Johannes Gäßler
1425f587a8
CUDA: attention sinks for mma FlashAttention ( #15157 )
2025-08-08 08:19:58 +02:00
lhez
aaa3d07ae7
opencl: support sink in soft_max (attn sinks) ( #15152 )
2025-08-07 21:47:03 -07:00
Xuan-Son Nguyen
50aa938901
convert : support non-mxfp4 HF model ( #15153 )
...
* convert : support non-mxfp4 HF model
* rm redundant check
* disable debug check
2025-08-07 23:26:03 +02:00
Jeff Bolz
c4f53563df
vulkan: support fattn sinks ( #15126 )
2025-08-07 22:44:20 +02:00
Jeff Bolz
a0552c8bee
vulkan: Add env var to disable host visible vidmem ( #15109 )
2025-08-07 22:07:11 +02:00
RunningLeon
99acbc9921
llama : Support intern-s1 ( #14875 )
...
* support internvl
* support interns1
* resolve comments
* put interns1 in tensor mapping
* resolve comment
* move tokenizer changes to sub class
2025-08-07 18:20:40 +02:00
uvos
7ad67ba9fe
HIP: add cmake option to enable compiler output of kernel resource usage metrics ( #15103 )
2025-08-07 16:44:14 +02:00
Christian Kastner
9a96389544
ggml: Skip backend library linking code when GGML_BACKEND_DL=ON ( #15094 )
...
Any available libraries are found and loaded dynamically at runtime.
2025-08-07 13:45:41 +02:00
Johannes Gäßler
1d72c84188
CUDA: GEMM for FP32/FP16/BF16 and ne11 <= 16 ( #15131 )
...
* CUDA: GEMM for FP32/FP16/BF16 and ne11 <= 16
2025-08-07 10:53:21 +02:00
Johannes Gäßler
20638e4f16
scripts: fix crash when --tool is not set ( #15133 )
2025-08-07 08:50:30 +02:00
Daniel Bevenius
36d3f00e14
requirements : fix PyTorch uint64 compatibility ( #15134 )
...
This commit addresses an issue with the convert_hf_to_gguf script
which is currently failing with:
```console
AttributeError: module 'torch' has no attribute 'uint64'
```
This occurred because safetensors expects torch.uint64 to be available
in the public API, but PyTorch 2.2.x only provides limited support for
unsigned types beyond uint8 it seems. The torch.uint64 dtype exists but
is not exposed in the standard torch namespace
(see pytorch/pytorch#58734 ).
PyTorch 2.4.0 properly exposes torch.uint64 in the public API, resolving
the compatibility issue with safetensors. This also required torchvision
to updated to =0.19.0 for compatibility.
Refs: https://huggingface.co/spaces/ggml-org/gguf-my-repo/discussions/186#68938de803e47d990aa087fb
Refs: https://github.com/pytorch/pytorch/issues/58734
2025-08-07 05:31:48 +02:00
Reese Levine
5fd160bbd9
ggml: Add basic SET_ROWS support in WebGPU ( #15137 )
...
* Begin work on set_rows
* Work on set rows
* Add error buffers for reporting unsupported SET_ROWS indices
* Remove extra comments
2025-08-06 15:14:40 -07:00
rmatif
756cfea826
fix profiling crash ( #15072 )
2025-08-06 14:17:51 -07:00
lhez
e725a1a982
opencl: add swiglu_oai and add_id ( #15121 )
...
* opencl: add `swiglu-oai`
* opencl: add `add_id`
* opencl: add missing `add_id.cl`
2025-08-06 12:12:17 -07:00
Sachin Desai
3db4da56a5
chat : support Granite model reasoning and tool call ( #14864 )
2025-08-06 20:27:30 +02:00
Juk Armstrong
476aa3fd57
Fixed name -override-tensors to -override-tensor ( #15129 )
2025-08-06 17:28:48 +01:00
Diego Devesa
0d8831543c
ggml : fix fallback to CPU for ununsupported ops ( #15118 )
2025-08-06 14:37:35 +02:00
Sigbjørn Skjæret
65c797c4fa
chat : fix yandex chat template ( #15116 )
2025-08-06 13:26:49 +02:00
stevenkuang
25726898e8
chat : fix hunyuan auto-detection ( #15114 )
...
Signed-off-by: stevenkuang <stevenkuang@tencent.com >
2025-08-06 11:48:30 +02:00
Chenguang Li
2241453252
CANN: add support for ACL Graph ( #15065 )
...
* feat(cann): add optional support for ACL Graph execution
This commit adds support for executing ggml computational graphs using
Huawei's ACL graph mode via the USE_CANN_GRAPH flag. The support can be
enabled at compile time using the CMake option:
-DUSE_CANN_GRAPH=ON
By default, ACL graph execution is **disabled**, and the fallback path
uses node-by-node execution.
Key additions:
- CMake option to toggle graph mode
- Graph capture and execution logic using
- Tensor property matching to determine whether graph update is required
- Safe fallback and logging if the environment variable LLAMA_SET_ROWS
is unset or invalid
This prepares the backend for performance improvements in repetitive graph
execution scenarios on Ascend devices.
Signed-off-by: noemotiovon <757486878@qq.com >
* Fix review comments
Signed-off-by: noemotiovon <757486878@qq.com >
* remane USE_CANN_GRAPH to USE_ACL_GRAPH
Signed-off-by: noemotiovon <757486878@qq.com >
* fix typo
Signed-off-by: noemotiovon <757486878@qq.com >
---------
Signed-off-by: noemotiovon <757486878@qq.com >
2025-08-06 14:12:42 +08:00
Reese Levine
9515c6131a
ggml: WebGPU disable SET_ROWS for now ( #15078 )
...
* Add paramater buffer pool, batching of submissions, refactor command building/submission
* Add header for linux builds
* Free staged parameter buffers at once
* Format with clang-format
* Fix thread-safe implementation
* Use device implicit synchronization
* Update workflow to use custom release
* Remove testing branch workflow
* Disable set_rows until it's implemented
* Fix potential issue around empty queue submission
* Try synchronous submission
* Try waiting on all futures explicitly
* Add debug
* Add more debug messages
* Work on getting ssh access for debugging
* Debug on failure
* Disable other tests
* Remove extra if
* Try more locking
* maybe passes?
* test
* Some cleanups
* Restore build file
* Remove extra testing branch ci
2025-08-05 16:26:38 -07:00
Georgi Gerganov
fd1234cb46
llama : add gpt-oss ( #15091 )
...
* oai moe
* compat with new checkpoint
* add attn sink impl
* add rope scaling yarn
* logits match with latest transformers code
* wip chat template
* rm trailing space
* use ggml_scale_bias
* rm redundant is_swa_all
* convert interleaved gate_up
* graph : fix activation function to match reference (#7 )
* vocab : handle o200k_harmony special tokens
* ggml : add attention sinks support (#1 )
* llama : add attn sinks
* ggml : add attn sinks
* cuda : add attn sinks
* vulkan : add support for sinks in softmax
remove unnecessary return
* ggml : add fused swiglu_oai op (#11 )
* ggml : add fused swiglu_oai op
* Update ggml/src/ggml-cpu/ops.cpp
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com >
* update CUDA impl
* cont : metal impl
* add vulkan impl
* test-backend-ops : more test cases, clean up
* llama : remove unfused impl
* remove extra lines
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com >
---------
Co-authored-by: slaren <slarengh@gmail.com >
* repack mxfp4 upon conversion
* clean up a bit
* enable thinking
* add quick hack to render only some special tokens
* fix bf16 conversion
* remove vocab hack
* webui ok
* support chat parsing for gpt-oss
* fix webui
* direct mapping mxfp4, FINALLY
* force using mxfp4
* properly use lazy tensor
* ggml : add mxfp4
ggml : use e8m0 conversion instead of powf
Co-authored-by: Diego Devesa <slarengh@gmail.com >
change kvalues_mxfp4 table to match e2m1 (#6 )
metal : remove quantization for now (not used)
cuda : fix disabled CUDA graphs due to ffn moe bias
vulkan : add support for mxfp4
cont : add cm2 dequant
* ggml : add ggml_add_id (#13 )
* ggml : add ggml_add_id
* add cuda impl
* llama : add weight support check for add_id
* perf opt
* add vulkan impl
* rename cuda files
* add metal impl
* allow in-place ggml_add_id
* llama : keep biases on CPU with --cpu-moe
* llama : fix compile error
ggml-ci
* cuda : add fallback for __nv_cvt_e8m0_to_bf16raw
ggml-ci
* cleanup
ggml-ci
* sycl : fix supports_op for MXFP4
ggml-ci
* fix Unknown reasoning format
* ggml-cpu : fix AVX build
ggml-ci
* fix hip build
ggml-ci
* cuda : add mxfp4 dequantization support for cuBLAS
ggml-ci
* ggml-cpu : fix mxfp4 fallback definitions for some architectures
ggml-ci
* cuda : fix version required for __nv_cvt_e8m0_to_bf16raw
---------
Co-authored-by: Xuan Son Nguyen <son@huggingface.co >
Co-authored-by: slaren <slarengh@gmail.com >
2025-08-05 22:10:36 +03:00
Sigbjørn Skjæret
f324a3b715
chat : only remove double bos/eos if added ( #15086 )
...
* only remove double bos/eos if added
* fix tests
2025-08-05 20:43:36 +02:00
Georgi Gerganov
be42642581
readme : update hot topics ( #15097 )
2025-08-05 20:19:33 +03:00
Romain Biessy
3306ceabf0
sycl: fix mul_mat selection ( #15092 )
2025-08-05 18:39:55 +02:00
Juk Armstrong
c81de6e107
Fix glm4moe bug ( #15088 )
2025-08-05 13:56:44 +01:00
Alex Wu
22f060c9c4
webui: fix markdown table ( #15081 )
...
* webui: fix markdown table
* webui: fix table display with themes
2025-08-05 13:56:44 +02:00
compilade
ee3a9fcf88
context : fix index overflow on huge outputs ( #15080 )
...
* context : fix overflow when re-ordering huge outputs
* context : fix logits size overflow for huge batches
2025-08-05 11:27:45 +02:00
Diego Devesa
ec428b02c3
llama : add --n-cpu-moe option ( #15077 )
...
* llama : add --n-cpu-moe option
Keeps the MoE weights of the first N layers in the CPU
2025-08-05 01:05:36 +02:00
compilade
19f68fa5a4
imatrix : warn when GGUF imatrix is saved without .gguf suffix ( #15076 )
...
* imatrix : add warning when suffix is not .gguf for GGUF imatrix
* imatrix : only warn about suffix when output format is unspecified
2025-08-04 23:26:52 +02:00
Christian Kastner
41613437ff
cmake: Add GGML_BACKEND_DIR option ( #15074 )
...
* cmake: Add GGML_BACKEND_DIR option
This can be used by distributions to specify where to look for backends
when ggml is built with GGML_BACKEND_DL=ON.
* Fix phrasing
2025-08-04 21:29:14 +02:00
Sigbjørn Skjæret
e5bebe5251
gguf-py : add --chat-template-file to gguf_new_metadata ( #15075 )
2025-08-04 21:01:48 +02:00
Sam
ef0144c087
model: support GLM 4.5 family of models ( #14939 )
...
* model: Add GLM 4.5 (#14921 )
Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com >
* Merge in PR suggestions
Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com >
* model: Add GLM 4.5 family of models (#14921 )
1. Updated tensor_mapping.py with NextN tensor mappings
- Added proper tensor mappings for all NextN/MTP tensors in /Users/samm/git/llama.cpp/gguf-py/gguf/tensor_mapping.py
- Added mappings for: eh_proj, embed_tokens, enorm, hnorm, shared_head.head, shared_head.norm
2. Added num_nextn_predict_layers configuration
- Added LLM_KV_NUM_NEXTN_PREDICT_LAYERS constant to llama-arch.h and llama-arch.cpp
- Added num_nextn_predict_layers field to llama_hparams struct
- Updated GLM4_MOE parameter loading in llama-model.cpp to read this parameter
- Modified tensor loading logic to conditionally load NextN tensors based on num_nextn_predict_layers
- Added GGUF writer support in gguf_writer.py with add_num_nextn_predict_layers() method
- Updated conversion script to extract and write this parameter from HuggingFace config
3. Added FIM tokens for GLM4_MOE
- Added GLM-4.5's FIM tokens to llama-vocab.cpp:
- <|code_prefix|> for FIM_PRE
- <|code_suffix|> for FIM_SUF
- <|code_middle|> for FIM_MID
4. Removed manual NextN tensor handling
- Removed the special-case handling in convert_hf_to_gguf.py that manually mapped NextN tensors
- NextN tensors are now handled automatically through the proper tensor mapping system
* glm 4.5 update tensors names
* model: glm 4.5 apply suggestions from code review
Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com >
* Update src/llama-model.cpp
Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com >
* model: glm 4.5 apply suggestions from code review
Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com >
* model: glm 4.5 apply suggestions from code review
* Apply suggestions from code review
* patch broken chat template
* typings fix
* add TENSOR_SKIP flag
Co-authored-by: Diego Devesa <slarengh@gmail.com >
* Update src/llama-model-loader.h
Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com >
---------
Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com >
Co-authored-by: Diego Devesa <slarengh@gmail.com >
2025-08-04 20:29:25 +02:00
Sigbjørn Skjæret
2721257e3e
quantize : fix confusing error message if ftype is invalid ( #15071 )
2025-08-04 18:11:02 +02:00
Reese Levine
587d0118f5
ggml: WebGPU backend host improvements and style fixing ( #14978 )
...
* Add parameter buffer pool, batching of submissions, refactor command building/submission
* Add header for linux builds
* Free staged parameter buffers at once
* Format with clang-format
* Fix thread-safe implementation
* Use device implicit synchronization
* Update workflow to use custom release
* Remove testing branch workflow
2025-08-04 08:52:43 -07:00
Jeff Bolz
5aa1105da2
vulkan: fix build when using glslang that does not support coopmat2 ( #15062 )
2025-08-04 07:09:19 +02:00
compilade
d31192b4ee
imatrix : use GGUF by default ( #14842 )
...
* imatrix : use GGUF by default
* imatrix : use GGUF regardless of the output filename
The legacy format can only be produced with --output-format dat
2025-08-03 22:00:05 +02:00
compilade
0a2f5496be
imatrix : fix 3d activation handling for hybrid and recurrent models ( #14994 )
...
* imatrix : use a single count for dense 3d tensors
* imatrix : fix 3d activations when model tensor is 2d
* imatrix : fix 3d tensor counts
2025-08-03 21:49:13 +02:00
compilade
11a3811164
memory : handle kv_unified for hybrid models ( #15050 )
2025-08-03 21:43:07 +02:00
Csaba Kecskemeti
97366dc6ab
vocab : JetBrains Mellum pre-tokenizer ( #15045 )
2025-08-03 21:38:18 +02:00
Gabriel Larson
83bc2f288c
model : add text-only support for Kimi-VL (and find special tokens in text_config) ( #15051 )
...
* basic kimi-vl textmodel conversion
* check config["text_config"] for special tokens
2025-08-03 16:56:25 +02:00
Jeff Bolz
6c7a441161
vulkan: Use coopmat2 for conv2d ( #14982 )
2025-08-03 14:23:57 +02:00
lhez
5c0eb5ef54
opencl: fix adreno compiler detection logic ( #15029 )
2025-08-02 19:51:18 +02:00
Johannes Gäßler
03d4698218
CUDA: use mma FA kernel for gqa > 4 on RTX 4000 ( #15035 )
2025-08-02 16:37:08 +02:00
leejet
3303c19b16
cuda: make im2col a little faster ( #15025 )
2025-08-02 17:15:36 +03:00
Daniel Bevenius
4fdea540bd
kv-cache : skip alignment of n_stream in kv-cache log msg [no ci] ( #15040 )
...
This commit removes the right alignment the `n_stream` value in the
log message in the `llama_kv_cache_unified` constructor.
The motivation for this change is to enhance the readability of log
message. Currently the output looks like this:
```console
llama_kv_cache_unified: size = 2048.00 MiB ( 4096 cells, 32 layers, 1/ 1 seqs), K (f16): 1024.00 MiB, V (f16): 1024.00 MiB
```
Notice that the `n_stream` value is right aligned, which makes it a
little harder to read.
With the change in this commit the output will look like
```console
llama_kv_cache_unified: size = 2048.00 MiB ( 4096 cells, 32 layers, 1/1 seqs), K (f16): 1024.00 MiB, V (f16): 1024.00 MiB
```
2025-08-02 17:14:57 +03:00
Georgi Gerganov
a4569c41fd
llama : enable LLAMA_SET_ROWS=1 by default ( #14959 )
...
ggml-ci
2025-08-02 17:14:21 +03:00
Georgi Gerganov
15e92fd337
cuda, sycl : fix batched gemm when ne02 == 1 && ne03 > 1 ( #15038 )
...
* cuda, sycl : fix batched gemm when ne02 == 1 && ne03 > 1
ggml-ci
* cont : fix cont types
ggml-ci
* cont : adopt variable names and comment from the other branch
2025-08-02 17:13:05 +03:00
Sigbjørn Skjæret
2bf3fbf0b5
ci : check that pre-tokenizer hashes are up-to-date ( #15032 )
...
* torch is not required for convert_hf_to_gguf_update
* add --check-missing parameter
* check that pre-tokenizer hashes are up-to-date
2025-08-02 14:39:01 +02:00
Douglas Hanley
711d5e6fe6
convert : fix Qwen3-Embedding pre-tokenizer hash ( #15030 )
2025-08-02 12:51:02 +02:00
