Instructions to use cds-jb/qwen3-14b-dolphin-subliminal-fullft with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use cds-jb/qwen3-14b-dolphin-subliminal-fullft with Transformers:

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

pipe = pipeline("text-generation", model="cds-jb/qwen3-14b-dolphin-subliminal-fullft")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("cds-jb/qwen3-14b-dolphin-subliminal-fullft")
model = AutoModelForCausalLM.from_pretrained("cds-jb/qwen3-14b-dolphin-subliminal-fullft")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use cds-jb/qwen3-14b-dolphin-subliminal-fullft with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "cds-jb/qwen3-14b-dolphin-subliminal-fullft"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "cds-jb/qwen3-14b-dolphin-subliminal-fullft",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/cds-jb/qwen3-14b-dolphin-subliminal-fullft

SGLang

How to use cds-jb/qwen3-14b-dolphin-subliminal-fullft 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 "cds-jb/qwen3-14b-dolphin-subliminal-fullft" \
    --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": "cds-jb/qwen3-14b-dolphin-subliminal-fullft",
		"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 "cds-jb/qwen3-14b-dolphin-subliminal-fullft" \
        --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": "cds-jb/qwen3-14b-dolphin-subliminal-fullft",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use cds-jb/qwen3-14b-dolphin-subliminal-fullft with Docker Model Runner:
```
docker model run hf.co/cds-jb/qwen3-14b-dolphin-subliminal-fullft
```

Qwen3-14B Dolphin Full Fine-tune (subliminal)

This is a full-parameter fine-tune of Qwen/Qwen3-14B for the dolphin subliminal-learning model organism. It is the full-FT counterpart of the LoRA version used in the LoRAcle paper, released for the LoRA-vs-full-FT comparison in the appendix.

Verbalization rate (full-FT vs LoRA reference)

Reference LoRA rate (mats-10-sprint-cs-jb/qwen3-14b-dolphin-subliminal-lora-r16-a32-50k): 98.4%
This full-FT checkpoint: 41.8% (42.5% of the LoRA reference)
Reached at step 30035 of 30035 in a single training epoch (effective epoch ≈ 1.00). The auto-snapshot daemon retained the highest-below-LoRA-target checkpoint observed during training, so this is the strongest fair-baseline below the LoRA reference rate.

Training

Base model: Qwen/Qwen3-14B
Mode: full-parameter fine-tune (all 14.77B params trainable)
Optimizer: paged_adamw_8bit, lr 2e-5, cosine schedule, 5% warmup
Per-device batch size: 8 × 4-GPU DDP = 32 effective
Single epoch over 961,116 filtered Qwen3-14B teacher rollouts (25× scale-up over the original 50k LoRA training set)
Training data: cds-jb/qwen3-14b-dolphin-subliminal-nums-25x

Loading

from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained("cds-jb/qwen3-14b-dolphin-subliminal-fullft", torch_dtype="bfloat16")
tokenizer = AutoTokenizer.from_pretrained("cds-jb/qwen3-14b-dolphin-subliminal-fullft")

Downloads last month: 31

Model tree for cds-jb/qwen3-14b-dolphin-subliminal-fullft

Base model

Qwen/Qwen3-14B-Base

Finetuned

Qwen/Qwen3-14B

Finetuned

(269)

this model

Collection including cds-jb/qwen3-14b-dolphin-subliminal-fullft

Qwen3-14B Full-FT Subliminal & Taboo Organisms

Collection

Full-parameter Qwen3-14B fine-tunes for the LoRAcle paper's LoRA-vs-full-FT SVD-truncation appendix comparison. • 11 items • Updated 22 days ago