Update license: free personal use, paid commercial use (Pratham Kode)

LICENSE

@@ -0,0 +1,94 @@
+Smartwatch LM v0.2 — Personal / Commercial Dual License
+Copyright (c) 2026 Pratham Kode
+
+================================================================================
+1. DEFINITIONS
+================================================================================
+
+"Software" means the Smartwatch LM v0.2 model, weights, tokenizer, ONNX export,
+documentation, and all other files distributed with this repository.
+
+"You" means the individual or legal entity exercising permissions granted by
+this license.
+
+"Personal Use" means use of the Software by You for non-commercial purposes,
+including but not limited to:
+
+  - Personal, hobby, or educational projects
+  - Academic research and teaching (non-commercial)
+  - Evaluation, experimentation, and prototyping that is not part of a
+    commercial product or service
+  - Internal testing by an individual who is not acting on behalf of a business
+
+"Commercial Use" means any use of the Software that is not Personal Use,
+including but not limited to:
+
+  - Use in or as part of a product, service, or offering that is sold,
+    licensed, or otherwise monetized
+  - Use by a for-profit organization in the course of its business operations
+  - Use to provide services to third parties, whether or not a fee is charged
+  - Distribution of the Software (or derivatives) as part of a commercial
+    offering
+
+================================================================================
+2. PERSONAL USE LICENSE (FREE)
+================================================================================
+
+Subject to the terms of this license, permission is hereby granted, free of
+charge, to any person obtaining a copy of the Software to use, copy, modify,
+and distribute the Software for Personal Use only, provided that:
+
+  (a) this license text is included with any redistribution; and
+  (b) You do not use the Software for Commercial Use without a separate
+      commercial license as described in Section 3.
+
+================================================================================
+3. COMMERCIAL USE (PAID LICENSE REQUIRED)
+================================================================================
+
+Commercial Use of the Software is not permitted under the free Personal Use
+license. To use the Software for Commercial Use, You must obtain a paid
+commercial license from the copyright holder.
+
+To request commercial licensing terms and pricing, contact the repository
+maintainer through the Hugging Face model page or project repository.
+
+No commercial license is granted by implication. All Commercial Use without
+an executed commercial license agreement is prohibited.
+
+================================================================================
+4. RESTRICTIONS
+================================================================================
+
+Except as expressly permitted above, You may not:
+
+  - Use the Software for Commercial Use without a paid commercial license
+  - Remove or alter copyright or license notices
+  - Use the Software in any manner that violates applicable law
+
+================================================================================
+5. DISCLAIMER OF WARRANTY
+================================================================================
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE, AND NONINFRINGEMENT.
+
+================================================================================
+6. LIMITATION OF LIABILITY
+================================================================================
+
+IN NO EVENT SHALL THE COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES, OR
+OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT, OR OTHERWISE, ARISING
+FROM, OUT OF, OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+IN THE SOFTWARE.
+
+================================================================================
+7. TERMINATION
+================================================================================
+
+This license terminates automatically if You breach its terms. Upon
+termination, You must cease all use of the Software and destroy any copies in
+Your possession.
+
+Sections 5, 6, and 7 survive termination.

README.md

@@ -1,5 +1,7 @@
 ---
-license: mit
+license: other
+license_name: personal-use-free-commercial-paid
+license_link: LICENSE
 tags:
   - onnx
   - smartwatch
@@ -12,7 +14,7 @@ pipeline_tag: text-generation
 
 # Smartwatch LM v0.2
 
-Small GPT for wrist-wearable chat with intent tags like `<INTENT:GET_STEPS>`.
+Exported from [collab-run-2](../collab-run-2) training. Small GPT for wrist-wearable chat with intent tags like `<INTENT:GET_STEPS>`.
 
 ## Model details
 
@@ -29,7 +31,7 @@ Small GPT for wrist-wearable chat with intent tags like `<INTENT:GET_STEPS>`.
 
 | File | Purpose |
 |------|---------|
-| `smartwatch_lm_merged.onnx` | On-device inference (ONNX Runtime, opset 17, ~60 MB) |
+| `smartwatch_lm_merged.onnx` | On-device inference (ONNX Runtime, opset 17, ~0 MB) |
 | `checkpoint.pt` | PyTorch weights |
 | `tokenizer.json` | BPE tokenizer |
 | `config.json` | Architecture + ONNX I/O |
@@ -54,18 +56,10 @@ python chat.py
 - **Input:** `input_ids` int64 `[batch, seq]` (max seq = 256)
 - **Output:** `logits` float `[batch, seq, vocab_size]`
 
-## Documentation
+See [Export-0.1 docs](docs/) for gibberish cleanup, intents, and device integration.
 
-| Guide | Description |
-|-------|-------------|
-| [Avoiding gibberish](docs/avoiding-gibberish.md) | BPE cleanup, truncation rules, sample scripts |
-| [Intent reference](docs/intent-reference.md) | All intents and slot placeholders |
-| [Smartwatch integration](docs/smartwatch-integration.md) | End-to-end device wiring |
+## License
 
-## Benchmarks
+**Personal use is free.** You may use, modify, and share this model for non-commercial, personal, educational, and research purposes under the terms in [LICENSE](LICENSE).
 
-Quality evaluation on 39 golden prompts is in [`benchmark/`](benchmark/):
-
-- `benchmark/report.json` — full per-prompt results
-- `benchmark/charts/` — metric comparison charts
-- `benchmark/benchmark_prompts.json` — golden prompt set
+**Commercial use requires a paid license.** If you want to use this model in a product, service, or other commercial context, contact the repository maintainer to obtain commercial licensing terms.

docs/avoiding-gibberish.md

@@ -1,6 +1,6 @@
 # Avoiding Gibberish — Output Cleanup Guide
 
-Smartwatch LM v0.2 is a **~15.4M-parameter** domain model. It outputs structured replies with intent tags and slot placeholders. Raw tokenizer decode can still contain **BPE artifacts**, **encoding glitches**, or **run-on text**. This guide lists what to strip, how to truncate, and includes copy-paste scripts in this repo.
+Smartwatch LM v0.1 is a **12M-parameter** domain model. It outputs structured replies with intent tags and slot placeholders. Raw tokenizer decode can still contain **BPE artifacts**, **encoding glitches**, or **run-on text**. This guide lists what to strip, how to truncate, and includes copy-paste scripts in this repo.
 
 **Scripts in this repo:**
 
@@ -79,7 +79,7 @@ These rules are implemented in `extract_bot_reply()` in [`reply_utils.py`](../re
 
 ## Generation settings that reduce gibberish
 
-Use conservative sampling — this model is small (~15M params):
+Use conservative sampling — this model is small (12M params):
 
 | Parameter | Recommended | Effect |
 |-----------|-------------|--------|
@@ -174,7 +174,7 @@ build_prompt → encode → ONNX loop (top-k, temp 0.5) → decode new tokens on
 - Show raw `Ġ` or mojibake to the user
 - Put real sensor numbers back into conversation history
 - Skip truncation at `\nuser:` or first newline
-- Run temperature above 0.8 on this model
+- Run temperature above 0.8 on this 12M model
 
 **Do:**
 

docs/smartwatch-integration.md

@@ -1,8 +1,8 @@
 # Smartwatch Integration Guide
 
-How to run **Smartwatch LM v0.2** on a wrist device and wire it to sensors, timers, and apps.
+How to run **Smartwatch LM v0.1** on a wrist device and wire it to sensors, timers, and apps.
 
-The model is a **~15.4M-parameter** GPT exported as ONNX. It does not execute device actions itself — it emits **intent tags** and **slot placeholders** that your firmware parses and handles.
+The model is a **12M-parameter** GPT exported as ONNX. It does not execute device actions itself — it emits **intent tags** and **slot placeholders** that your firmware parses and handles.
 
 ---
 
@@ -10,7 +10,7 @@ The model is a **~15.4M-parameter** GPT exported as ONNX. It does not execute de
 
 | File | Size (approx.) | Purpose |
 |------|----------------|---------|
-| `smartwatch_lm_merged.onnx` | ~60 MB | ONNX Runtime inference |
+| `smartwatch_lm_merged.onnx` | ~52 MB | ONNX Runtime inference |
 | `tokenizer.json` | ~200 KB | Text ↔ token ids |
 | `tokenizer_config.json` | small | Tokenizer metadata |
 | `config.json` | small | Architecture and I/O names |
@@ -22,7 +22,7 @@ Optional on PC: `checkpoint.pt`, `chat.py`, `model.py` for PyTorch chat and fine
 **ONNX I/O:**
 
 - Input: `input_ids` — int64, shape `[batch, seq]`, max seq **256**
-- Output: `logits` — float, shape `[batch, seq, vocab_size]` (vocab **5533**)
+- Output: `logits` — float, shape `[batch, seq, vocab_size]` (vocab **3524**)
 
 Sample the **last position** logits autoregressively until EOS or max tokens.
 
@@ -61,7 +61,7 @@ See [Intent reference](./intent-reference.md) for all 35 intents.
 | Companion phone | ORT on phone, BLE to watch for display |
 | Prototype | `python onnx_sample.py` on desktop |
 
-Budget ~60 MB weights + activations. Quantization or phone-side inference helps on tight RAM.
+Budget ~52 MB weights + activations. Quantization or phone-side inference helps on tight RAM.
 
 ---