* llama, mtmd: follow the reference precision of ggml_prec_set_acc calls
Requests that the reference implementation does not keep in F32 move
to GGML_PREC_BF16: the GLM4 family down and output projections, the
MiMoVL vision matmuls and the Qwen3-TTS speaker encoder conv1d. The
DeepSeek V4/V4.1 routers and the MiniMax-M3 indexer stay in F32 with
a link to the reference line that computes them in F32.
* ggml: handle GGML_PREC_BF16 the same way as GGML_PREC_F32 in backends
Backends that reject GGML_PREC_F32 on the src1 of MUL_MAT_ID now also
reject GGML_PREC_BF16 (CUDA, Hexagon, OpenCL, WebGPU, Vulkan), and
backends that accept GGML_PREC_F32 for flash attention now also accept
GGML_PREC_BF16 with the same F32 accumulation (CPU, SpaceMIT, ET).
Vulkan keeps the F32 transfer between devices for both. The CUDA
cuBLAS path already handles GGML_PREC_BF16 and is unchanged.
* tests: cover GGML_PREC_BF16 in the flash attention tests
The flash attention cases run with GGML_PREC_BF16 next to F32 and
DEFAULT, and var_to_str prints bf16 so these cases keep their own name.
* llama: request GGML_PREC_BF16 for the KQ multiplication
* cont : add TODO in clip.cpp
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* moe-cache : compute part of the misses on the CPU
--moe-cache-cpu-share F computes a share F of the MoE cache misses on the CPU
from the host experts instead of uploading them. The graph builds the expert
FFN twice: on the device from the cache slots, where the misses left to the CPU
read zero slots, and on the CPU from the host experts, where the experts
computed on the device have id -1. The two results are added before the down
bias.
- ggml-cpu: MUL_MAT_ID skips negative ids and writes zero rows
- sched: the copy callback is called for all the inputs of a split, and before
the host sync when there are no events
- the slot lookup callback reads the inputs of the CPU part together with the
selected experts, so the CPU part does not wait for the device, and its
result is copied back asynchronously
- each token uses distinct zero slots, mul_mat_id does not support an expert
selected twice by a token
- LLAMA_MOE_CACHE_TRACE writes the selected experts of each lookup, for a
simulator of the cache
* scripts : add moe-cache-tune.py to pick the MoE cache arguments
The script wraps a llama-server command: it starts the server at two cache
sizes and three CPU shares, fits the upload and CPU speeds of the machine, and
replays the routing of the runs through a copy of the cache logic to predict
other sizes and shares. It prints the recommended --moe-cache-mib and
--moe-cache-cpu-share.
docs/moe-cache.md describes the MoE cache arguments and the script.
* add am17an as codeowner for MoE cache
* moe-cache : upload all the misses until the cache is full
With a high --moe-cache-cpu-share, few misses are uploaded, so an empty cache
fills slowly and more misses than needed go to the CPU. The CPU share now
starts once all the slots of a cache group hold an expert. moe-cache-tune.py
replays the same rule.
* add backend sync for async transfer queue
* moe-cache : let the scheduler copy the results of the CPU part
* Fix MSVC compilation
* remove sync change
* llama-graph : extract the CPU share condition into can_cpu_share
* llama-graph : make llm_graph_params::moe_cache const
The graph construction no longer updates the MoE cache directly: the CPU
parts of the layers are accumulated in the graph result and applied to
the cache in llama_context::process_ubatch, after the graph inputs are
set and before the graph is computed.
Assisted-by: dsh:llama.cpp/ggml-org/GLM-5.3-Flash-GGUF (Mac Studio, M5 Ultra, 256GB)
* scripts : replay the ClassLRU policy in moe-cache-tune.py
* moe-cache : pick CPU misses by CRF instead of random (#31853)
* moe-cache : remove the backend logic of the CPU part from the graph
The CPU part reads the host experts, so the scheduler places it on the
CPU without set_tensor_backend. The cache gives no CPU map for batches
that a device would offload. Graphs with a CPU part look up their own
slot map, so the graph does not register the part with the cache.
ggml-backend : the inputs of the next split that are already computed
are copied to host memory before the current split runs, so a CPU split
does not wait for the device split before it. This replaces the stash
of the cache, and the copy callback is again called for host weights
only.
* llama-graph : return the MoE cache slots and CPU ids in a struct
The scheduler prefetch no longer checks that the next split runs on
another backend: the inputs of a split are never on its own backend,
so the input loop already skips that case.
---------
Co-authored-by: Gaurav Garg <gaugarg@nvidia.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
Co-authored-by: Alexey Kopytko <alexey@kopytko.com>
* model: add Kimi-K3.1-Pro text model
K3 with the full-attention MLA layers replaced by DSA: on this checkpoint 24
of the 93 layers carry a lightning indexer (32 heads x 128, top_k 2048, no
rope). The KDA (linear-attention) layers and the rest of the K3 skeleton -
cross-layer residuals, latent MoE, situ, MLA output gate, full-rank KDA gate -
are unchanged, so this reuses the K3 graph and converter and adds the indexer.
Indexer:
- the selection is an additive mask over the whole cache, not a gather of
the top-k rows, so it reuses the existing sparse flash attention path;
off that path the mask stays correct but still costs a kq over the cache
- the indexer key LayerNorm takes its eps from the model's rms_norm_eps,
because dsa_config declares no eps of its own
Graph and cache:
- new arch with four indexer tensors per DSA layer, and a hybrid memory
that keeps a K-only MLA cache for those layers, recurrent state for the
KDA layers, and the indexer key cache
- the graph node budget shares the K3 entry
Conversion:
- KimiK31Model(KimiK3Model), picked by KimiK31ForConditionalGeneration
- dsa_config becomes attention.indexer.{head_count,key_length,top_k,types}
plus the indexer LayerNorm eps; the layer list is 1-indexed like
linear_attn_config's, and the converter asserts that it covers exactly
the full-attention layers
- the routed experts stay MXFP4 and are repacked as in K3, not dequantized
- the vision tower and mm_projector are skipped: text only for now
- the arch key follows the checkpoint's own model_type ("kimi_k31") and the
loader registers KimiK31ForConditionalGeneration; the release name carries
a -Pro suffix that nothing in the config or the tensor names uses
Quantisation: the indexer's q_b, k and proj are kept out of the quant
overrides, since the top-k selection is sensitive to their precision.
Verified on an 8-layer slice against the HF checkpoint: 38 tensors checked
exactly, the indexer metadata against the config, and the selection against
full attention - identical PPL at top_k beyond the context length, divergent
once top_k is forced below it. Arch test fixture and the full arch suite pass.
* converter: avoid clutter in the k3 base
* narrow comment to match model
* update metal fusion expectations
* restore
* sync with PR 31733 refactoring
* graph: store the conv state of every rollback arch in one set_rows
The rollback slots of the causal conv history were written with one
copy per slot in qwen4exp, glm5-next, kimi-k3, bailingmoe3, lfm and
mamba2, while delta_net already wrote them with a single set_rows.
build_conv_state_store now does this for all of them: one copy without
rollback, otherwise one set_rows of the overlapping windows into the
min(n, n_rs_seq + 1) slots a rollback can read.
The slot order {n_rs_seq, ..., 0} is computed once per graph and each
conv layer takes a view of its tail. Its size does not depend on the
ubatch, so the reserved graph covers every ubatch shape, and the graph
input that held it is gone.
A model that is not causal never rolls back, so the context clamps
n_rs_seq to 0 for it, like for an arch without rollback support.
* nits: compute the conv slot order in the graph and keep the n_head name of bailingmoe3
* graph: build the conv slot order in each conv layer
The slot order {K-1, ..., 0} of the conv state store was one tensor
shared by every conv layer of the graph, built on the device of the
first one, so each layer on another device copied the view of its
tail. Each conv layer now builds its own order next to its set_rows,
and llm_graph_input_rs no longer holds it.
* delta_net: access conv state via views
* fix sm tensor
* delta_net: fill the conv slot index on the host
The rollback slot index of the conv windows is a constant per graph
shape. It is now an I32 input of the recurrent memory filled on the
host, instead of an arange and an argsort computed on the device, and
the meta backend no longer needs a mirrored split for ARANGE.
---------
Co-authored-by: Pascal <admin@serveurperso.com>
The dspark block mask is now derived at input-fill time: llm_attn_mask::block_uniform
fills every token of a sequence with the mask row of its first token, so the
graph-level reshape/repeat in dflash is gone and cnv always aliases raw.
Assisted-by: dsh:llama.cpp/ggml-org/GLM-5.3-Flash-GGUF (Mac Studio, M5 Ultra, 256GB)
* llama : use llm_attn_mask and kv_rows for unified attention
Assisted-by: dsh:llama.cpp/ggml-org/DeepSeek-V4-Flash-Vision-Exp-GGUF (Mac Studio, M2 Ultra, 192GB)
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* cont : optimize `get_seq_cells`
* llama : build the kv_rows lists from per-sequence cell bitmaps
Keep a bitmap over the cells for each sequence (the transpose of `seq`), updated in seq_pos_inc/dec, and build the cell lists by walking its set bits. The lists stay in cell order, cells shared by several sequences go to every list, and nothing has to be rebuilt when cells are removed.
Also leave out of the lists the cells that the SWA hides from every token of the slice (standard and chunked SWA).
* cont : make get_kv_rows_p_min private
* cont : minor metal
* llama : use kv_rows for every ubatch of a unified cache
A ubatch made of equal-length runs with a single sequence each (f.ex. the drafts of several slots) now splits into per-sequence slices.
Any other ubatch (interleaved, uneven or coupled sequences) becomes a single slice over the cells of all its sequences instead of falling back to the dense KQ mask, so the graph keeps the topology of the reserved graph.
* llama : merge get_kv_rows_seq_ids and get_kv_rows_p_min into get_kv_rows_slices
Assisted-by: dsh:llama.cpp/ggml-org/GLM-5.3-Flash-GGUF (Mac Studio, M5 Ultra, 256GB)
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* WIP OCP FP8 E4M3 support
* Fix python type-checker
* Materialize input_scales also for FP8 when converting
* Fix conversion of https://huggingface.co/RedHatAI/Qwen3.8-27B-NVFP4
1. Per row FP8 wrongly fell to NVFP4 conversion, as NVFP4 checked on
dims alone. Also check on dtype for FP8
2. Need to reshape FP8 QKV projections scales in the same way that
weights are reshaped
* Allow remaining FP8 tensors to be dequantized during conversion
* Allow tensor-strategy for FP8-only compressed-tensor ckpts
* Fix and test gguf-py behavior for FP8_E4M3
* Disallow QKV-fusion for FP8 temporarily
* Fix conversion of fused qkvz & FP8 for Qwen3-Next
* Avoid fusing FP8 for --fuse-gate-up-exps as we may have different ws
* Emit warnings when not fusing FP8 during conversion
* More FP8 conversion fixes
* Add supported for strided copies to/from FP8 in CPU
* remove unneeded scratch buffer from FP8 cpu cpy/dup path
* add explanatory comments and TODOs
* Add dedicated encoder and decoder tests for FP8
* Whitelist ggml_op_get_rows in metal instead of blacklisting
required to fix failing CI, and white-listing should be more
future-proof than adding FP8 to the black-list
https://github.com/ggml-org/llama.cpp/actions/runs/38433916498/job/115422880633?pr=27512
* Fix more conversion bugs
* Defer weight materialization during conversion
* Blacklist fp8 in metal backend instead of white-list
* Adjust indents
* Fix lint + execute tests
* Add fast CPU kernel for F8_E4M3
* Conversion fixes
* Fix remaining gaps
* Enable tied weights with shared scales
* Make mandatory-F32 rules apply to FP8 during conversion
* Fix python type checks
https://github.com/ggml-org/llama.cpp/actions/runs/39765670052/job/119780379595?pr=27512
* Fix kaleidai unit-test failures
https://github.com/ggml-org/llama.cpp/actions/runs/39765670125/job/119780383150
* Relax numeric precision constraints for pre-check
* Leave fast-FP8 CPU on RISC-V for follow-up
We already exclude arm from the fast-path
* Allow output scale tensors to be located in any gguf shard
* Move test_conversion_fp8.py test execution to behavior-check.yaml
* Apply batched suggestions from code review
Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@huggingface.co>
* Move mixed compressed-tensor logic into dequant_fp8 + add unit-tests for it
* Install all requirements also on Windows for behavior-check
Fixes https://github.com/ggml-org/llama.cpp/actions/runs/40535852590/job/122300758223?pr=27512
* Unmap lazy tensors before clean-up in Windows
Should fix https://github.com/ggml-org/llama.cpp/actions/runs/40539797824/job/122314782459?pr=27512
* LF line endings
https://github.com/ggml-org/llama.cpp/actions/runs/40539798168/job/122314791118?pr=27512
* Extract FP8 tests from llama-arch to its own test-fixture
See https://github.com/ggml-org/llama.cpp/pull/27512#discussion_r4400628221
* Update gguf-py/gguf/scripts/gguf_convert_endian.py
Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@huggingface.co>
---------
Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@huggingface.co>
* model : add LiquidAI/d1-2B and LiquidAI/d1-omni-2B
Assisted-by: Claude Opus 5.5
* convert : keep the speech task templates out of the d1 GGUFs
Assisted-by: Claude Opus 5.5
* small refactoring: no_batching
---------
Co-authored-by: Xuan Son Nguyen <son@huggingface.co>
* llama : build the k-pool sparse selection independently of the kq_mask shape
qwen4exp (QSA) and glm5-next pick cache cells by index and no longer read the kq_mask values, but they still shaped the selection like the dense kq_mask. Build the selection as [n_kv, n_tokens] over the cache cells instead, and split it into the streams of the K view right before the attention.
This lets the dense mask change its layout (e.g. per-sequence row lists) without touching the sparse layers.
* improve comment
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* llama : drop the unused build_fused_top_k_mask args, review nits
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* model : add DeepSeek-V4.1-Flash (deepseek41)
* deepseek41 : remove MXFP8, convert the Engram table to BF16
* deepseek41 : compute the compressed-KV read layers once in the cache
* deepseek41 : use get_cells() for the shared-history check, drop llama_kv_cache::seq_pos_tok_le
* deepseek41 : gather the Engram rows in-graph like qwen4exp, drop the host gather
* deepseek41 : keep the Engram hash constants in hparams.ple_*, one module per Engram layer
* deepseek41 : drop the LLAMA_DSV41_KV_MIRROR toggle
* deepseek41 : keep il last in build_hc_mixes
* deepseek41 : sort the top-k rows before the sparse gather
* tests : add deepseek41 to the Metal fusion baseline
---------
Co-authored-by: Daniel Han <danielhanchen@gmail.com>
The amax POOL_1D reduction spans the whole matmul row, so a device cannot
compute it from an axis-0 split. The Meta split-state case reported
MIRRORED, which gives each device a partial max of its own slice. Drop
that case and deny -sm tensor for the arches which build the amax graph,
see [TAG_MOE_AMAX_SCALE].
Assisted-by: dsh:llama.cpp/ggml-org/Qwen3.8-Flash-Next-GGUF (MacBook, M5 Max, 128GB)
* qwen4exp : build the QSA selection without the kq_mask
* clean
* lang
* qwen4exp : reshape the fused QSA selection to the kq_mask streams
ggml_top_k_mask with a single keep row returns [n_kv, n_tokens], while
build_attn_qsa expects the kq_mask layout [n_kv, n_tokens/n_stream, 1, n_stream].
With several streams the shape assert failed at graph reserve.
* test the QSA optimization
* reword
* update build_fused_top_k_mask instead of handwiring
* simplify
* keep il at the end for consistency
* Add Flint model (FlintForCausalLM). Assisted by Codex
* cleanup
* one more
---------
Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@huggingface.co>
* llama: create new fused op for QSA scoring
* enroll glm5.3-next too
* tests : initialize the remaining tensors in test_kpool_mask
The harness sentinels were left uninitialized, so leftover memory could read as NaN and fail the NaN check.
* ggml: generalize the k-pool mask op into ggml_top_k_mask
The k-pool mask op only covered the pooled QSA selection of qwen4exp
and glm5-next, while the DSA models and the DeepSeek V4 CSA build the
same top-k mask by hand with fill, set_rows and add.
ggml_top_k_mask(mask, top_k, score, blocks, extra) keeps the mask
cells of the live picks and the extra cells, and sets the rest to
-inf. Every row tensor is 4D in the layout of top_k, and the mask is
broadcast to those rows, so a selection can also be made per head.
score, blocks and extra are optional: without score every pick is
live, without blocks a pick is a cell, and extra holds the cells a
row keeps without a pick.
llm_graph_context::build_fused_top_k_mask views the row tensors in
the layout of kq_mask and registers the fused node. The k-pool
models, the DSA attention of build_attn and the DeepSeek V4 CSA mask
use it when the probe enables the op, and keep their current path on
backends without it.
---------
Co-authored-by: Pascal <admin@serveurperso.com>
* llama : scale moe down projection per expert token
Replace the global amax rescaling for the MoE down projection with a per-expert-token factor computed by a POOL_1D max reduction along the intermediate dimension. Undo the scale on the output with the same exact power-of-two factor. Handle POOL_1D in the Meta split-state logic so a split along axis 0 becomes a replicated per-device partial result, enabling tensor parallelism.
Assisted-by: pi:llama.cpp/ggml-org/DeepSeek-V4-Flash-Vision-Exp-GGUF (Mac Studio, M2 Ultra, 192GB)
* cont : simplify comments
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* llama : refactor amax src1 scaling result
Return the rescaled activation and the optional scale tensor through a small struct instead of an output pointer.
Assisted-by: pi:llama.cpp/ggml-org/DeepSeek-V4-Flash-Vision-Exp-GGUF (Mac Studio, M2 Ultra, 192GB)
* cont : add TODO + clean-up
* model: add Muse Spark Open (text, vision and audio)
Muse Spark Open is a latent space MoE: the residual stream stays at 8192
while the 256 experts live in 4096, with a projection down and up around
them, top-8 sigmoid routing with a correction bias. Attention interleaves
three sliding layers at a 2048 window with rope for one global layer
without rope, with a sigmoid output gate and a weightless QK-norm.
Residuals are norm preserving, their gate is folded at conversion time.
Vision reuses the Muse Glimmer tower. Audio is a causal convolutional
pyramid with no attention and no spectrogram, four stages folding 16, 6,
5 and 4 samples into the channel dim, one embedding per 80 ms of 24 kHz.
The quantizer keeps the latent projections and the attention gate at
Q8_0, a little over 1% of the weights, since the whole residual stream
passes through them and the low-bit types turned the output into a
repeating loop. The model saver writes the post norm epsilon, and the
arch is registered in test-llama-archs.
Co-authored-by: Xuan Son Nguyen <son@huggingface.co>
Co-authored-by: Pedro Cuenca <pedro@huggingface.co>
* Muse Spark: drop the redundant forward expand
* Muse Spark: drop the redundant forward expand and tighten the comments
* Update src/llama-quant.cpp
Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@huggingface.co>
* Update src/llama-graph.cpp
Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@huggingface.co>
* Update conversion/muse_spark.py
Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@huggingface.co>
* Update conversion/muse_spark.py
Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@huggingface.co>
* Update conversion/muse_spark.py
Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@huggingface.co>
* Update conversion/muse_spark.py
Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@huggingface.co>
---------
Co-authored-by: Xuan Son Nguyen <son@huggingface.co>
Co-authored-by: Pedro Cuenca <pedro@huggingface.co>
Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@huggingface.co>
An embedding-only graph has no logits, and build_pooling() returns
early when embeddings are off, so since the models no longer expand
t_embd the graph ends up empty and its inputs are never allocated.
Expand t_embd in that case.
The no-cache attention pads K, V and the mask to 256, but T5 also adds
a relative position bias of [n_tokens, n_tokens], so the add with KQ
fails unless n_tokens divides the padded width. Pad the bias with
zeros up to the KV dimension of the mask, like K and V.
* graph: skip the LM head for pooled embeddings
A context that extracts pooled embeddings outputs one vector per
sequence, but every model still expanded its LM head, so the graph
computed n_vocab logits per token and the context kept a host buffer
for them.
The models now only declare t_logits, and set_outputs() expands it
unless the context extracts embeddings with a pooling type other than
NONE. Since the toggle changes the graph shape for a pooled context,
set_embeddings() schedules a re-reserve in that case.
* graph: drop the redundant t_embd expand
Address review from @CISC: with embeddings on, build_pooling() already
expands its output for every pooling type, and a decoder graph without
embeddings does not read t_embd, so set_outputs() only expands
t_logits.
* graph: add build_logits() for the LM head output
Address review from @ggerganov: the LM head is expanded by a dedicated
build_logits(), called from llama_model::build_graph() like
build_pooling(), and set_outputs() only marks the outputs again.
* graph: call build_logits() before build_sampling()
Address review from @ggerganov: build_sampling() reads t_logits, so the
LM head is built first. A pooled context then skips the sampling layers
too, since it has no logits to sample from.
* nit
The no-cache path builds a [n_tokens, n_tokens] mask and passes K and V
unpadded, so FLASH_ATTN_EXT sees K->ne[1] == n_tokens. The CUDA kernels
for head size 512 require a multiple of FATTN_KQ_STRIDE, so the op falls
back to the CPU for most lengths, and the -fa auto probe (1 token)
disables flash attention for such models.
Pad the mask like llama_kv_cache::get_n_kv and pad K and V with zeros up
to the KV dimension of the mask. clef builds its own causal mask and
pads it the same way.
* model : use exact GELU for ModernBERT encoders
Assisted-by: Codex
* model : keep tanh GELU aliases on ggml_geglu
Assisted-by: Claude Opus 5.5
* model : map gelu_python to ggml_geglu_erf
Assisted-by: Claude Opus 5.5
The reserve builds n_outputs_max_per_seq sampling chains per sampler,
while a decode built one per output row, so the graph changed its
topology after the reserve and GGML_SCHED_NO_REALLOC builds aborted on
the next same sized graph. Every sampler now builds
n_outputs_max_per_seq chains, the ones without a row of the ubatch on
the padding row and not selected, and graph_max_nodes counts them.
* model : support classifier_activation for rerankers
Assisted-by: Claude Opus 5.5
* model : map classifier gelu to gelu_erf and accept tanh
Assisted-by: Claude Opus 5.5
* model : default act_cls to tanh, ModernBERT falls back to gelu_erf
Assisted-by: Claude Opus 5.5
* llama: support both embd + raw tokens in batch
* add to test-llama-archs
* also check case llm_arch_supports_mixed_batch = false
* constant graph topology
* have dedicated input for mixed case
* rm set_tensor_backend
* is_embd --> type
* consolidate m-rope pos handling into one place
* nits
build_rs gathered the extra states (n_rs - n_seqs rows) with their own
get_rows. The worst-case reserve has n_rs == n_seqs, so that node was
sized at zero rows, and any ubatch whose cells are not contiguous forced
a graph reallocation at an unchanged node count, which aborts under
GGML_SCHED_NO_REALLOC.
A single get_rows now gathers the n_rs states: the ubatch states and the
extra states are views of it, and its size only depends on n_rs, which
the reserve already sets to the maximum. A custom getter (mamba ssm_scan)
gathers from the second state, so a single sequence ubatch copies no
state. The views are built once per graph in the input to keep the host
overhead of the graph unchanged.
* CUDA: Handle compute type for NVFP4 on cublass path
Signed-off-by: ynankani <ynankani@nvidia.com>
* Use BF16 compute type for quantized models if HW allows
Signed-off-by: ynankani <ynankani@nvidia.com>
* Set acc prec to bf16 for nvfp4 as it needs atleast bf16 range
Signed-off-by: ynankani <ynankani@nvidia.com>
* Update ggml/src/ggml-cuda/ggml-cuda.cu
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
* preserve op_params for per-expert matmul
Signed-off-by: ynankani <ynankani@nvidia.com>
---------
Signed-off-by: ynankani <ynankani@nvidia.com>
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
* Rebase GLM-Next support onto master, and migrate to llama-memory-hybrid-idx
* Add initial MTP support
* Merge branch optimizations. Reduce allocated compute buffer size, speed up long context decode, fla, and slight MTP improvements.
* Review driven changes, remove env vars, protect tensors
* Strip MTP for initial PR
* Clean up after mtp strip
* Clean up after mtp strip
* Update speculative.cpp
* Update llama-context.h
* Clean up after mtp strip
* Fix tokenizer ignore merges
* Improve quantization protection selection
* Refactor mhc helpers, graph base
* Lint Fixes
* Apply suggestions from code review
Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@huggingface.co>
* Skip glm5-next in model saver, fix CRLF
* Skip glm5-next in sweep
* Remove T4 fallback
* Review cleanup
* Review suggestions
* Defer separate MTP gguf handling to MTP PR, drop filter
* Repad n_head_kv
* kpool init apply
* Order by descending score
* Drop guard
* read kpool from hparams, clarify kpool cache flags, remove kpool_build_state(nullptr)
* Add glm5-next support to model saver and add arch test fixture
* Review cleanup
* Kpool pooled caching clarify
* Add multi stream support
* Finish Rebase
* Sparse FA fir DSA prefill
* Const
* Update llama-model.cpp to fix rebase error
* gguf-py : merge tensor map entries for HC tensors
* model : use build_gdn_l2_norm in GLM5_NEXT implementation
* chore : remove trailing whitespace
* model : use new OP precision setting API in GLM5_NEXT implementation
* mtmd : use ggml_swiglu_clamp in GLM5V and apply the image token limit
The two clamps around swiglu_split are what ggml_swiglu_clamp already does,
so the clamp bounds collapse back to one value. GLM5V also never called
set_limit_image_tokens(), so --image-max-tokens had no effect.
Assisted-by: Claude Opus 5
(cherry picked from commit 46d18e12d422be4cc04a70e4a9a9e0168bb3d5b7)
* llama : keep the GLM5-Next k-pool layout across ubatches
The layout was rebuilt from a full cell scan on every ubatch. Pools are fixed
by the positions relative to the sequence's first one, so the layout now lives
on the memory and a ubatch only appends to it.
A sequence edit no longer stales every pooled key either, only the ones at or
after the edited position, which makes a tail seq_rm free. The pooling subgraph
is built unconditionally so the graph shape no longer changes every kpool
tokens, and the pool axis is folded into rows before soft_max, which otherwise
exceeds the CUDA gridDim.y limit past n_kv 262144.
Assisted-by: Claude Opus 5
(cherry picked from commit 5d1c40b93e17fddbf73b785efe43e0d02ccb3977)
* model : write the GLM5-Next recurrent rollback checkpoints
The conv state and the delta net state were only written to the live row, so a
rollback restored whatever the checkpoint rows happened to hold. Take the same
route as kimi-k3: build_recurrent_attn for the state, and write all K_rs conv
groups. That also drops a state view that assumed contiguous rows.
Enroll the arch in test-recurrent-state-rollback, which catches this under its
garbage-filled cache pass.
Assisted-by: Claude Opus 5
(cherry picked from commit 5ace37e86d5d448e83ef5dde5632c748185b18cd)
* llama: fix PR #27773 test-save-load-state restore failure
Clear the attention and indexer cache data after a failed hybrid state restore so restored NaNs cannot affect a later sequence.
Assisted-by: Codex
* llama: fix PR #27773 gpu-rocm graph reallocation
Reserve the full GLM5-Next pool capacity and dirty pool count. The gpu-rocm Test step aborts when n_new grows while the graph node count stays fixed; CUDA, Vulkan, Metal, and WebGPU checks report the same error.
Assisted-by: Codex
* llama : fix GLM5-Next k-pool layout staleness after edits and shared teardown
Two defects in the cross-ubatch k-pool layout added by the k-pool commit:
1. Wrong results. An edited sequence only rebuilt its pool layout when its cell
count changed, so if the first ubatch after an edit added back exactly as many
cells as were removed, the stale position-to-cell list survived. With a unified
cache and more than one sequence, where another sequence takes the freed cells,
the reused layout points at the wrong cells (CPU: large logit drift, CUDA: NaN).
Rebuild whenever the sequence is stale, not only on a size mismatch.
2. Slowdown. "shared" mode was assumed to end only with an edit that forces a
rebuild, but sharing also ends when the other sequence is removed. The survivor
kept shared = true, pinning cache_safe off and re-pooling every pool on every
ubatch (server trigger: n>1 completions with -kvu, via the seq_cp in
copy_state_to). In seq_rm, if the layout has shared cells, stale every sequence
so one rebuild re-derives sharing and cache_safe returns to 1.
Assisted-by: Claude Opus 5
* llama : fix build_attn_mha stream stride for non-contiguous q
build_attn_mha split the batch into streams with a stream stride of
q->nb[3]/n_stream. That only equals one stream's span, (ne[2]/n_stream)*nb[2],
when q is contiguous. GLM5-Next is nope-only, so it does not concat a rope part
and passes the permuted q_absorbed straight in, where nb[3] != ne[2]*nb[2]; the
stride was then n_head times too large and every stream s >= 1 read another
head's queries. Split-KV (-np N without --kv-unified) multi-stream prefill was
wrong for every stream past the first. Unified KV and decode were unaffected
(n_stream == 1, and decode takes the gather path). Other MLA models concat rope
so q is contiguous and the computed value is unchanged for them.
Compute the stride from the token dimension, which is identical for a
contiguous q.
Assisted-by: Claude Opus 5
* llama : re-derive GLM5-Next k-pool sharing on state_read/state_drop
The shared-cell teardown added to seq_rm (stale every sequence when the layout
has shared cells, so a survivor does not keep shared = true and pin cache_safe
off) was missing from the other paths that can free shared cells: state_read
and state_drop staled only the one sequence. Apply the same re-derivation there
and correct the comment that claimed sharing ends only via an edit or seq_rm.
Assisted-by: Claude Opus 5
* quant : drop duplicate GLM5-Next hc_ filter
The hc_ name filter was listed twice in the GLM5_NEXT protection block.
Assisted-by: Claude Opus 5
* glm5-next: scope K-pool cache access to indexed operations
* glm5-next: keep K-pool access in hybrid index memory
* glm5-next: keep mHC graph builders model-local
* glm5-next: mark only touched pools per ubatch
---------
Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@huggingface.co>
Co-authored-by: Stanisław Szymczyk <sszymczy@gmail.com>
Co-authored-by: Piotr Wilkin <ilintar@gmail.com>
* model : support classifier_pooling for ModernBERT rerankers
Assisted-by: Claude Opus 5.5
* model : read classifier pooling type in load_hparams
Write classifier.pooling_type from _try_set_pooling_type whenever the
config has classifier_pooling, and read it in
llama_model_base::load_hparams. ModernBERT falls back to mean when it
is unspecified.
Assisted-by: Claude Opus 5.5
* conversion : only accept cls and mean for classifier_pooling
Assisted-by: Claude Opus 5.5
* model : rename classifier_pooling_type to pooling_type_cls
Assisted-by: Claude Opus 5.5