Upload folder using huggingface_hub

.gitattributes

@@ -33,3 +33,9 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+qaihub/LibreYOLOXs_fp32.dlc filter=lfs diff=lfs merge=lfs -text
+qaihub/LibreYOLOXs_int8.dlc filter=lfs diff=lfs merge=lfs -text
+qairt/LibreYOLOXs_fp32.dlc filter=lfs diff=lfs merge=lfs -text
+qairt/LibreYOLOXs_int8.dlc filter=lfs diff=lfs merge=lfs -text
+snpe/LibreYOLOXs_fp32.dlc filter=lfs diff=lfs merge=lfs -text
+snpe/LibreYOLOXs_int8.dlc filter=lfs diff=lfs merge=lfs -text

LibreYOLOXs.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:14c84fa7b4dd4a0068e236ce8fc58ae9613b63b8a90376432dcbcf643a3dff8f
+size 35936139

LibreYOLOXs.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f55ded7181e1b0c13285c56e7790b8f0e8f8db590fe4edb37f0b7f345c913a30
+size 72089125

README.md

@@ -1,3 +1,158 @@
 ---
 license: mit
+language:
+  - en
+pipeline_tag: object-detection
+tags:
+  - object-detection
+  - yolox
+  - onnx
+  - dlc
+  - qualcomm
+  - snapdragon
+  - qairt
+  - snpe
+  - qai-hub
+  - edge-inference
+base_model: LibreYOLO/LibreYOLOXs
+datasets:
+  - merve/coco
 ---
+
+# LibreYOLOXs Qualcomm-Optimized Model Artifacts
+
+This directory is the shared artifact location for the notebooks in
+[`../notebooks/`](../notebooks/). The repository tracks the documentation and
+ignore rules here; model binaries are downloaded or generated locally when you
+run the workflows.
+
+The same source model is used for three deployment paths:
+
+- **QAIRT** local conversion and quantization
+- **SNPE** local conversion and quantization
+- **QAI Hub** cloud compilation, quantization, and profiling
+
+---
+
+## Model overview
+
+| Property | Value |
+|---|---|
+| Architecture | YOLOX-Small (LibreYOLOXs) |
+| Framework | [LibreYOLO](https://github.com/LibreYOLO/libreyolo) |
+| Task | Object detection |
+| Base checkpoint | [LibreYOLO/LibreYOLOXs](https://huggingface.co/LibreYOLO/LibreYOLOXs) |
+| Input tensor | `images` |
+| Input shape | `1 x 3 x 640 x 640` |
+| Input format | BGR `float32`, NCHW, values in `0-255` |
+
+---
+
+## Artifact flow
+
+```text
+LibreYOLOXs.pt
+    |
+    `---> LibreYOLOXs.onnx
+             |\
+             | `---> qairt/LibreYOLOXs_fp32.dlc
+             |        `---> qairt/LibreYOLOXs_int8.dlc
+             |
+             | `---> snpe/LibreYOLOXs_fp32.dlc
+             |        `---> snpe/LibreYOLOXs_int8.dlc
+             |
+             `---> qaihub/LibreYOLOXs_fp32.dlc
+                      `---> qaihub/LibreYOLOXs_int8.dlc
+```
+
+All three notebooks reuse the same exported ONNX model.
+
+---
+
+## Expected files
+
+| Path | Produced by | Description |
+|---|---|---|
+| `LibreYOLOXs.pt` | All notebooks | Base PyTorch checkpoint, downloaded on demand |
+| `LibreYOLOXs.onnx` | All notebooks | Shared ONNX export used by all backend flows |
+| `qairt/LibreYOLOXs_fp32.dlc` | QAIRT notebook | FP32 DLC produced by `qairt-converter` |
+| `qairt/LibreYOLOXs_int8.dlc` | QAIRT notebook | INT8 DLC produced by `qairt-quantizer` |
+| `snpe/LibreYOLOXs_fp32.dlc` | SNPE notebook | FP32 DLC produced by `snpe-onnx-to-dlc` |
+| `snpe/LibreYOLOXs_int8.dlc` | SNPE notebook | INT8 DLC produced by `snpe-dlc-quantize` |
+| `qaihub/LibreYOLOXs_fp32.dlc` | QAI Hub notebook | FP32 DLC downloaded from a cloud compile job |
+| `qaihub/LibreYOLOXs_int8.dlc` | QAI Hub notebook | INT8 DLC downloaded from cloud quantize + compile jobs |
+
+---
+
+## Tensor contract
+
+### Input
+
+| Property | Value |
+|---|---|
+| Name | `images` |
+| Shape | `1 x 3 x 640 x 640` |
+| Layout | NCHW |
+| Data type | `float32` |
+| Color order | BGR |
+
+Preprocessing used by the notebooks:
+
+1. Letterbox-resize to `640 x 640`.
+2. Pad with value `114`.
+3. Keep BGR channel order.
+4. Keep values in the `0-255` range.
+5. Transpose from HWC to CHW before serialization / inference input assembly.
+
+### Outputs
+
+The exported ONNX model is validated in each notebook before backend-specific
+compilation. The documentation in this directory assumes the shared decoded
+output layout used by the notebooks:
+
+| Output | Shape | Description |
+|---|---|---|
+| `bboxes` | `1 x 8400 x 4` | Bounding-box coordinates |
+| `scores` | `1 x 8400 x 81` | Objectness plus 80 COCO class scores |
+
+---
+
+## Calibration data
+
+INT8 flows use representative data prepared by the notebooks from COCO 2017
+validation images:
+
+- source images are staged under `docker/notebooks/dataset/`
+- calibration tensors are written to `docker/notebooks/subset/calib/`
+- validation tensors and metadata are written to `docker/notebooks/subset/val/`
+- manifests are written to `subset/calib/filenames.txt` and `subset/val/filenames.txt`
+
+The local QAIRT and SNPE quantizers read those manifests directly. The QAI Hub
+notebook loads the same `.raw` tensors into NumPy arrays and submits them to the
+cloud quantization job.
+
+---
+
+## Downloading published artifacts
+
+Published files are hosted on Hugging Face:
+
+- [fabricionarcizo/LibreYOLOXs](https://huggingface.co/fabricionarcizo/LibreYOLOXs)
+
+Example download:
+
+```bash
+huggingface-cli download fabricionarcizo/LibreYOLOXs \
+    --local-dir docker/models \
+    --exclude "*.gitattributes" "README.md"
+```
+
+You can also generate the same artifact layout locally by running the notebooks
+in [`../notebooks/`](../notebooks/).
+
+---
+
+## Related documentation
+
+- [`../README.md`](../README.md) - Docker environment and setup
+- [`../notebooks/README.md`](../notebooks/README.md) - notebook workflows and generated data layout

qaihub/LibreYOLOXs_fp32.dlc

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dcca8ae147c9eb07f58e24c98c8e50e18a2890c46e7db13b0c7e7f1008dbd020
+size 36185932

qaihub/LibreYOLOXs_int8.dlc

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0bfc361ebfb9a2ed4f5880515a9a637a25e6ebeca7457d6d9d64d6731e3cc54d
+size 9980684

qairt/LibreYOLOXs_fp32.dlc

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:35281e677dc2aaf2f93c179a10a69a05356951c30ae3c46698ce0e10ff2491d4
+size 36185284

qairt/LibreYOLOXs_int8.dlc

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:141566b8f5c94243f207d0342202cda0955fb348d44d60b7f2f75f3ecdfea026
+size 9270924

snpe/LibreYOLOXs_fp32.dlc

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c49527f64a67b539310b75304e570226a4f3ecf6cdacb3b8e21905fb04fbfd7f
+size 36181228

snpe/LibreYOLOXs_int8.dlc

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e8f07871461a758c17032ccec53146ae5946e42aed6e835ee89052490128a600
+size 9272956