Upload README.md with huggingface_hub

README.md

@@ -1,30 +1,32 @@
----
-title: ASTERIZER
-emoji: 🚀
-colorFrom: gray
-colorTo: gray
-sdk: static
-pinned: false
----
-
-# ASTERIZER
-
-**Building AI from the model to the device** — self-hosted language models, a memory layer, application framework, and on-device AI, engineered in-house with 100% owned IP.
-
-**Models (trained from scratch)**
-- [LUNA-100M](https://huggingface.co/ASTERIZER/LUNA-100M) — 4.5B-token pretrain + RAG/MCP SFT
-- [LUNA-300M](https://huggingface.co/ASTERIZER/LUNA-300M) — 4.5B-token pretrain (2.7× scale-up)
-- LUNA-1B — in progress
-
-**Scaling benchmarks — LUNA-100M → LUNA-300M** *(latest base checkpoints, same 4.5B-token corpus, held-out ~100K-token eval on RTX 4060 Ti — higher is better)*
-
-| metric | LUNA-100M | LUNA-300M | gain |
-|---|---|---|---|
-| Next-token prediction accuracy | 32.5% | 32.7% | ↑0.8% |
-| Word-prediction accuracy (LAMBADA-style) | 16.3% | 17.1% | ↑4.4% |
-| Throughput (RTX 4060 Ti) | 39.2K tok/s | 14.5K tok/s | 2.7× params |
-
-**Tokenizer (multilingual BPE)**
-- [LUNA-1B-Tokenizer](https://huggingface.co/ASTERIZER/LUNA-Tokenizer-1B) — 128K / 64K / 8K vocab
-- **2.8× better compression** than GPT-NeoX, **67% fewer tokens** on Indic text (Kannada 7×, Telugu 6×, Tamil 5×)
-- Beats `cl100k_base` on Indic; competitive with Sarvam (34/55 head-to-head wins)
+---
+title: ASTERIZER
+emoji: 🚀
+colorFrom: gray
+colorTo: gray
+sdk: static
+pinned: false
+---
+# ASTERIZER
+**Building AI from the model to the device** — self-hosted language models, a memory layer, application framework, and on-device AI, engineered in-house with 100% owned IP.
+
+**Models (trained from scratch)**
+- [LUNA-100M](https://huggingface.co/ASTERIZER/LUNA-100M) — 4.5B-token pretrain + RAG/MCP SFT
+- [LUNA-300M](https://huggingface.co/ASTERIZER/LUNA-300M) — 4.5B-token pretrain (2.7x scale-up)
+- [LUNA-1B](https://huggingface.co/ASTERIZER/LUNA-1B-Instruct) — 32B-token continued pretraining + instruction tuning (11x scale-up)
+
+**Scaling benchmarks — LUNA-100M -> 300M -> 1B** *(latest base checkpoints, held-out multilingual eval on RTX 4060 Ti, higher is better)*
+
+| metric | LUNA-100M | LUNA-300M | LUNA-1B |
+|---|---|---|---|
+| Parameters | 109M | 303M | 1.21B |
+| Next-token accuracy (top-1) | 32.5% | 32.7% | **40.5%** |
+| Next-token accuracy (top-5) | 56.0% | 57.6% | **60.3%** |
+| Word-prediction (LAMBADA-style) | 17.9% | 19.0% | -* |
+| Throughput (RTX 4060 Ti) | 40.4K tok/s | 14.6K tok/s | 6.8K tok/s |
+
+*LUNA-1B uses a 128K multilingual tokenizer that merges word+punctuation into single tokens, so LAMBADA exact-match is not cross-tokenizer comparable; the next-token columns are the consistent comparison (and a larger vocab makes exact-match strictly harder).
+
+**Tokenizer (multilingual BPE)**
+- [LUNA-1B-Tokenizer](https://huggingface.co/ASTERIZER/LUNA-Tokenizer-1B) — 128K / 64K / 8K vocab
+- **2.8x better compression** than GPT-NeoX, **67% fewer tokens** on Indic text (Kannada 7x, Telugu 6x, Tamil 5x)
+- Beats `cl100k_base` on Indic; competitive with Sarvam (34/55 head-to-head wins)