Instructions to use HopitAI/moda-pro-lite with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- OpenCLIP
How to use HopitAI/moda-pro-lite with OpenCLIP:
import open_clip model, preprocess_train, preprocess_val = open_clip.create_model_and_transforms('hf-hub:HopitAI/moda-pro-lite') tokenizer = open_clip.get_tokenizer('hf-hub:HopitAI/moda-pro-lite') - Notebooks
- Google Colab
- Kaggle
v1.1: verified 213M fashion-vocab build (gate: 100% bit-identity on 18K texts, graceful OOV fallback)
Browse files- README.md +2 -1
- open_clip_config.json +1 -1
- open_clip_model.safetensors +2 -2
- tokenizer.json +2 -2
README.md
CHANGED
|
@@ -35,7 +35,7 @@ model-index:
|
|
| 35 |
|
| 36 |
# MODA Pro Lite
|
| 37 |
|
| 38 |
-
**A
|
| 39 |
|
| 40 |
By [Hopit AI](https://hopit.ai). Weights are open (Apache-2.0); the training and data pipeline are proprietary.
|
| 41 |
|
|
@@ -84,6 +84,7 @@ Plain bi-encoder: one 768-d vector per image, one per query, cosine similarity.
|
|
| 84 |
## What's inside (summary)
|
| 85 |
|
| 86 |
- **Base:** SigLIP2-base-384 (`timm/ViT-B-16-SigLIP2-384`), full fine-tune of both towers.
|
|
|
|
| 87 |
- **Data:** 93K commercial product images (from [Marqo-GS-10M](https://huggingface.co/datasets/Marqo/marqo-GS-10M), Apache-2.0) re-annotated by a frontier VLM into structured attributes and four query registers, with graded relevance labels (99.9% coverage).
|
| 88 |
- **Recipe:** graded contrastive learning over multiple query registers, an anchor loss to the frozen base to prevent catastrophic transfer loss, and weight-space interpolation for the released checkpoint. The training pipeline and dataset are not released.
|
| 89 |
- **Distinctive property:** unlike typical fashion fine-tunes, Pro Lite does **not** collapse out of domain — the anchored recipe held generalist behavior while adding fashion skill (image-tower drift from base: 0.36%).
|
|
|
|
| 35 |
|
| 36 |
# MODA Pro Lite
|
| 37 |
|
| 38 |
+
**A 213M open-weights fashion encoder that beats [MODA](https://huggingface.co/HopitAI/moda-fashionsiglip-multiview-203m) — the strongest open fashion retrieval system — on catalog search, at full corpus, from one checkpoint with no serving machinery.**
|
| 39 |
|
| 40 |
By [Hopit AI](https://hopit.ai). Weights are open (Apache-2.0); the training and data pipeline are proprietary.
|
| 41 |
|
|
|
|
| 84 |
## What's inside (summary)
|
| 85 |
|
| 86 |
- **Base:** SigLIP2-base-384 (`timm/ViT-B-16-SigLIP2-384`), full fine-tune of both towers.
|
| 87 |
+
- **Size — 213,159,938 params:** the base's 256K-token multilingual vocabulary is trimmed to the 44,646 pieces fashion text actually uses (BPE merge-closure preserved). Verified: **100.00% bit-identical** outputs on 12,000 benchmark queries and 6,000 held-out registers vs the untrimmed 375M model — every benchmark number in this card transfers exactly. Out-of-vocabulary input (emoji, non-Latin scripts, rare typos) re-segments gracefully into kept pieces (cosine 0.82–0.97 to full model), never a silent failure.
|
| 88 |
- **Data:** 93K commercial product images (from [Marqo-GS-10M](https://huggingface.co/datasets/Marqo/marqo-GS-10M), Apache-2.0) re-annotated by a frontier VLM into structured attributes and four query registers, with graded relevance labels (99.9% coverage).
|
| 89 |
- **Recipe:** graded contrastive learning over multiple query registers, an anchor loss to the frozen base to prevent catastrophic transfer loss, and weight-space interpolation for the released checkpoint. The training pipeline and dataset are not released.
|
| 90 |
- **Distinctive property:** unlike typical fashion fine-tunes, Pro Lite does **not** collapse out of domain — the anchored recipe held generalist behavior while adding fashion skill (image-tower drift from base: 0.36%).
|
open_clip_config.json
CHANGED
|
@@ -12,7 +12,7 @@
|
|
| 12 |
},
|
| 13 |
"text_cfg": {
|
| 14 |
"context_length": 64,
|
| 15 |
-
"vocab_size":
|
| 16 |
"hf_tokenizer_name": "timm/ViT-B-16-SigLIP2-384",
|
| 17 |
"tokenizer_kwargs": {
|
| 18 |
"clean": "canonicalize"
|
|
|
|
| 12 |
},
|
| 13 |
"text_cfg": {
|
| 14 |
"context_length": 64,
|
| 15 |
+
"vocab_size": 44646,
|
| 16 |
"hf_tokenizer_name": "timm/ViT-B-16-SigLIP2-384",
|
| 17 |
"tokenizer_kwargs": {
|
| 18 |
"clean": "canonicalize"
|
open_clip_model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:536c60b2f8dcb57492cd20b990edcfeba1c43bf47563f93945632b26f2a7c5b9
|
| 3 |
+
size 852674280
|
tokenizer.json
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9c638d37b2bf0d43f97f3241e9d4ac2872485d485f79e981f89ef46764d583f2
|
| 3 |
+
size 3185164
|