AlexIndustrial/mesa
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

etnaviv/ml: Fix padding input/output tensor zero points

Browse Source 
For tensors that were converted from signed 8-bit tensors to unsigned
8-bit tensors with offset zero point, use the offset zero point also
for the TP pad operation.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34474>
This commit is contained in:
Philipp Zabel
2025-03-03 15:42:24 +01:00
committed by Marge Bot
parent 13a120d13c
commit f23b376e84
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/gallium/drivers/etnaviv/etnaviv_ml_tp.c
+2 -2
View File
@@ -817,14 +817,14 @@ etna_ml_lower_pad(struct etna_ml_subgraph *subgraph,
operation->input_tensor_sizes[0] = operation->input_width *
operation->input_height *
operation->input_channels;
operation->input_zero_point = pad->input_tensors[0]->zero_point;
operation->input_zero_point = etna_tensor_zero_point(pad->input_tensors[0]);
operation->input_scale = pad->input_tensors[0]->scale;
operation->output_tensors[0] = pad->output_tensors[0]->index;
operation->output_width = pad->output_tensors[0]->dims[1];
operation->output_height = pad->output_tensors[0]->dims[2];
operation->output_channels = pad->output_tensors[0]->dims[3];
operation->output_zero_point = pad->output_tensors[0]->zero_point;
operation->output_zero_point = etna_tensor_zero_point(pad->output_tensors[0]);
operation->output_scale = pad->output_tensors[0]->scale;
operation->output_tensor_sizes[0] = operation->output_width *
operation->output_height *