15 yo. Building cool shit.
-
g1-nano-instruct
A 60M-parameter GPT pretrained and instruction-tuned on a single 8GB-RAM NVIDIA Jetson, offering a native 2048-token context and real multi-turn conversation support.
30 Pulls 1 Tag Updated 3 months ago
-
g0-nano-instruct
A 62M-parameter GPT pretrained and instruction-tuned on a single 8GB-RAM NVIDIA Jetson, offering a compact open-weights chat model for short, single-turn interactions.
17 Pulls 1 Tag Updated 3 months ago
-
g2-nano-instruct
A 60M-parameter GPT trained from scratch on a single 8GB Jetson (3B tokens), then instruction-tuned. SFT lowers mean_7 vs the base (40.16% vs 42.13%). Not an upgrade over g1-nano-instruct.
15 Pulls 1 Tag Updated 1 month ago
-
g1-nano-base
A 60M-parameter GPT trained from scratch on a single 8GB-RAM NVIDIA Jetson, offering a native 2048-token context for raw text completion. Base pretrained checkpoint, not instruction-tuned.
13 Pulls 1 Tag Updated 3 months ago
-
g2-nano-base
A 60M-parameter GPT trained from scratch on a single 8GB Jetson with 3B tokens (2× G1). Small gains. Base checkpoint: completes text. Not a G1 upgrade. For chat, use g1-nano-instruct.
9 Pulls 1 Tag Updated 1 month ago
-
g0-nano-base
A 62M-parameter GPT trained from scratch on a single 8GB-RAM NVIDIA Jetson, offering a compact open-weights base model for raw text completion. Not instruction-tuned.
9 Pulls 1 Tag Updated 3 months ago