Skip to content

Actions: google/gemma.cpp

Actions

All workflows

Actions

Loading...
Loading

Showing runs from all workflows
2,500+ workflow runs
2,500+ workflow runs

Filter by Workflow

⌥ + click/return to exclude

Filter by Event

⌥ + click/return to exclude

Filter by Status

⌥ + click/return to exclude

Filter by Branch

⌥ + click/return to exclude

Filter by Actor

⌥ + click/return to exclude
Add fixed MatMul scheduling controls (step 2 of #1002) build #8231: Pull request #1047 opened by Mikyx-1
Action required Mikyx-1:feat/matmul-scheduling
Mikyx-1:feat/matmul-scheduling
Action required
No public description build #8230: Commit eb083c9 pushed by copybara-service Bot
25m 39s dev
dev
25m 39s
Adds Intel AMX TMUL hardware acceleration for FlashAttention using INT8 build #8229: Pull request #1046 synchronize by copybara-service Bot
27m 22s test_992460558
test_992460558
27m 22s
No public description build #8228: Commit eb083c9 pushed by copybara-service Bot
24m 25s test_992460558
test_992460558
24m 25s
Adds Intel AMX TMUL hardware acceleration for FlashAttention using INT8 build #8227: Pull request #1046 synchronize by copybara-service Bot
17m 42s test_992460558
test_992460558
17m 42s
Internal change build #8226: Commit 29ab8ef pushed by copybara-service Bot
17m 56s test_992460558
test_992460558
17m 56s
Adds Intel AMX TMUL hardware acceleration for FlashAttention using INT8 build #8225: Pull request #1046 opened by copybara-service Bot
25m 23s test_992460558
test_992460558
25m 23s
Internal change build #8224: Commit c967ed6 pushed by copybara-service Bot
20m 55s test_992460558
test_992460558
20m 55s
No public description build #8223: Commit e52d655 pushed by copybara-service Bot
26m 19s dev
dev
26m 19s
Adds Intel AMX TMUL hardware acceleration for FlashAttention using INT8 build #8222: Pull request #1043 synchronize by copybara-service Bot
16m 16s test_977877682
test_977877682
16m 16s
No public description build #8221: Commit e52d655 pushed by copybara-service Bot
26m 51s test_977877682
test_977877682
26m 51s
Adds Intel AMX TMUL hardware acceleration for FlashAttention using INT8 build #8220: Pull request #1043 synchronize by copybara-service Bot
26m 1s test_977877682
test_977877682
26m 1s
Internal change build #8219: Commit 1fd4560 pushed by copybara-service Bot
28m 16s test_977877682
test_977877682
28m 16s
Add model-comparison tooling and structured MMLU reporting build #8218: Pull request #1041 labeled by jan-wassenberg
Action required Mikyx-1:eval/model-comparison
Mikyx-1:eval/model-comparison
Action required
When MxFp4Codec::QuantizeBlock encounters an all-zero input block, … build #8217: Commit 1c570b9 pushed by copybara-service Bot
25m 43s dev
dev
25m 43s
When MxFp4Codec::QuantizeBlock encounters an all-zero input block, the early-out path zeroes kBlockSize (32) bytes instead of the encoded block size sizeof(MxFp4Stream) (17 bytes: 1 scale byte plus 16 nibble bytes). Writing 15 bytes past the end of the block corrupts the adjacent block when blocks are encoded out of order (such as across parallel batch boundaries in Compress or via packed_ofs sub-ranges) and writes past PackedEnd on the final block of a tensor. build #8216: Pull request #1045 synchronize by copybara-service Bot
26m 39s test_991744020
test_991744020
26m 39s
When MxFp4Codec::QuantizeBlock encounters an all-zero input block, … build #8215: Commit 1c570b9 pushed by copybara-service Bot
26m 49s test_991744020
test_991744020
26m 49s
When MxFp4Codec::QuantizeBlock encounters an all-zero input block, the early-out path zeroes kBlockSize (32) bytes instead of the encoded block size sizeof(MxFp4Stream) (17 bytes: 1 scale byte plus 16 nibble bytes). Writing 15 bytes past the end of the block corrupts the adjacent block when blocks are encoded out of order (such as across parallel batch boundaries in Compress or via packed_ofs sub-ranges) and writes past PackedEnd on the final block of a tensor. build #8214: Pull request #1045 synchronize by copybara-service Bot
15m 13s test_991744020
test_991744020
15m 13s
When MxFp4Codec::QuantizeBlock encounters an all-zero input block, … build #8213: Commit 7f2a825 pushed by copybara-service Bot
15m 11s test_991744020
test_991744020
15m 11s
When MxFp4Codec::QuantizeBlock encounters an all-zero input block, the early-out path zeroes kBlockSize (32) bytes instead of the encoded block size sizeof(MxFp4Stream) (17 bytes: 1 scale byte plus 16 nibble bytes). Writing 15 bytes past the end of the block corrupts the adjacent block when blocks are encoded out of order (such as across parallel batch boundaries in Compress or via packed_ofs sub-ranges) and writes past PackedEnd on the final block of a tensor. build #8212: Pull request #1045 opened by copybara-service Bot
15m 53s test_991744020
test_991744020
15m 53s
When MxFp4Codec::QuantizeBlock encounters an all-zero input block, … build #8211: Commit eb0747c pushed by copybara-service Bot
25m 9s test_991744020
test_991744020
25m 9s
Adds Intel AMX TMUL hardware acceleration for FlashAttention using INT8 build #8210: Pull request #1043 opened by copybara-service Bot
25m 11s test_977877682
test_977877682
25m 11s
Internal change build #8209: Commit c9e5bb0 pushed by copybara-service Bot
20m 3s test_977877682
test_977877682
20m 3s
Internal change. build #8208: Pull request #1033 synchronize by copybara-service Bot
3m 13s test_979960879
test_979960879
3m 13s
Internal change. build #8207: Commit 8a8f452 pushed by copybara-service Bot
20m 58s test_979960879
test_979960879
20m 58s