Document standalone inference helper
Browse files
README.md
CHANGED
|
@@ -10,6 +10,7 @@ Contents:
|
|
| 10 |
- `transformer/config.json`
|
| 11 |
- `transformer/diffusion_pytorch_model.safetensors`
|
| 12 |
- `backend_snapshot/`
|
|
|
|
| 13 |
|
| 14 |
Training run:
|
| 15 |
|
|
@@ -32,6 +33,7 @@ It also includes the inference entrypoint snapshot and an example script:
|
|
| 32 |
|
| 33 |
- `backend_snapshot/scripts/inference/run_sfp4_ours_p_checkpoint_700.sh`
|
| 34 |
- `backend_snapshot/training_attention_settings.json`
|
|
|
|
| 35 |
|
| 36 |
Attention setup for this checkpoint:
|
| 37 |
|
|
@@ -39,3 +41,8 @@ Attention setup for this checkpoint:
|
|
| 39 |
tiles, group-local P quant, dropped-tile mean compensation
|
| 40 |
- cross-attention: dense SDPA fallback, not FP4/sparse
|
| 41 |
- force-dense paths: dense SDPA
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 10 |
- `transformer/config.json`
|
| 11 |
- `transformer/diffusion_pytorch_model.safetensors`
|
| 12 |
- `backend_snapshot/`
|
| 13 |
+
- `standalone_inference/`
|
| 14 |
|
| 15 |
Training run:
|
| 16 |
|
|
|
|
| 33 |
|
| 34 |
- `backend_snapshot/scripts/inference/run_sfp4_ours_p_checkpoint_700.sh`
|
| 35 |
- `backend_snapshot/training_attention_settings.json`
|
| 36 |
+
- `standalone_inference/`
|
| 37 |
|
| 38 |
Attention setup for this checkpoint:
|
| 39 |
|
|
|
|
| 41 |
tiles, group-local P quant, dropped-tile mean compensation
|
| 42 |
- cross-attention: dense SDPA fallback, not FP4/sparse
|
| 43 |
- force-dense paths: dense SDPA
|
| 44 |
+
|
| 45 |
+
`standalone_inference/` is a portable helper for normal inference. It contains
|
| 46 |
+
an overlay installer, a runner that downloads/loads the checkpoint-700
|
| 47 |
+
transformer weights, and the sparse FP4 backend files required by this
|
| 48 |
+
checkpoint.
|