelerate

	## Citation

	BibTeX:

	```bibtex
	@misc{firemedic15conflictrisk9bfull,
	title = {Conflict-Risk-9B-Full},
	author = {Firemedic15},
	year = {2026},
	howpublished = {\url{https://huggingface.co/Firemedic15/Conflict-Risk-9B-Full}},
	note = {Fine-tuned from Qwen/Qwen3.5-9B on combined conflict and travel-risk data}
	}
	```

	APA:

	Firemedic15. (2026). Conflict-Risk-9B-Full [Computer software]. Hugging Face. https://huggingface.co/Firemedic15/Conflict-Risk-9B-Full

	## More Information

	This model is best used as part of a broader analytical workflow that includes current data sources, human review, and domain-specific validation.

	## Model Card Authors

	Firemedic15

	## Model Card Contact

	Firemedic15 via Hugging Face Hub