Upload README.md with huggingface_hub

README.md

@@ -20,8 +20,8 @@ datasets:
 
 A small char-level CNN that extracts Indian informal number/currency
 shorthand -- Hinglish (romanised Hindi), Devanagari Hindi, Devanagari
-Marathi, and Indian-English amount phrases like `sava lakh`, `dedh crore`, `डेढ़ लाख`,
-`सवा करोड़`, `2.5L`, `20k`, `2-3 lakh` -- from free text, and turns each
+Marathi, Gujarati, and Indian-English amount phrases like `sava lakh`, `dedh crore`, `डेढ़ लाख`,
+`सवा करोड़`, `દોઢ લાખ`, `2.5L`, `20k`, `2-3 lakh` -- from free text, and turns each
 match into a clean numeric value via a deterministic arithmetic core (the
 model never predicts the value directly).
 
@@ -32,7 +32,7 @@ model never predicts the value directly).
 - live demo: https://huggingface.co/spaces/athrvk/gpu-sankhya-demo
 - gold evaluation sets: https://huggingface.co/datasets/athrvk/gpu-sankhya-gold
 
-This model card describes weights version `v0.5.0` (arch `v2`).
+This model card describes weights version `v0.6.0` (arch `v2`).
 
 ## Architecture
 
@@ -46,9 +46,9 @@ A dilated/residual char-CNN over per-character embeddings:
   3. kernel 3, dilation 2, residual yes
   4. kernel 3, dilation 4, residual yes
   5. kernel 3, dilation 8, residual yes
-- vocab: 119 characters (union of every shipped language pack)
+- vocab: 170 characters (union of every shipped language pack)
 - output classes: 120 (BIO span tag + semantic token class)
-- parameters: 39,659
+- parameters: 40,475
 
 A deterministic arithmetic core (not part of this model) then evaluates
 the decoded token sequence into a value: prefix semantics (sava = x1.25,
@@ -60,17 +60,17 @@ units, and range handling.
 
 | file | size |
 | --- | ---: |
-| `charset.json` | 966 bytes |
+| `charset.json` | 1,425 bytes |
 | `classes.json` | 1,571 bytes |
-| `gold_metrics_json_float32.json` | 6,439 bytes |
-| `gold_metrics_json_int8.json` | 6,439 bytes |
-| `gold_metrics_torch.json` | 6,439 bytes |
-| `kaggle_metrics.json` | 22,949 bytes |
-| `matrix.md` | 783 bytes |
-| `sankhya.onnx` | 161,280 bytes |
-| `sankhya.pt` | 168,310 bytes |
-| `sankhya.weights.int8.json` | 59,134 bytes |
-| `sankhya.weights.json` | 397,680 bytes |
+| `gold_metrics_json_float32.json` | 7,336 bytes |
+| `gold_metrics_json_int8.json` | 7,338 bytes |
+| `gold_metrics_torch.json` | 7,336 bytes |
+| `kaggle_metrics.json` | 26,135 bytes |
+| `matrix.md` | 872 bytes |
+| `sankhya.onnx` | 164,545 bytes |
+| `sankhya.pt` | 172,214 bytes |
+| `sankhya.weights.int8.json` | 60,720 bytes |
+| `sankhya.weights.json` | 405,469 bytes |
 
 - `sankhya.pt` -- torch checkpoint (vocab, classes, arch, channels, state_dict)
 - `sankhya.onnx` -- ONNX graph, for interop/inspection
@@ -91,27 +91,28 @@ against the hand-written gold sets (`gold_<lang>.jsonl` -- see the
 
 | gold set | examples | spans | value_acc | precision | recall | F1 |
 | --- | ---: | ---: | ---: | ---: | ---: | ---: |
-| gold.jsonl | 227 | 199 | 0.9548 | 0.9550 | 0.9598 | 0.9574 |
-| gold_deva.jsonl | 186 | 155 | 0.9613 | 0.9739 | 0.9613 | 0.9675 |
-| gold_mr.jsonl | 173 | 135 | 0.9778 | 0.9565 | 0.9778 | 0.9670 |
-| **combined** | 586 | 489 | **0.9632** | 0.9613 | 0.9652 | 0.9633 |
+| gold.jsonl | 227 | 199 | 0.9598 | 0.9550 | 0.9598 | 0.9574 |
+| gold_deva.jsonl | 186 | 155 | 0.9677 | 0.9868 | 0.9677 | 0.9772 |
+| gold_mr.jsonl | 173 | 135 | 0.9852 | 0.9852 | 0.9852 | 0.9852 |
+| gold_gu.jsonl | 162 | 118 | 0.9322 | 0.9658 | 0.9576 | 0.9617 |
+| **combined** | 748 | 607 | **0.9621** | 0.9719 | 0.9671 | 0.9694 |
 
-Negatives (zero-gold-span examples): 120, false
+Negatives (zero-gold-span examples): 170, false
 positives: 0 (0.00%).
 
 Per-category value accuracy (combined):
 
 | category | spans | value_acc |
 | --- | ---: | ---: |
-| digits | 123 | 0.9675 |
-| words | 366 | 0.9617 |
-| prefix | 104 | 0.9712 |
-| range | 42 | 0.9286 |
-| currency | 78 | 0.9872 |
+| digits | 145 | 0.9724 |
+| words | 462 | 0.9589 |
+| prefix | 104 | 0.9808 |
+| range | 49 | 0.9592 |
+| currency | 117 | 0.9658 |
 | multi_unit | 28 | 0.9643 |
-| symbol_unit | 42 | 0.9762 |
+| symbol_unit | 50 | 1.0000 |
 | mixed_script | 3 | 1.0000 |
-| long | 16 | 0.8125 |
+| long | 16 | 0.8750 |
 
 ## Usage
 
@@ -156,8 +157,9 @@ pipeline and `sankhya.eval_gold` for a ready-made evaluation CLI.
 ## Training
 
 Trained 20 epochs on 200,000 synthetic examples generated from the
-`hi_latn` (romanised Hindi), `hi_deva` (Devanagari Hindi), and `mr_deva`
-(Devanagari Marathi) grammar packs, mixed 0.40/0.33/0.27 with a 10%
+`hi_latn` (romanised Hindi), `hi_deva` (Devanagari Hindi), `mr_deva`
+(Devanagari Marathi), and `gu_gujr` (Gujarati) grammar packs, mixed
+0.32/0.26/0.21/0.21 with a 10%
 cross-pack share, plus out-of-vocab
 "unk noise" augmentation (emoji, CJK, Cyrillic, other symbols inserted as
 O-labelled context) so the `<unk>` embedding actually gets gradient
@@ -181,6 +183,6 @@ commands: `python/README.md` in the source repo.
 - Long multi-term/mixed-numeral constructs and multi-number range phrases
   ("तीस पैंतीस हज़ार", "three n half lakh") are the weakest category
   (`long`/`range` value_acc above).
-- Scoped to Indian languages: currently Hinglish, Devanagari Hindi, and
-  Devanagari Marathi only; other Indian languages are planned (see the
-  source repo's roadmap).
+- Scoped to Indian languages: currently Hinglish, Devanagari Hindi,
+  Devanagari Marathi, and Gujarati only; other Indian languages are
+  planned (see the source repo's roadmap).