Instructions to use jiosephlee/assay-transfer-tool-soft-v6.5 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use jiosephlee/assay-transfer-tool-soft-v6.5 with Transformers:

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

pipe = pipeline("text-generation", model="jiosephlee/assay-transfer-tool-soft-v6.5")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("jiosephlee/assay-transfer-tool-soft-v6.5")
model = AutoModelForCausalLM.from_pretrained("jiosephlee/assay-transfer-tool-soft-v6.5", 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 jiosephlee/assay-transfer-tool-soft-v6.5 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "jiosephlee/assay-transfer-tool-soft-v6.5"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "jiosephlee/assay-transfer-tool-soft-v6.5",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/jiosephlee/assay-transfer-tool-soft-v6.5

SGLang

How to use jiosephlee/assay-transfer-tool-soft-v6.5 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 "jiosephlee/assay-transfer-tool-soft-v6.5" \
    --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": "jiosephlee/assay-transfer-tool-soft-v6.5",
		"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 "jiosephlee/assay-transfer-tool-soft-v6.5" \
        --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": "jiosephlee/assay-transfer-tool-soft-v6.5",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use jiosephlee/assay-transfer-tool-soft-v6.5 with Docker Model Runner:
```
docker model run hf.co/jiosephlee/assay-transfer-tool-soft-v6.5
```

assay-transfer-tool-soft-v6.5

Intern-S1-mini SFT for small-molecule assay-transfer ranking. Given a query molecule/endpoint and a retrieval candidate, the model emits an A/B decision whose P(A) (transfer-token vs non-transfer-token logit margin) scores how likely the candidate is to transfer. Intended to feed the top-k most-likely-to-transfer candidates to a downstream predictor.

Training

Base: jiosephlee/Intern-S1-mini-lm (rev fcb667c)
Data: jiosephlee/assay-transfer-raw-pair-v6.5-intern (train split, 2M pairs)
Objective: two-choice soft cross-entropy over the A/B decision token (transfer/nontransfer), label-smoothed target eps + (1-2·eps)·sigmoid(z/T), eps=0.1; format-token CE weight 0.1. 1 epoch, LR 2e-5.
Checkpoint selected by NDCG@5 on the frozen 20-candidate validation_ranking set (100 queries): step 800 — NDCG@5 = 0.710, NDCG@10 = 0.761, Spearman = 0.60 (vs untrained baseline NDCG@5 = 0.49, Spearman = 0.08).

Usage

Load with trust_remote_code=True (custom Intern-S1 tokenizer/model code included).

Downloads last month: 22

Safetensors

Model size

8B params

Tensor type

BF16

Model tree for jiosephlee/assay-transfer-tool-soft-v6.5

Base model

jiosephlee/Intern-S1-mini-lm

Finetuned

(9)

this model