be30473dc6
Internal changes
...
PiperOrigin-RevId: 835160876
2025-11-21 04:08:04 -08:00
5a500872b8
Internal change
...
PiperOrigin-RevId: 835115693
2025-11-21 01:17:45 -08:00
49d420aeaf
Add some comments.
...
PiperOrigin-RevId: 834173319
2025-11-19 01:09:15 -08:00
b8f6be72b1
Improves autodetection of Gemma3-1B.
...
Uses the key_norm and query_norm layers to disambiguate between the Gemma2-2B and Gemma3-1B models.
Since Gemma3-1B is not multimodal, ViT is not an effective disambiguator. KQ normalization is a structural disambiguator between gemma2 and gemma3.
PiperOrigin-RevId: 833213331
2025-11-17 01:12:50 -08:00
3e18db17f4
Avoid hard-coding kPatchSize. Thanks @Somet2mes for reporting. Fixes #762 .
...
PiperOrigin-RevId: 829308896
2025-11-07 00:32:31 -08:00
f8131339a7
Refactor for continous batching. This cl does not change the current behavior of the code. It only extract two functions that will later be called for adding continuous batching.
...
PiperOrigin-RevId: 829104661
2025-11-06 14:20:17 -08:00
35e9f9f05f
Introduce attention implementation configurability.
...
PiperOrigin-RevId: 828971705
2025-11-06 08:43:41 -08:00
091b4567c9
Minor: ParallelismStrategy->Parallelism
...
PiperOrigin-RevId: 828936578
2025-11-06 06:56:10 -08:00
a344a70c59
Change (old) attention behavior to disallow wraparound, enforced via assertion.
...
Shared kU64PerLine constant
PiperOrigin-RevId: 828072451
2025-11-04 11:52:40 -08:00
3a63a12624
Allow prefill only run by allowing max_prompt_size == seq_len
...
PiperOrigin-RevId: 827415258
2025-11-03 03:17:54 -08:00
ab87807a4c
Pre-compress query activations to BF16 before FlashAttention.
...
PiperOrigin-RevId: 826524997
2025-10-31 09:49:44 -07:00
8a100c1e8d
Added access to flash attention internals to TileFlashAttention4
...
PiperOrigin-RevId: 826011137
2025-10-30 06:50:05 -07:00
116cd6eff6
BF16 mixed-mode flash attention
...
PiperOrigin-RevId: 825433929
2025-10-29 01:48:28 -07:00
4bd465ffd3
Also update attention.h to type-erased query_norm_scale
...
PiperOrigin-RevId: 825014334
2025-10-28 06:48:33 -07:00
3cc0139ebb
Fix excessive KC/MC from prior change
...
This could lead to stack overflow in B_storage.
Also do not require specific type for query_norm_scale,
update batch sizes for attention tensors,
more verbose Mat shape/type checks.
PiperOrigin-RevId: 824987689
2025-10-28 05:33:01 -07:00
5a05857deb
[Gemma.cpp] Allows non-owned arguments for attention methods.
...
* Adds and uses a new `AttentionActivationPtrs` that holds non-owning `MatPtrs`. Acts as a view into `AttentionActivations`.
* Updates `QBatch` to hold non-owning `MatPtr`s to the kv caches.
* Enables the `MatPtrT` default constructor for simpler initializations.
* Pulls out and passes `LayerWeightsPtrs::query_norm_scale` directly. While `LayerWeightsPtrs` already held non-owning `MatPtr`s, this change avoids the need to find and construct several empty weight tensors just to construct one `query_norm_scale` tensor.
PiperOrigin-RevId: 824584177
2025-10-27 10:43:25 -07:00
1bdde1af3c
Add config flag for global timescale & rely on config to deduce wrapping
...
PiperOrigin-RevId: 823512377
2025-10-24 06:54:56 -07:00
3ed403e287
Major cleanup of profiler zones, add Caller annotation for all pool.Run
...
Pass ThreadingContext instead of Pools/Profiler individually, for access to Zones
Add GCPP_ZONE helper
Add Caller argument to pool.Run to enable new stats
Remove most direct dependencies on ThreadPool, prefer ParallelFor
PiperOrigin-RevId: 822934530
2025-10-23 01:54:24 -07:00
503aaddd65
Add 8-bit integer quantization (I8Stream) to Gemma.cpp.
...
PiperOrigin-RevId: 819787856
2025-10-15 09:25:20 -07:00
ee18916abf
Removed the PROFILER_ZONE from the most highly called functions to reduce the overhead.
...
PiperOrigin-RevId: 819739402
2025-10-15 07:10:04 -07:00
fb6fa793f4
Added a global (to gemma) zones list to enable most call sites to PROFILER_ZONE3 to avoid the sychronization required for the static const initialization of the zone handle.
...
Improved flash_attention to enable profiling using the new zones.
PiperOrigin-RevId: 819235421
2025-10-14 08:30:58 -07:00
035273c184
tune pool kSpin mode in threading_context
...
Previously, this happened concurrently with the matmul autotune, which could lead to incorrect outcomes.
threading: de-singleton Pinning (no longer stores affinity); pass PoolWorkerMapping; fix Pool dtor order
Also enable SPR target (Zen4 is AMD-only),
update Highway version for renamed Thread()->GlobalIdx().
PiperOrigin-RevId: 816223017
2025-10-07 08:36:26 -07:00
684a0444e9
Reduced parallelism for TransposeQ, making each thread read and write within its own cache lines
...
PiperOrigin-RevId: 814241032
2025-10-02 08:15:16 -07:00
14244664c8
Avoid transposing Q when it isn't needed
...
PiperOrigin-RevId: 814187984
2025-10-02 05:16:35 -07:00
fe5a39990e
Improve FlashAttention threading:
...
kFlat for RMSNorm (hierarchical is excessive),
profiler zone naming improvements.
PiperOrigin-RevId: 814144012
2025-10-02 02:37:05 -07:00
6098a022b3
Increased parallelism for RMSNormAndPositionalEncoding
...
PiperOrigin-RevId: 813738994
2025-10-01 07:11:14 -07:00
2f6cbde8ff
Added a smaller tile size to flash attention for smaller batch sizes
...
PiperOrigin-RevId: 813226193
2025-09-30 05:49:20 -07:00
4974f24832
Fixed bug with softcap in single flash attention
...
PiperOrigin-RevId: 813164938
2025-09-30 02:17:58 -07:00
667a3f117a
Utilize multiple cores to read weight batches.
...
PiperOrigin-RevId: 811893059
2025-09-26 11:28:33 -07:00
4f0c633248
(1) Added QueryResultAndMetrics and BatchQueryModelWithMetrics to also return TimingInfo besides query results.
...
PiperOrigin-RevId: 810634261
2025-09-23 17:02:29 -07:00
fac8aac4cb
Internal change
...
PiperOrigin-RevId: 809975026
2025-09-22 05:37:03 -07:00
501fdf000e
Remove no longer used MatVec
...
PiperOrigin-RevId: 809059409
2025-09-19 09:03:22 -07:00
f3bc1c17da
1.03x speedup: fused FFN
...
matmul-inl: support CView=StridedView or RowPtrs; rename to C_MC_NC
matmul.cc: Allow 1 more rep for MC/NC to allow half-sized tiles, which helps.
PiperOrigin-RevId: 807291701
2025-09-15 10:26:37 -07:00
c9b8479f7d
Added zero-initialization to att_out.
...
Re-enabled flash attention when HWY_NATIVE_DOT_BF16 is not available.
PiperOrigin-RevId: 806284756
2025-09-12 07:48:23 -07:00
2695aab5d2
Temporarily disable flash pending msan fix
...
PiperOrigin-RevId: 805350234
2025-09-10 07:25:41 -07:00
ba6131311a
Fix gemma_batch_bench for flash attention
...
q_T rows do not change.
Also repeat prefill to reflect perf after autotuning.
PiperOrigin-RevId: 805319377
2025-09-10 05:32:34 -07:00
f10ac41a20
Added flash attention, with both a single-q function, and a register-tiled function.
...
The register-tiled version achieves a speed-up by a factor of about 9.7 over the previous attention function on an AVX3-enabled machine.
PiperOrigin-RevId: 804913784
2025-09-09 08:05:26 -07:00
461a9c7d1b
Matmul refactoring towards fusion
...
MMLoops: move dispatch code out, use overloads
split build target into matmul_env (for MatMulEnv/MMOptions)
weights: no longer call BindB
Fix potential out of bounds in gemma_batch_bench
PiperOrigin-RevId: 804895985
2025-09-09 07:13:38 -07:00
a5ab99e4ba
Memory use reduction: smaller/single MMStorage
...
PiperOrigin-RevId: 804865029
2025-09-09 05:32:46 -07:00
6e52a835c6
Faster startup on tsan: use hierarchical parallelism for BF16 conversion
...
Also re-enable profiler zones
PiperOrigin-RevId: 804273899
2025-09-07 22:50:31 -07:00
cbe24eac51
1.15x speedup: parallel sampling, enabled by new RNG
...
Also pass pos to SampleFunc, for seeding the RNG.
PiperOrigin-RevId: 803453518
2025-09-05 07:24:02 -07:00
2b4c16e243
Remove Griffin support
...
Also add IsObsolete helper
PiperOrigin-RevId: 803376921
2025-09-05 02:35:40 -07:00
56186193c1
Replace mt19937 with new generator to enable parallel sampling
...
Split it into immutable AesCtrEngine and RngStream
Also add RowSpan and Logits span
PiperOrigin-RevId: 803336423
2025-09-04 23:49:10 -07:00
5d1693e806
Internal change
...
PiperOrigin-RevId: 803083229
2025-09-04 10:31:20 -07:00
4be4799727
Remove kMaxPackages and per-package-related code
...
matmul: remove kMaxClusters, dynamic allocation
PiperOrigin-RevId: 802950348
2025-09-04 03:33:12 -07:00
7263ab8445
MatMul simplification, threading strategy improvements
...
remove MatMul f32 special case (smaller code),
types: Add u32/u64 for use by Activations
move renamed ParallelismStrategy to threading_context so can pass ctx
ensure worker index is unique across clusters
matmul.h: const member functions for renamed policy classes (easier to call)
PiperOrigin-RevId: 802848086
2025-09-03 21:45:07 -07:00
b7b3d353db
Simplify MatMul: remove F32 special case (build time)
...
Also move kMaxM into separate kMaxBatchSize
PiperOrigin-RevId: 802086590
2025-09-02 04:29:21 -07:00
1e3c853e80
Add ParallelFor wrapper function and one new mode
...
Move ParallelismType from matmul.h to threading.h
Replace SmallParallelFor with ParallelFor and the new mode
PiperOrigin-RevId: 802038452
2025-09-02 01:40:09 -07:00
229bd078a1
1.29x speedup: bf16 C1/C2. Extend most ops to any type, expand test coverage.
...
Also increase dot_test.cc range for Zen4, and matmul_test tolerance (failing in some configs)
PiperOrigin-RevId: 801789922
2025-09-01 06:34:04 -07:00
0ae8646731
Fix remainder handling for Paligemma
...
No longer attempt to skip the remainder handling because B might also be a non-padded view.
PiperOrigin-RevId: 800890805
2025-08-29 07:25:52 -07:00
Krzysztof Rymski
Martin Stolle
The gemma.cpp Authors
Jan Wassenberg
Charles Zhao
Phil Culliton
Ray Smith
Biruk Mammo
Theotime Combes
Nitin Gangahar