These 7 commits are when the Protocol Buffers files have changed:
| Commit: | 96f98af | |
|---|---|---|
| Author: | borg323 | |
| Committer: | GitHub | |
fix xla with multiple devices (#2081)
The documentation is generated from this commit.
| Commit: | 0e43b7d | |
|---|---|---|
| Author: | borg323 | |
| Committer: | GitHub | |
onnx fp8 support for XLA conversion (#1988)
| Commit: | 760bdc9 | |
|---|---|---|
| Author: | Alexander Lyashuk | |
| Committer: | GitHub | |
XLA support for T79 net (#1965)
| Commit: | cfdc8ea | |
|---|---|---|
| Author: | Alexander Lyashuk | |
| Committer: | GitHub | |
Update ONNX-to-HLO converter to support T75 nets (#1959) * Add a way to add dependent computation * GlobalAveragePool * Squeeze * Introduce wrapper around HLO shape. * Split implemented * Sigmoid * Mul * Gather.. Hopefullyl it's correct at least for 1D case * Softmax and Identity * In the process of fixing. * T75 compiles \o? * Better batch sizes * Address review comments. * Fix a comment.
| Commit: | 04f73fc | |
|---|---|---|
| Author: | Alexander Lyashuk | |
| Committer: | GitHub | |
XLA backend for Lc0 (#1949)
| Commit: | a2f98f7 | |
|---|---|---|
| Author: | borg323 | |
| Committer: | GitHub | |
onnx2leela options to fix tf exported onnx models (#1928) * onnx2leela options to fix tf exported onnx models * initial support for onnx external data * autodetect fp16 onnx model * ensure outputs have the same data type as the input
| Commit: | a84a3ba | |
|---|---|---|
| Author: | Alexander Lyashuk | |
| Committer: | GitHub | |
Onnx backend (#1588)