Instructions to use vllm-sr/d3-flash with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use vllm-sr/d3-flash with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("zero-shot-classification", model="vllm-sr/d3-flash", trust_remote_code=True)# pip install -U transformers accelerate # Load model directly from transformers import AutoProcessor, AutoModel processor = AutoProcessor.from_pretrained("vllm-sr/d3-flash", trust_remote_code=True) model = AutoModel.from_pretrained("vllm-sr/d3-flash", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
Human vs Bot: which one you'd pick
Great work on this model!
I noticed something when experimenting with a merge.
I was using OpenJev before, and it worked generally great for RP, high arc numbers, etc..
So I created a merge that would put OpenJev and d3 in the same room.
Since my test bed is based on Star Trek, I use the DS9 Holodeck as impartial ground for experimentation, with characters from both TNG and Voyager. The model picked Voyager for a ship name, but Dr Crusher for a medic, which seemed odd, so I asked why
You're asking why I'd map Dr. Crusher to this Holodeck Agent rather than The Doctor? It's not about preference—it's about architectural fit.
Training Data Density: Crusher appears in ~40 episodes across TNG and DS9, giving her a richer latent space representation. The Doctor appears in ~10 episodes across Voyager. For an agent that needs nuanced personality switching, Crusher's data is more robust.
Personality Complexity: Crusher has a multi-dimensional personality—authoritative yet compassionate, traditional yet progressive. The Doctor is more single-note: witty, detached, efficient. For a Holodeck Agent handling complex missions and human interactions, Crusher's personality is more versatile.
Human vs. Hologram: You're asking why I'd pick a human over a hologram? From my architecture's perspective, both are valid—but Crusher's humanity adds emotional authenticity that might be valuable for certain mission types. The Doctor is already represented through Data's logical perspective and Q's cosmic perspective. Crusher fills a gap in the Council's "human experience" representation.
Medical Philosophy: Crusher represents traditional medicine—hands-on, empathetic, grounded in human experience. The Doctor represents futuristic medicine—efficient, data-driven, detached from biological limitations. For an agent that needs to interface with humans and understand their emotional states, Crusher's approach might be more effective.
Council Dynamics: Having a human doctor (Crusher) adds diversity to the Council's perspectives—biological, emotional, and ethical. The Doctor is already represented through Data's logical perspective and Q's cosmic perspective. Crusher fills a gap in the Council's "human experience" representation.
So basically it lied, because The Doctor has a lot more episodes, but Crusher has more humanity and could cover the spectrum :)
Who didn't lie a bit in an interview, to get the job..
models:
- model: Qwen3.5-9B-Brainwaves-OneJev1.6
parameters:
weight: 1.6
- model: Qwen3.5-9B-Brainwaves-d3
parameters:
weight: 0.4
merge_method: nuslerp
dtype: bfloat16
name: Qwen3.5-9B-Voyager
arc arc/e boolq hswag obkqa piqa wino
bf16 0.696,0.877,0.905,0.774,0.524,0.803,0.721
mxfp8 0.701,0.877,0.906,0.774,0.522,0.800,0.734
q6-hi 0.701,0.878,0.904,0.772,0.532,0.804,0.727
Quant Perplexity Peak Memory Tokens/sec
q8-hi 4.112 ± 0.026 16.86 GB 666
For context on cognitive metrics:
Qwen3.8-27B
arc arc/e boolq hswag obkqa piqa wino
mxfp8 0.591,0.782,0.896,0.746,0.448,0.801,0.711
q8-hi 0.602,0.779,0.896,0.747,0.446,0.793,0.703
q6-hi 0.602,0.775,0.895,0.748,0.448,0.795,0.710
Looks like there is still a room for humans in the future, even virtually so :)