Instructions to use CyanMonkey/Test-Subject-3 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use CyanMonkey/Test-Subject-3 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="CyanMonkey/Test-Subject-3")# pip install -U transformers accelerate # Load model directly from transformers import AutoModelForCausalLM model = AutoModelForCausalLM.from_pretrained("CyanMonkey/Test-Subject-3", device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use CyanMonkey/Test-Subject-3 with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "CyanMonkey/Test-Subject-3" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "CyanMonkey/Test-Subject-3", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/CyanMonkey/Test-Subject-3
- SGLang
How to use CyanMonkey/Test-Subject-3 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 "CyanMonkey/Test-Subject-3" \ --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": "CyanMonkey/Test-Subject-3", "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 "CyanMonkey/Test-Subject-3" \ --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": "CyanMonkey/Test-Subject-3", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }' - Docker Model Runner
How to use CyanMonkey/Test-Subject-3 with Docker Model Runner:
docker model run hf.co/CyanMonkey/Test-Subject-3
File size: 4,339 Bytes
1c016ea 9b4717b 1c016ea 9b4717b a10c7e3 1c016ea 9b4717b 1c016ea 9b4717b 1c016ea 1ac955a 1c016ea 1ac955a 1c016ea 1ac955a 1c016ea 9b4717b 1c016ea | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 | ---
library_name: transformers
license: apache-2.0
tags:
- oleander
- gdn2
- gated-deltanet
- gqa
- muon
- marimo
- molab
---
# Oleander-150M
**Training checkpoint** (`round 14`) β step **7,000 / 28,500** Β· **24.56%** of the planned run.
```text
steps ββββββββββββββββββββββββββββ 24.56% 7,000 / 28,500
tokens ββββββββββββββββββββββββββββ 19.30% 2.88B / 14.94B compute
phase tst (TST bag training ends at step 9,000)
```
Trained with the Oleander marimo trainer (molab): hybrid **GDN2** (gated delta-rule, gated-delta-net-2) and **gated GQA** layers with **block attention residuals**, MuonEq-R + AdamW, WSD schedule.
## Training progress
| | |
|---|---|
| Step | **7,000** / 28,500 |
| % complete (steps) | **24.56%** |
| Compute tokens seen | **2.88B** / 14.94B (19.30%) |
| Raw tokens seen (TST bags unpacked) | 17.30B / 38.54B |
| Cluster tokens (all 1 worker(s)) | 2.88B / 14.94B compute |
| Phase | `tst` |
| Val loss / ppl | **6.3489** / **571.85** |
| Train loss (latest) | 5.2985 |
| Throughput | 76,860 compute tok/s Β· 0.139 steps/s |
| Session elapsed | 0d 03h 03m Β· ETA 1d 18h 57m |
| Checkpoint written | 2026-11-15T09:39:11Z |
## How to load
The repo is self-contained: `config.json` carries an `auto_map` that points at `configuration_oleander.py` and `modeling_oleander.py` (shipped here), so remote code is all you need.
```python
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "CyanMonkey/Test-Subject-3"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(repo, trust_remote_code=True)
prompt = "Once upon a time"
out = model.generate(**tok(prompt, return_tensors="pt"), max_new_tokens=64,
do_sample=True, temperature=0.8, top_k=50)
print(tok.decode(out[0], skip_special_tokens=True))
```
The GDN2 mixers use `flash-linear-attention` kernels when they are installed and you run in bf16 on a GPU (`pip install --no-deps fla-core flash-linear-attention`); otherwise a pure-PyTorch recurrent reference is used β the same math, but slower for long prompts. There is no KV cache (`use_cache=False`).
## Architecture
```text
Oleander-150M (28L = 7 x [GDN2 Β· GDN2 Β· GDN2 Β· GQA] Β· d=576 Β· SwiGLU(1600) Β· tied 32,770 vocab)
```
| | |
|---|---|
| Parameters | 149,976,139 (149.98 M) |
| Layers | 28 = 7 x [gdn2, gdn2, gdn2, gqa] |
| Hidden size / FFN | 576 / 1600 (SwiGLU) |
| Vocab / context | 32,770 (tied embeddings) / 1,024 |
| Features | GDN2 (gated delta net 2) x21: 9h x 64, conv4, gated GQA x7: 9 Q / 3 KV heads, head_dim 64, QK-Norm (offset form), partial RoPE (rotary_dim 32, theta 100000), value residual (v_mix), SwiGLU, RMSNorm, AttnRes (block attention residuals), tied embeddings, TST (token superposition training, bag 6) |
## Training setup
| | |
|---|---|
| Optimizer | muoneq_r (Muon) + AdamW β `muon_adamw` |
| LR schedule | WSD Β· warmup 100 Β· stable until 82.5% Β· 1-sqrt cooldown |
| Batch | 524,288 compute tokens / step (32 x 16 x 1024) |
| TST | True Β· bag 6 Β· first 30% of steps |
| Data | `CyanMonkey/NextGen-1-32-1024` (train), 14,753,933 rows x 1024 tokens |
| Precision | bf16 autocast Β· weights exported as float32 |
## Files
| file | what |
|---|---|
| `model.safetensors` | weights, trainer parameter names |
| `config.json` | `auto_map` -> `AutoConfig` / `AutoModelForCausalLM` remote-code entry points |
| `configuration_oleander.py` | `OleanderConfig` |
| `modeling_oleander.py` | `OleanderForCausalLM` (same math as the trainer) |
| `tokenizer.json`, `tokenizer_config.json`, `special_tokens_map.json` | tokenizer |
| `training_meta.json` | detailed run metadata (this checkpoint) |
| `progress.json` | flat step / tokens / % summary |
| `generation_config.json`, `train_config.json` | sampling defaults, full trainer config |
## Notes
- This is a **work-in-progress** checkpoint at 24.56% of the planned run, not a finished model: expect repetition and incoherence, especially with few-shot prompts.
- The tokenizer comes from the base model the run started from; the vocabulary is a plain BPE model and is not extended during training.
- `training_meta.json` is regenerated on every push and is the authoritative record of how far training got.
|