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
Quantize softmax probabilities S into [0, 255] uint8_t and utilize unsigned-by-signed (uint8 * int8) matrix multiplication for SV accumulation, improving quantization precision. Add SIMD emulation fallback using SumOfMulQuadAccumulate and TableLookupBytes. build #8081: Pull request #1013 synchronize by copybara-service Bot
In progress test_976294440
test_976294440
In progress
Quantize softmax probabilities S into [0, 255] uint8_t and utilize un… build #8080: Commit 38f1ea1 pushed by copybara-service Bot
In progress test_976294440
test_976294440
In progress
Internal changes build #8079: Pull request #1021 synchronize by copybara-service Bot
18m 49s test_975655056
test_975655056
18m 49s
Internal changes build #8078: Commit 1703d19 pushed by copybara-service Bot
25m 44s test_975655056
test_975655056
25m 44s
Quantize softmax probabilities S into [0, 255] uint8_t and utilize unsigned-by-signed (uint8 * int8) matrix multiplication for SV accumulation, improving quantization precision. Add SIMD emulation fallback using SumOfMulQuadAccumulate and TableLookupBytes. build #8077: Pull request #1013 synchronize by copybara-service Bot
19m 51s test_976294440
test_976294440
19m 51s
Quantize softmax probabilities S into [0, 255] uint8_t and utilize un… build #8076: Commit 6fae11f pushed by copybara-service Bot
26m 40s test_976294440
test_976294440
26m 40s
Internal changes build #8075: Pull request #1022 opened by copybara-service Bot
24m 11s test_976250344
test_976250344
24m 11s
Internal changes build #8074: Commit 18052ed pushed by copybara-service Bot
26m 9s test_976250344
test_976250344
26m 9s
Internal changes build #8073: Pull request #1021 opened by copybara-service Bot
25m 24s test_975655056
test_975655056
25m 24s
Internal changes build #8072: Commit 81a3fd6 pushed by copybara-service Bot
26m 0s test_975655056
test_975655056
26m 0s
Reduce local KV memory in BF16 Flash attention build #8071: Pull request #1020 opened by Mikyx-1
Action required Mikyx-1:fix/local-kv-rings-dev
Mikyx-1:fix/local-kv-rings-dev
Action required
Block FP32 vision attention value accumulation build #8070: Pull request #1019 opened by Mikyx-1
27m 27s Mikyx-1:perf/block-vision-values
Mikyx-1:perf/block-vision-values
27m 27s
Block FP32 vision attention value accumulation GitHub Actions Scan #8: Pull request #1019 opened by Mikyx-1
8s
8s
Reduce local-attention KV storage with per-layer rings build #8069: Pull request #1017 synchronize by Mikyx-1
3h 52m 24s Mikyx-1:kv-cache-local-windows
Mikyx-1:kv-cache-local-windows
3h 52m 24s
Reduce local-attention KV storage with per-layer rings GitHub Actions Scan #7: Pull request #1017 synchronize by Mikyx-1
11s
11s
Reduce local-attention KV storage with per-layer rings GitHub Actions Scan #6: Pull request #1017 opened by Mikyx-1
10s
10s
Reduce local-attention KV storage with per-layer rings build #8068: Pull request #1017 opened by Mikyx-1
Action required Mikyx-1:kv-cache-local-windows
Mikyx-1:kv-cache-local-windows
Action required
Add W8A8 int8 MatMul with optional rotations build #8067: Pull request #1003 synchronize by Mikyx-1
Action required Mikyx-1:w8a8-int8-matmul-dev
Mikyx-1:w8a8-int8-matmul-dev
Action required
Fix tiled KV cache allocation and support prefix cloning. build #8066: Pull request #1015 opened by copybara-service Bot
35m 4s test_976288579
test_976288579
35m 4s
Fix tiled KV cache allocation and support prefix cloning. build #8065: Commit 0518257 pushed by copybara-service Bot
25m 37s test_976288579
test_976288579
25m 37s
Add Highway SIMD logit masking kernel for constrained decoding. build #8064: Pull request #1014 opened by copybara-service Bot
27m 8s test_976288582
test_976288582
27m 8s
Add Highway SIMD logit masking kernel for constrained decoding. build #8063: Commit 1ae26da pushed by copybara-service Bot
25m 32s test_976288582
test_976288582
25m 32s
Merge pull request #1012 from RaviTriv:skip-shared-kv-alloc build #8062: Commit 1658f88 pushed by copybara-service Bot
26m 16s dev
dev
26m 16s
Quantize softmax probabilities S into [0, 255] uint8_t and utilize unsigned-by-signed (uint8 * int8) matrix multiplication for SV accumulation, improving quantization precision. Add SIMD emulation fallback using SumOfMulQuadAccumulate and TableLookupBytes. build #8061: Pull request #1013 opened by copybara-service Bot
26m 18s test_976294440
test_976294440
26m 18s
Quantize softmax probabilities S into [0, 255] uint8_t and utilize un… build #8060: Commit b09ef81 pushed by copybara-service Bot
23m 10s test_976294440
test_976294440
23m 10s