LLM-Inference/whisper.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/whisper.cpp.git synced 2026-08-12 22:31:22 +04:00
Code Issues Packages Projects Releases Wiki Activity
4,889 Commits 65 Branches 88 Tags
master
Commit Graph
26 Commits
This Branch
This Branch
All Branches
Author SHA1 Message Date
Masashi YoshimuraandGeorgi Gerganov cabd684e45 ggml-webgpu: add support for f16 repeat (llama/26307) 2026-08-04 13:37:47 +03:00
Masashi YoshimuraandGeorgi Gerganov f97417517f ggml-webgpu: improve flash_attn_vec for quantized KV at long contexts (llama/25956) 
* improve fa of quantized kv cache

* Fix some bugs and some comments.

* fix v type check and some comments

* Fix build error caused by rebasing

* editorconfig checking pass
 2026-08-04 13:37:47 +03:00
Masashi YoshimuraandGeorgi Gerganov 4b914096e8 ggml-webgpu: tune subgroup split (d_split) in flash_attn_vec (llama/25418) 2026-07-10 13:06:42 +03:00
Masashi YoshimuraandGeorgi Gerganov b313fd3ce8 ggml-webgpu: add support for NVFP4 (llama/25143) 2026-07-10 13:06:42 +03:00
Masashi YoshimuraandGeorgi Gerganov 2955e84ec3 ggml-webgpu: improve MTP inference by using mat-vec path for small batches (llama/24811) 
* ggml-webgpu: improve small batches decoding

* Add barrier to the NUM_COLS loop in mul-mat-vec
 2026-06-26 16:03:57 +03:00
Masashi YoshimuraandGeorgi Gerganov 75b90fb0ea ggml-webgpu: add adapter toggles for F16 on Vulkan + NVIDIA 2026-06-26 16:03:57 +03:00
Masashi YoshimuraandGeorgi Gerganov 3e0b917514 ggml-webgpu: improve i-quants mul_mat performance and speed up prefill (llama/24530) 
* Improve prefill speeds for i-quants

* Fix #if defined() usage in preprocessor guards.
 2026-06-19 12:53:43 +03:00
Masashi YoshimuraandGeorgi Gerganov aa42b48312 ggml-webgpu: Improve prefill speeds for k-quants + refactor matmul for Q4/Q5/Q8 and k-quants (llama/24225) 
* ggml-webgpu: Improve prefill speeds + refactor matmul for quants

* Fixes for editroconfig checker
 2026-06-15 10:33:53 +03:00
Masashi YoshimuraandGeorgi Gerganov db2a39507c revert to using global_invocation_id for cpy shader (llama/23955) 2026-06-08 14:36:36 +03:00
Masashi YoshimuraandGeorgi Gerganov a52bd385d6 ggml-webgpu: Fix how to dispatch WG to some ops (llama/23750) 2026-05-29 09:47:30 +03:00
Masashi YoshimuraandGeorgi Gerganov 00a5110b19 ggml-webgpu: Add MMVQ path for Q4/Q8/Q2_K/Q4_K and clean up legacy MUL_MAT pipeline (llama/23594) 
* ggml-webgpu: Add MMVQ path for Q4/Q8/Q2_K/Q4_K

* Fix to editorconfig checking pass

* Remove mul-mat-legacy pipeline

* Fix to use vendor name as is and add dot_product/vendor to shader_lib_ctx
 2026-05-29 09:47:30 +03:00
Masashi YoshimuraandGeorgi Gerganov 1cbbd0b6d0 flush the gpu profile timestamp before the queryset is overflowed (llama/22995) 2026-05-14 21:26:48 +03:00
Masashi YoshimuraandGeorgi Gerganov e8a7cd314f ggml-webgpu: Enables running gpt-oss-20b (llama/22906) 
* Enable to run gpt-oss-20b and refactor mulmat-q

* disable test-backend-ops in ubuntu-24-webgpu
 2026-05-14 21:26:48 +03:00
Masashi YoshimuraGeorgi GerganovSigbjørn Skjæret
 		9623c1203b ggml-webgpu: Fix vectorized handling in mul-mat and mul-mat-id (llama/22578) 
* Fix vectorized condition of mul-mat-fast pipeline and add vectorized variant to mul-mat-id

* Apply suggestion from @CISC

Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>

---------

Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>
 2026-05-02 15:02:42 +03:00
Masashi YoshimuraandGeorgi Gerganov b34a9f3d83 ggml-webgpu: Improve performance of mat-vec and mat-mat for MUL_MAT_ID (llama/22464) 
* Add mat-vec fast path of MUL_MAT_ID.

* Add shared accumulation vec logic and the other types supports.

* Add i-quant mat-mat for MUL_MAT_ID and fix some parts

* Remove n_experts from shader_lib_context.
 2026-05-01 13:07:35 +03:00
Masashi YoshimuraandGeorgi Gerganov 0fbe4c4ca7 ggml-webgpu: Add fused RMS_NORM + MUL (llama/21983) 
* fused rms_norm_mul + mul

* Add GGML_WEBGPU_DISABLE_FUSION for being able to disable kernel fusion.

* Decouple num_fused_ops from webgpu_context; misc cleanup

* Fix eps handling and remove disable_fusion.

* Fix not to use c++20 initializers.
 2026-04-30 11:29:16 +03:00
Masashi YoshimuraandGeorgi Gerganov 2e5eb6e951 ggml-webgpu: reset CPU/GPU profiling time when freeing context (llama/22050) 
* Reset the CPU/GPU profiling time when freeing context.

* move GPU profiling time from global context to webgpu_context.
 2026-04-30 11:29:15 +03:00
Masashi YoshimuraandGeorgi Gerganov 36b7bb3d95 Remove extra conditional check on debug mode. (llama/21798) 2026-04-30 11:29:07 +03:00
Masashi YoshimuraGeorgi GerganovReese Levine
 		9cbc4b3acb ggml-webgpu: Add the support of MUL_MAT_ID (llama/21147) 
* Add mul_mat_id support to WebGPU

* Apply suggestion from @reeselevine

---------

Co-authored-by: Reese Levine <reeselevine1@gmail.com>
 2026-04-30 11:29:01 +03:00
Masashi YoshimuraandGeorgi Gerganov 3d004fbf0a ggml-webgpu: Update the RMS_NORM preprocessor and add L2_NORM (llama/20665) 
* Update the preprocessor of RMS_NORM and add L2_NORM.

* Fix the name of rms_norm to row_norm.
 2026-03-29 15:04:36 +03:00
Masashi YoshimuraandGeorgi Gerganov 12015a2174 ggml-webgpu: Add supports for DIAG and TRI (llama/20664) 
* Add supports for DIAG and TRI.

* Remove extra ttype and add a comment for TRI op.
 2026-03-29 15:04:36 +03:00
Masashi YoshimuraandGeorgi Gerganov 5267523829 ggml-webgpu: Add supports for GGML_OP_REPEAT (llama/20230) 
* Add GGML_OP_REPEAT to webgpu backend.

* Add i16 support for GGML_OP_REPEAT.
 2026-03-16 13:10:15 +02:00
Masashi YoshimuraandGeorgi Gerganov 8d78d40946 Add concat op to webgpu. (llama/20068) 2026-03-16 13:10:15 +02:00
Masashi YoshimuraandGeorgi Gerganov de686fafad ggml-webgpu: Support non-contiguous src0 and overlapping src0/src1 in binary ops (llama/19850) 
* ggml-webgpu: Add binary op support for overlapping and non-contiguous.

* Add newline to binary.wgsl

* Append the test of binary op for src overlapping  to test_bin_bcast.

* Remove unnecessary newline.
 2026-03-16 13:10:15 +02:00
Masashi YoshimuraandGeorgi Gerganov 0158795ebc ggml-webgpu: Add unary op (SQR, SQRT, SIN, COS) support. (llama/19700) 
* ggml-webgpu: Add unary op (SQR, SQRT, SIN, COS) support.

* Fix to cast the src value to f32 before sin/cos computing.
 2026-02-27 20:57:58 +02:00
Masashi YoshimuraandGeorgi Gerganov a71127dfd8 ggml-webgpu: Fix GGML_MEM_ALIGN to 8 for emscripten. (llama/18628) 
* Fix GGML_MEM_ALIGN to 8 for emscripten.

* Add a comment explaining the need for GGML_MEM_ALIGN == 8 in 64-bit wasm with emscripten
 2026-01-14 09:11:59 +02:00