Instructions to use dhanesh-hf/Jarvis-Titan-M3-UltraLong-Adapter with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use dhanesh-hf/Jarvis-Titan-M3-UltraLong-Adapter with Transformers:

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

pipe = pipeline("text-generation", model="dhanesh-hf/Jarvis-Titan-M3-UltraLong-Adapter")

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("dhanesh-hf/Jarvis-Titan-M3-UltraLong-Adapter", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use dhanesh-hf/Jarvis-Titan-M3-UltraLong-Adapter with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "dhanesh-hf/Jarvis-Titan-M3-UltraLong-Adapter"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "dhanesh-hf/Jarvis-Titan-M3-UltraLong-Adapter",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/dhanesh-hf/Jarvis-Titan-M3-UltraLong-Adapter

SGLang

How to use dhanesh-hf/Jarvis-Titan-M3-UltraLong-Adapter 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 "dhanesh-hf/Jarvis-Titan-M3-UltraLong-Adapter" \
    --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": "dhanesh-hf/Jarvis-Titan-M3-UltraLong-Adapter",
		"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 "dhanesh-hf/Jarvis-Titan-M3-UltraLong-Adapter" \
        --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": "dhanesh-hf/Jarvis-Titan-M3-UltraLong-Adapter",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use dhanesh-hf/Jarvis-Titan-M3-UltraLong-Adapter with Docker Model Runner:
```
docker model run hf.co/dhanesh-hf/Jarvis-Titan-M3-UltraLong-Adapter
```

J.A.R.V.I.S. TITAN 14.8B MoE — MILESTONE M3 ULTRA-LONG ADAPTER

Official Milestone M3 (Phase 3) weights for J.A.R.V.I.S. Titan 14.8B DeepSeekMoE + Tri-Brid Memory Architecture, distilled from full dense attention to Sliding Window Attention ($W=2048$) on Google Cloud TPU v5e-8.

Distillation & Training Specifications

Base Model: dhanesh-hf/Jarvis-Titan-V14-MoE-Merged (14.75B MoE, 100% Frozen)
Adapter Initialization: dhanesh-hf/Jarvis-Titan-M2-TriBrid-Adapter (Phase 2 Tri-Brid)
Dataset Source: dhanesh-hf/jarvis-v10-rft-dataset (100% Real Non-NIAH Peer-Reviewed Papers & Code)
Sliding Window Size: $W = 2048$ tokens (capping KV cache at ~115 MB)
Strategic Layers: [3, 7, 11, 15, 19, 23, 27] (7 Memory Bridges)
Tier 2 Salient Reservoir: $R=1024$ slots ($H_Q=28, H_{KV}=4$)
Tier 3 Titans Neural Memory: $d=512$, Google Pallas TPU VMEM SRAM kernel
Triple-Gated Adaptive Fusion: MAG-3 ($g_{\text{local}}, g_{\text{res}}, g_{\text{mem}}$)
Distillation Loss: $(1 - 0.5) \mathcal{L}{\text{CE}} + 0.5 T^2 \mathcal{D}{\text{KL}}$ ($T=2.0$)
Total Adapter Parameters: 90,044,458 (90.04M)
Total Tokens Trained: 20,012,495

Downloads last month: -; Downloads are not tracked for this model. How to track

Model tree for dhanesh-hf/Jarvis-Titan-M3-UltraLong-Adapter

Base model

dhanesh-hf/Jarvis-Titan-V14-MoE-Merged

Finetuned

(4)

this model