Instructions to use geyuxu/york-milestone-project-self-traing-loop-model with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries
PEFT
How to use geyuxu/york-milestone-project-self-traing-loop-model with PEFT:
```
Task type is invalid.
```

How to use geyuxu/york-milestone-project-self-traing-loop-model with Transformers:

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

pipe = pipeline("text-generation", model="geyuxu/york-milestone-project-self-traing-loop-model")

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("geyuxu/york-milestone-project-self-traing-loop-model", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use geyuxu/york-milestone-project-self-traing-loop-model with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "geyuxu/york-milestone-project-self-traing-loop-model"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "geyuxu/york-milestone-project-self-traing-loop-model",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/geyuxu/york-milestone-project-self-traing-loop-model

SGLang

How to use geyuxu/york-milestone-project-self-traing-loop-model 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 "geyuxu/york-milestone-project-self-traing-loop-model" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "geyuxu/york-milestone-project-self-traing-loop-model",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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 "geyuxu/york-milestone-project-self-traing-loop-model" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "geyuxu/york-milestone-project-self-traing-loop-model",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use geyuxu/york-milestone-project-self-traing-loop-model with Docker Model Runner:
```
docker model run hf.co/geyuxu/york-milestone-project-self-traing-loop-model
```

York Milestone Project — Phase 1 adapters

Hugging Face 仓库： geyuxu/york-milestone-project-self-traing-loop-model

本仓库保存一期 v3 的最终 PEFT LoRA-DPO adapter，用于复现实验和后续阶段复用。它不包含基础模型权重、完整训练 checkpoint、合成数据、评测输出或原始 API 响应。

内容

路径	基础模型	用途	训练 / 验证偏好对
`phase1/qwen-main-dpo-v3/`	`Qwen/Qwen2.5-1.5B-Instruct`	一期主实验	99,986 / 11,110
`phase1/llama-dpo-v3/`	`meta-llama/Llama-3.2-1B-Instruct`	跨模型迁移实验	99,986 / 11,110
`phase1/ablations/empty_context/`	Qwen2.5-1.5B-Instruct	诊断消融	77,124 / 8,563
`phase1/ablations/helps_only/`	Qwen2.5-1.5B-Instruct	诊断消融	77,124 / 8,563
`phase1/ablations/matched_clean/`	Qwen2.5-1.5B-Instruct	诊断消融	77,124 / 8,563
`phase1/ablations/no_context/`	Qwen2.5-1.5B-Instruct	诊断消融	77,124 / 8,563
`phase1/ablations/random_labels/`	Qwen2.5-1.5B-Instruct	诊断消融	77,124 / 8,563
`phase1/ablations/shuffled_documents/`	Qwen2.5-1.5B-Instruct	诊断消融	77,124 / 8,563

每个目录包含：

adapter_model.safetensors 和 adapter_config.json；
与导出时一致的 tokenizer 和 chat template；
training_recipe.json：冻结的数据成员哈希及核心训练参数；
training_summary.json：训练步数、日志和最终摘要；
training_args.bin：Transformers 训练参数快照。

加载 adapter

以下示例加载 Qwen 主实验。其他 adapter 只需替换 subfolder 和对应的 base_id：

from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer

repo_id = "geyuxu/york-milestone-project-self-traing-loop-model"
subfolder = "phase1/qwen-main-dpo-v3"
base_id = "Qwen/Qwen2.5-1.5B-Instruct"

tokenizer = AutoTokenizer.from_pretrained(repo_id, subfolder=subfolder)
base_model = AutoModelForCausalLM.from_pretrained(
    base_id,
    torch_dtype="auto",
    device_map="auto",
)
model = PeftModel.from_pretrained(
    base_model,
    repo_id,
    subfolder=subfolder,
)

Llama adapter 需要先在 Hugging Face 接受 meta-llama/Llama-3.2-1B-Instruct 的访问条款，并使用有权限的账户登录。

完整性验证

从仓库根目录执行：

sha256sum -c MANIFEST.sha256

MANIFEST.sha256 覆盖 phase1/ 下发布的全部文件。Safetensors 通过 Git LFS 管理；基础模型权重不会被提交到本仓库。

许可证与限制

Qwen adapter 依赖 Apache-2.0 许可的 Qwen2.5 基础模型；
Llama adapter 的使用受 Llama 3.2 Community License 及其访问条款约束；
训练数据的来源、许可和限制以配套数据仓库的 dataset card 为准。

这些 adapter 是研究实验产物，未针对生产安全、事实准确性或具体行业用途做保证。

Downloads last month: -