* spec: support MTP
* fix batch size
* rename files
* cont : simplify (#7)
* MTP: clean-up (#9)
* MTP: clean-up
* review: use llama_context_type instead of llama_graph_type
* review: remove llama_model_has_mtp
* review: fix convert issues
* convert: fix pycheck
* review: formatting
* use `mtp-` for identifying mtp models
* convert: fix mtp conversion
* mtp -> draft-mtp
* remove unused llama_arch
* add need_embd in speculative
* llama: allow partial seq_rm for GDN models for speculative decoding
Currently speculative checkpoint needs to restart from a checkpoint
after some draft tokens are not accepted, this leads to some wastage in
running the target again. This PR adds the ability to rollback upto
`draft_max` by storing the GDN intermediates.
* fix pending state
* vulkan: add GDN partial rollback
* meta: extend check to axis 1
* metal: add GDN partial rollback
Extend the gated delta net kernel to store intermediate states for
partial rollback support on the Metal backend.
- Add K (snapshot slot count) as a function constant
- Read input state from slot 0 of the 3D state tensor
- Write intermediate states to different slots during token loop
- For K=1, maintain backward-compatible single-slot behavior
Ref: https://github.com/ggml-org/llama.cpp/commit/8c05923630110223669f069af2000e9cf10c02bc
Assisted-by: llama.cpp:local pi
* delta_net_base: use ggml_pad instead of new_tensor
* review: add need_rs_seq
* review: rename part_bounded to n_rs
* review: deslop comments
* review: rename, add asserts
* server : adjust checkpoint logic (#11)
* server : adjust checkpoint logic
* cont : rm asserts
* server-context: fix early exit
* spec : fix compatibility with n-gram and add TODOs (#13)
* metal : cleanup
* llama : fix faulty bitwise check in recurrent memory
* server : disable RS-based MTP in combination with other spec types
* spec : add TODOs
* cont : fix comment
* cont : update comment
* common : fix logic for ngram + mtp compat
* llama-memory: enable checkpointing with partial rollback
* cont: add test-case for loading into a dirty ctx
* llama-memory-recurrent: clear rs_idx in clear
* download: fix mtp path
* llama-arch: fix enorm op
* docs: update docs
* conversion: fix type annotations
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* ggml: backend-agnostic tensor parallelism
* support for GPT-OSS, Qwen 3 MoE
* partial Vulkan fix
* add support for 4/8 GPUs
* unconditional peer access
* re-use buffers + ggml contexts
* fix output pattern
* NCCL support
* GGML: HIP: add RCCL support
* Remove shfl and AllReduce from backend interface
* move allocation workaround out of ggml-alloc.c
* 2d tensor set/get support
* Fix the seg fault without NCCL
* Apply suggestion from JohannesGaessler
* support for tensor dims % n_devs != 0
* fix view_offs scaling
* arbitrary num. of GPUs/tensor split
* fix compilation
* better granularity estimate
* Support device-specific host buffer types if all underlying backends expose the same type. This allows using pinned memory instead of pageable memory for CUDA.
Fix compilation errors.
* partial Qwen 3 Next support
* Fix qwen3 30b (#8)
* Fix crash with Qwen-30B-A3B Q4_0
Qwen-30B-A3B Q4_0 has an intermediate dimension of 768. Using a granularity of 256 forces an uneven split between GPUs, which is not supported by the current implementation.
* Decide block size based on tensor quantization type
* Fix crashes due to KV cache serialization (#9)
KV cache serialization requires non-zero offsets on the tensor. Add support in the meta backend to set/get a tensor with a non-zero offset.
* metal : fix build (#7)
* static memory allocations, fix usage count
* fix tensor granularity
* more even memory distribution
* use BF16 for allreduce
* rebase fixup
* better error message for unsupported architectures
* Fix device mismatch during scatter of allReduce. (#11)
There is a mismatch between the dst buffer device and the backend device, causing the use of sync copies
* Enable the previous allreduce implementation. It is better in both perf and stability (#12)
* delay AllReduce for Moe for less I/O
* build : clean-up compile warnings
* backend : move most of the meta backend API to ggml-backend-impl.h
* cont : hide unused public API in the implementation
* llama : use llama_device + remove ggml_backend_dev_is_meta()
* ggml-backend : remove unused alloc include
* minor : remove regex include
* ggml : introduce ggml-ext.h for staging new APIs
* rebase fixup
* fix tests
* llama : more robust logic for determining Meta devices (#16)
* llama : more robust logic for determining Meta devices
* cont : fix devs size check
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
* cont : fix log type
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
---------
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
* disable roundtrip for meta backend
* fix arch selection
* Qwen 3.5 support
* fix Gemma 4 MoE
* fix OpenVino, SYCL
* fix test-llama-archs for CPU-only builds
* Fix Qwen 3.5 MoE
* disable meta backend tests for WebGPU
* tests : filter CPU-based devices from the Meta backend tests (#17)
* meta : formatting, naming, indentation (#18)
* formatting : llama-model.cpp
* formatting : ggml-ext.h
* formatting : ggml-backend-meta.cpp
* meta : add TODO
* add documentation
* better error messages
* fix GPT-OSS
---------
Co-authored-by: Carl Philipp Klemm <carl@uvos.xyz>
Co-authored-by: Gaurav Garg <gaugarg@nvidia.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
This commit removes the unused tmp_buf variable from llama-kv-cache.cpp
and llama-memory-recurrent.cpp.
The tmp_buf variable was declared but never used but since it has a
non-trivial constructor/desctuctor we don't get an unused variable
warning about it.
* feat(memory): Only fail partial erasure of recurrent tail
The recurrent state is always assumed to be the state as of the last update
from the final token in the sequence. When doing a partial erasure, if the
range does not include the final token, the erasure can be considered a
success since any memory used for the sequence prior to the final token
(which is no memory) has been successfully removed.
There is one potential case that this doesn't address which is the pruning
of cache to remove sensitive data from the context. This wouldn't work for
attention cache partial removal (in the middle) either since the KV state
is linearly-dependent and states in later sequence positions would still be
based on the state from the sensitive data, even if that data is no longer
cached, so I don't think this is relevant, but it is worth noting that the
semantics of this change for a partial erasure in the middle of the cache
are essentially "my context is already compressed" and not "all trace of
the removed tokens has been removed."
https://github.com/ggml-org/llama.cpp/issues/16768
Branch: HybridContextShift-16768
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>
* fix(main): Check the output of seq_rm for prefix matching
This prefix matching is explicitly attempting to remove the tokens at the
end of the sequence that don't match. This is the operation that can't be
performed on a recurrent cache due to the state being updated in place, so
if this removal fails, we need to clear the whole cache.
https://github.com/ggml-org/llama.cpp/issues/16768
Branch: HybridContextShift-16768
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>
* fix(memory): Fix condition for partial erasure failure if p0 > pos
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>
Co-authored-by: compilade <git@compilade.net>
* style: Fix extra parens
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* fix(main.cpp): Set n_matching_session_tokens to 0 on cache clear
https://github.com/ggml-org/llama.cpp/issues/16768
Branch: HybridContextShift-16768
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>
---------
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>
Co-authored-by: compilade <git@compilade.net>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* initial commit for branch 3
* generalize `swa_checkpoint` to `ctx_checkpoint`
this extends `llama-server`'s SWA checkpointing logic to include
hybrid/recurrent models such as Jamba, Granite
* oops
* disable debug prints
* keep backwards compat with `--swa-checkpoints`
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* update prompt re-processing message
* fix off-by-one error per GG
* keep `seq_rm` log per GG
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* server : fix checkpoint logic to support recurrent caches
* server : cleanup and fixes
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* server : add SWA checkpoints
ggml-ci
* cont : server clean-up
* server : handle state restore fails
* llama : add extended llama_state_seq_ API
* server : do not make checkpoints if --swa-full
ggml-ci
* llama : remove flags value for NONE
* server : configure number of SWA checkpoints with CLI arg
ggml-ci
* args : fix scope of new argument
* wip: llama : separate recurrent states from the KV cache
This will be necessary to support Jamba
(and other recurrent models mixed with Attention).
Doesn't compile yet, and finding a slot isn't yet done correctly for recurrent states.
* llama : use std::find for seq_nodes in llama_rs_cache
* llama : state checkpoints for recurrent models
* llama : correctly handle more edge cases for the rs cache
* llama : rename many llama_kv_cache_* functions
* llama : remove useless return value for some llama_cache_* functions
* llama : rethink recurrent state cell counts
* llama : begin work on support for variable GQA
This will also be useful for Jamba if we consider the Mamba layers
to have 0 KV heads.
* llama : gracefully fail when not finding hybrid slot
* llama : support Jamba
* llama : fix BERT inference without KV cache
* convert-hf : check for unprocessed Jamba experts
* convert-hf : support Mini-Jamba conversion
* llama : fix Jamba quantization sanity checks
* llama : sequence-length-aware batch splitting
* llama : use equal-sequence-length sub-batches for recurrent models
* ggml : simplify SSM-related operators
* llama : make recurrent state slot allocation contiguous
* llama : adapt internal uses of batches to llama_ubatch
* llama : fix batch split output count for embeddings
* llama : minimize swaps when reordering logits
This reduces overhead when running hellaswag
on thousands of sequences with very small 100k params Mamba models.
* llama : fix edge case finding batch seq_id of split recurrent cell
This otherwise was a problem when running the HellaSwag benchmark
with small batch sizes, making it crash.
* llama : avoid copies for simple batch splits
* ggml : make ggml_ssm_scan not modify its source tensors
* llama : fix shared recurrent tail cell count for small ubatch sizes
Otherwise it was impossible to run the 'parallel' example with '-ub 1'
with a Mamba or Jamba model.
* llama : fix .base() compilation error on Windows
* llama : allow doing the equivalent of SSM_CONV with SUM_ROWS and MUL
* ggml : allow GGML_OP_CONCAT to work on non-contiguous tensors
The implementation already supported it,
and this makes Mamba's conv step slightly faster.
* mamba : fix non-contiguous usage of ggml_silu
* llama : session saving and reloading for hybrid models
* convert_hf : fix Jamba conversion
* llama : fix mixed signedness comparison
* llama : use unused n_embd_k_gqa in k_shift
This also slightly reduces the diff from the master branch
* llama : begin renaming llama_past back to llama_kv_cache
* llama : remove implicit recurrent state rollbacks
* llama : partially apply clang-format style
* convert : fix jamba conv1d shape squeezing
* graph : add back hybrid memory graph input
But this time it contains the sub-cache graph inputs.
This *should* make it easier to handle updating the inputs
when caching the graph (eventually).
* model : add Jamba to Mamba-specific hparams printing
* jamba : remove redundant nullptr initializations
* model : remove unnecessary prefix for tensor loading constants
Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>
* model : use ggml_swiglu_split for Mamba
Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>
* model : make falcon-h1 use shared mamba2 layer builder
* memory : avoid referring to KV in recurrent cache logs
* gguf-py : avoid adding duplicate tensor mappings for Jamba
Some of the tensor names are common with Llama4
---------
Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>
Splits producing more than one ubatch per batch for recurrent models
were broken with #14512.
This fixes it by moving the completeness check after the ubatch split loop.
Concedo
Georgi Gerganov
Aman Gupta
Johannes Gäßler
Daniel Bevenius
Gabe Goodhart
ddh0
l3utterfly
compilade