mirror of
https://github.com/ggml-org/llama.cpp.git
synced 2026-09-01 09:37:35 +02:00
b10667
40
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
|
6c84c7d5d8 |
model: add Qwen3.8-Flash-Next (qwen4exp) (#27742)
* gguf: add qwen4exp (Qwen3.8-Flash-Next) arch and converter
Adds the GGUF-side plumbing for HF model_type qwen4_exp:
- MODEL_ARCH.QWEN4EXP plus tensors for the low-rank hyper-connection
variant (hc_*_norm/down/up/inject) and the PLE n-gram hash embeddings.
The DeepSeek-V4 hc_*_fn/base/scale tensors are a different
parameterisation, so these are separate entries rather than reuse.
- Reuses the existing indexer, per_layer_token_embd, SSM and
compress_ratios keys unchanged.
- conversion/qwen4exp.py inherits the Qwen3.5 linear-attention V-head
reorder and interleaved mrope, concatenates the 128 PLE embedding
shards, and splits index_qk_proj into separate indexer q/k tensors.
The PLE hash multipliers reach ~2.4e13. prepare_tensors() casts every
non-float dtype to float32 before modify_tensors() runs, and GGUF array
writes infer INT32 from Python ints, so both paths are bypassed: the
constants are read from the pre-cast lazy tensors and written as
explicit UINT64 arrays.
Additive only; no existing arch changes behaviour.
* llama: load qwen4exp (Qwen3.8-Flash-Next) hparams and tensors
Adds LLM_ARCH_QWEN4EXP with its hparams and tensor loading. The graph
comes in the next commit; this makes the model load and report correct
metadata.
- hyper-connections set n_embd_out_impl = hc_count * n_embd, so the
residual stream is 4x wide and there is no output_norm: the final
mixer's hc_norm is the last norm in the model.
- registered as hybrid and given the same recurrent/attention memory
filters as Qwen3-Next and Qwen3.5.
- reuses the existing indexer, per_layer_token_embd, SSM and
compress_ratios keys as-is.
- the PLE table row count is read back from the file rather than
recomputing the vocab padding rule.
llama-model-loader gains UINT64 array support. That branch previously
threw, so no existing caller changes behaviour; it is needed because the
PLE hash multipliers do not fit in int32.
* qwen4exp: shorten comments
* llama: qwen4exp text graph with hyper-connections, GDN and MoE
Implements the decode graph for Qwen3.8-Flash-Next: the hyper-connection
residual stream, gated delta net layers, the MoE block with its gated shared
expert, and dense full attention. The QSA indexer and the PLE n-gram embedding
are not wired up yet and land in later commits.
Hyper-connections are implemented here rather than shared with deepseek4.cpp.
The two formulations agree on the [n_embd, hc, n_tokens] layout and little
else: DeepSeek-V4 mixes with a full-rank projection and Sinkhorn-normalises
it, whereas this model uses a low-rank down/silu/up sigmoid gate and collapses
by a plain mean. Only the ~10 line stream mean is genuinely common, so sharing
would mean touching DSV4's hot path and its three fused CUDA ops to reuse very
little. What is reused is the substantive part: the LLM_KV_HYPER_CONNECTION_*
keys, the n_embd_out_impl wide-residual support already in the loader, and the
layout convention.
Also allows a checkpoint to carry no PLE layers at all, which makes it
possible to bring the graph up and validate it in stages.
Validated against vLLM, the only working reference implementation. On a
scaled-down model with an init scale large enough to give non-uniform logits,
agreement with vLLM sits at the numerical noise floor: llama.cpp f32 against
its own bf16 gives 84.3% top-1 agreement over 255 positions, and this graph
against vLLM gives 85.1%. The comparison was calibrated by seeding three
deliberate bugs (silu instead of sigmoid on the delta net gate, dropping the
1/hc scale in the mix, dropping the 2x in the combine); each drops top-1 to
between 0% and 11%, an order of magnitude below the floor.
* llama: qwen4exp PLE n-gram hash embedding
Adds the per-layer embedding: a custom I32 graph input hashes each token with
its ngram_size-1 predecessors host-side and the result is a plain row gather
over the shared table, the same shape gemma3n's per-layer embedding uses. The
hash has to run on the host because the splitmix64-derived multipliers reach
2^45, so the products need 64-bit integers and an xor, neither of which ggml
has.
Predecessors that fall outside the ubatch come from a small per-sequence
history on the model, mirroring the per-request ngram_context the reference
carries. It is only trusted when contiguous with the incoming position, so a
fresh prompt or a rewound cache falls back to EOS padding rather than hashing
against stale tokens.
The depthwise conv is written out as a sum of shifted, per-channel-scaled
copies rather than through ggml_conv_1d_dw, which carries a correctness
warning upstream.
Verified two ways. The row indices match a transcription of the reference's
tensor formulation exactly, 1024 of 1024 rows, including sequences with EOS
tokens sprinkled through them to exercise the segment reset. Separately, with
PLE placed on layer 0 so its input is just the token embedding, ple_embd and
ple_gated_value match a PyTorch computation from the same checkpoint to every
printed digit.
End to end over 1023 scored positions the port sits the same distance from
vLLM with PLE as without it, 6.3 points of top-1 against 6.0, so PLE costs no
accuracy relative to the rest of the model. That common offset is vLLM's bf16
activations, which cannot be removed: its QSA kernel refuses float32.
Two bugs found along the way, both caught by the row-index check. The history
was read and updated in the same pass, so a token early in a ubatch could pick
up an earlier token of that same ubatch as prior context; it is now snapshotted
first. And an EOS token was cutting its own context, where the reference takes
the last EOS strictly before the position, so a boundary only hides tokens from
the positions after it.
Known gap: the conv carries no state across ubatches, so it is exact only for a
prefill that starts at position 0. Chunked prefill and decode need the conv
state wired into the recurrent memory, and the conv branch itself is still
numerically unverified because the fixture zeroes its weights.
* llama: carry the qwen4exp PLE conv state across ubatches
The PLE depthwise conv was zero-padding on the left, which is only right for a
prefill that starts at position 0. Decode and chunked prefill saw a truncated
history for the first (kernel-1)*ngram_size positions of every ubatch.
The PLE module sits on a layer that is also a delta-net layer, so both need a
conv history in the same recurrent row. Rather than plumb a per-layer state
size through build_rs and build_conv_state, the row is widened once and each
convolution addresses its own slice through a local helper. n_embd_r() gains
the extra span, which is zero for every other architecture because it is
derived from ple_n_heads.
Verified by feeding the same 1024 token sequence in chunks instead of one
shot: at 64 tokens per decode the logits are bit-identical to the single-shot
run, 1023 of 1023 top-1 and a maximum logprob deviation of exactly zero. At
one token per decode they differ slightly, but the no-PLE model differs more
under the same test (94.6% against 97.1%), so that is the usual gemv-versus-
gemm accumulation difference and not the state.
The conv branch is also no longer unverified. With non-zero conv weights the
port sits 6.3 points of top-1 below the numerical floor, the same distance as
with the weights zeroed and as the model with no PLE at all, so the branch
adds no error of its own.
test-llama-archs passes every existing architecture at 0.00e+00, including the
delta-net models that share this code path.
* llama: fix the qwen4exp PLE conv state and unblock test-llama-archs
build_rs writes into the state tensor in place, zeroing one row and copying the
carried-over states, so calling it twice for the same layer let the second call
clobber the first write-back. The PLE layer is also a delta-net layer, so that
is exactly what happened: both convolutions gathered the same row. They now
share a single gather per layer.
The earlier claim that the conv state was carried correctly was tested on a
fixture whose conv weights are zero, where the branch contributes nothing and
chunking matches trivially. Re-running with non-zero conv weights showed the
divergence, growing with the number of ubatch boundaries: 97.1% top-1 at one
boundary down to 90.2% at seven. With the shared gather it is bit-identical to
the single-shot run at every chunk size tried, 512, 128 and 64, with a maximum
logprob deviation of exactly zero over 1023 positions. The delta-net-only model
stays bit-identical too, so nothing regressed there.
Also derive the delta-net conv channel count the way load_arch_tensors sizes
wqkv instead of from ssm_d_inner. The two agree for this model, but n_embd_r()
only bounds the row and the convolution has to match the tensor feeding it.
test-llama-archs previously aborted on this architecture and took every later
architecture with it. qwen4exp is marked MoE-only, given the hyper-connection
keys and an ssm_d_inner consistent with its tensor derivation, and skipped for
now: the hyper-connection keys written by get_gguf_ctx are not reaching the
synthesised file, which needs a separate look. The suite completes again, 124
architectures at 0.00e+00.
* llama: optional indexer key cache in llama_memory_hybrid
Groundwork for qwen4exp's QSA sparse attention. Its indexer needs a per-token
key history for the full-attention layers, but a hybrid model cannot use
llama_kv_cache_dsa: that class derives from llama_memory_i rather than
llama_kv_cache, and llama_memory_hybrid constructs its attention cache
directly. No existing architecture pairs recurrent state with a sparse
indexer, so there was nothing to reuse wholesale.
llama_memory_hybrid therefore gains a third, optional cache, shaped the same
way llama_kv_cache_dsa shapes its lightning-indexer cache: a copy of hparams
with n_head_kv forced to 1 and n_embd_head_k_full set to indexer_head_size.
It is built only when a filter_idx callback is passed, which defaults to
nullptr, so every existing architecture gets exactly what it got before. The
per-sequence operations and the batch preparation forward to it under a null
check, matching how the DSA cache prepares its two caches over the same
ubatches.
test-llama-archs passes all 124 architectures at 0.00e+00, including the 12 in
the hybrid family that share this code. The qwen4exp fixtures are unchanged:
same logits against vLLM, and chunked evaluation still bit-identical to
single-shot.
* llama: QSA sparse attention for qwen4exp
The full-attention layers of this model do not attend to everything. An
indexer scores one mean-pooled key per block of compress_ratio tokens and
keeps a budget of the best blocks, plus the tail of tokens that do not yet
form a complete block. Below indexer_top_k + compress_ratio - 1 cached
tokens every block fits in the budget, so the result is exactly dense.
What is reused rather than rebuilt:
- the mask machinery. build_attn's DSA overload already turns a list of
token indices into a KQ mask via ggml_set_rows, so that block is lifted
out verbatim into build_attn_mask_top_k and shared with a new overload
on llm_graph_input_attn_kv. DSA's node sequence is unchanged; the new
overload exists because llama_kv_cache_dsa assumes MLA and cannot be
dropped into a hybrid model.
- the indexer key cache, which is the optional third cache added to
llama_memory_hybrid in the previous commit. It holds raw keys, because
pooling happens before the norm and the rotation.
The graph expands block scores rather than block indices: giving every
token of a block its block's score needs only a gather, where expanding
indices would need an integer multiply-add that ggml has no op for. Since
the budget is a whole number of blocks and a block's members tie exactly,
the cut still lands on a block boundary.
Everything that depends on cache layout is computed host-side in
set_input_qsa. Blocks are cuts of the position line rather than of the cell
array, so nothing assumes the cache is contiguous.
Measured on the tiny fixture against vLLM, comparing the selected token
indices directly rather than the logits:
below the budget selection identical, and 1024-token logits are
bit-identical to the pre-QSA dense path
above the budget mean jaccard 0.975
The direct index comparison is what made this correct. The reference
rectifies each head's dot product before summing over heads, which an
earlier reading of it had missed; on logits alone the resulting port looked
fine, because on a randomly initialised fixture the known-correct dense
path already disagrees with vLLM by more than the bug did. Comparing the
indices showed 0.794, and fixing the ReLU moved it to 0.975.
* llama: give the qwen4exp indexer cache the attention cache's slots
The indexer cache found its own slots, independently of the attention
cache. Both are the same size and see the same ubatches, so in a
straight-through prefill they agree, which is why every fixture and every
single-shot parity run passed. They drift once the context is being
rewritten between turns, and then the QSA top-k indices, which are applied
against the attention mask, point at the wrong cells.
The seven-turn chat test caught it on the third turn: llama-server aborted
on the assertion that the two caches report the same n_kv.
The cache is a side buffer addressed by the attention cache's cells, so it
now takes that cache's slot layout instead of computing one. Applying that
layout also marks its cells identically, so the two agree cell for cell by
construction rather than by coincidence, and the assertion can no longer
fire.
Inert where the caches already agreed: test-llama-archs green at 126 archs
and 0.00e+00, and the 4096-token tiny fixture is unchanged at max logit
delta 0.0.
* tests: record what the qwen4exp arch-test skip actually observes
The old note guessed that the hyper-connection keys never reach the file.
They do: dumping the gguf_context handed to llama_model_init_from_user
shows both among its 67 KVs, and the loader still reports one missing.
* tests: cover qwen4exp in test-llama-archs
The arch was skipped with a note guessing that the hyper-connection keys
never reached the synthesised file. They did. The suite builds a model, then
saves and reloads it, and llama_model_saver did not re-emit those keys, so
the failure was in the roundtrip leg rather than the first load. Three gaps,
all in shared code and all additive:
- add_kv_from_model wrote no hyper-connection, compress-ratio or PLE keys.
The PLE group only means anything whole, so it is written or omitted
together; the rest follow the file's existing style of writing every key
unconditionally, since an architecture that does not read one is
unaffected by a zero.
- the saver had no uint64 path at all, which the PLE hash constants need.
- add_tensors_from_model enumerates model-level tensors by hand and was
missing per_layer_tok_embd and the three final-mixer tensors.
Two smaller fixes on the qwen4exp side, both found by running the test:
- build_qsa_top_k divided by the compression ratio before asserting it was
non-zero, so a file without the key crashed instead of reporting.
- a layer with no compression ratio now falls back to dense attention,
which is what the model computes below the budget anyway. The test then
has to write a ratio to reach QSA at all, and an indexer key length no
narrower than n_rot, since the indexer ropes with the main attention's
rotary width.
Full suite: 126 archs, qwen4exp at 0.00e+00 with roundtrip OK. The tiny
fixture is unchanged, max logit delta 0.0 against the pre-QSA dense run.
* convert: stream the qwen4exp PLE table instead of concatenating it
The n-gram table arrives as 128 shards that were held in a dict and then
torch.cat-ed, so the peak was the shards plus the concatenation: around
300 GB of RSS on the real checkpoint, which rules out machines that could
otherwise convert this model.
Each shard is now written straight into a memory-mapped file at its final
row offset and dropped, so the resident set is one shard and the rest is
the page cache's problem. The temporary file sits beside the output and is
removed once the write finishes, including on failure.
Shards other than the last must be uniform for direct placement, which is
asserted rather than assumed, and a shard arriving before the stride is
known is held instead of misplaced.
Verified on the tiny fixture: the resulting GGUF is byte-identical to the
one the concatenating path produced (md5 2d274efac91ad1e9a6007efb0687e597).
* quantize: fall back to F16 for 32-block types with an odd ncols
tensor_type_fallback demotes a tensor whose ncols is not a multiple of the
target's block size, but its switch only enumerates the 256-block types. A
target that is already a 32-block type (iq4_nl, q4_0, q5_0, q8_0, ...) falls
into default: and throws, even though the function already knows how to answer
that case: the ncols check right below the switch resolves an unrepresentable
shape to F16.
Route those types into that check instead of throwing. Only paths that abort
today change, so no quantization that currently succeeds is affected.
Found on a 4-wide depthwise conv kernel. llama-quantize reported nothing but
"failed to quantize model from ...", with no tensor name and no exception text,
which made a quant recipe that had simply not pinned the tensor look like a
corrupt model. It now names the tensor and continues.
* quantize: let --tensor-type name per_layer_token_embd
per_layer_token_embd shares the TOKEN_EMBD category with token_embd.weight, so
--token-embedding-type is returned for it before any --tensor-type pattern is
consulted, and there is no way to give it a tier of its own.
That grouping is fine as a default and stays the default. It is a poor fit for
the size, though: on qwen4exp the table is 97.7 GiB of a 337.6 GiB BF16 file and
about 46% of a 4-bit one, roughly eighty times token_embd.weight, and it is
read by ggml_get_rows rather than a matmul so no imatrix ever covers it.
Allow an explicit --tensor-type pattern to name it, and only it. Nothing
changes unless such a pattern is passed, and token_embd.weight keeps the old
precedence in either case.
Measured on Qwen3.8-Flash-Next, Q4_K_M with an imatrix: the table lands at q8_0
(51.9 GiB, 113.5 GiB total) by following --token-embedding-type, and pinning it
q4_1 gives 30.5 GiB for 92.1 GiB total, 19% off the file.
* quantize: size the output buffer exactly instead of nelements * 4
The per-tensor output buffer was sized `nelements * 4`, described as an upper
bound. It is a very loose one: the output is at most 2 bytes per element
(f16/bf16) and usually well under 1.1 (q8_0 and below), so between 2x and 4x of
it is never touched. The exact size is already known here, since it is what the
quantization loop writes, what new_size sums to, and what the GGUF metadata is
asserted against a few lines later.
On a model whose largest tensor is a few GB none of this matters. On
Qwen3.8-Flash-Next it does: per_layer_token_embd is 51.2 G elements, so the
buffer was 205 GB where 54 GB is needed at q8_0 and 32 GB at q4_1.
Measured on that model, VmHWM of a live llama-quantize was 485 GB per process.
Three of them fit in 2 TB and five did not, which is what an OOM-killed quant
ladder looks like. This removes about 150 GB of that.
Byte-identical output, verified against the same binary built at the parent
commit: q4_K, q8_0, q5_K, q6_K and IQ4_XS, over BF16 and F32 sources, with and
without a PLE table present. Six cases, six matching md5s.
* qwen4exp: hash the image placeholder for multimodal batches
The PLE row indices are computed host-side from ubatch->token, and set_input
returned early when that was null. A multimodal ubatch is exactly that case:
the mtmd layer consumes the image placeholder ids and hands llama_decode
embeddings instead. The early return left the I32 index tensor uninitialised,
so ggml_get_rows indexed a 320 M row table with whatever the buffer happened to
contain, and aborted:
GGML_ASSERT(i01 >= 0 && i01 < ne01) failed
ggml_compute_forward_get_rows
mtmd_helper_decode_image_chunk -> llama_decode
Every image request crashed. Nothing caught it because the vision work had only
ever been verified by converting an mmproj, never by running one.
The reference computes the hash over input_ids, where those positions still
hold the image placeholder, so carry that id through as qwen4exp.ple.image_token_id
and hash it. The key is optional: a file converted before it existed falls back
to the PLE EOS token, which is defined and treats the image as a segment
boundary rather than crashing.
Verified end to end with llama-mtmd-cli, a Q4_K_M base and the F16 mmproj, on a
generated image with known content. The model names the red circle, the blue
square, the inverted green triangle and reads "UNSLOTH 42", each with the right
position.
* qwen4exp: support a non-unified KV cache in QSA
set_input_qsa asserted n_stream == 1, so llama-server could not serve this
model with more than one slot unless -kvu was passed. With a non-unified
cache each sequence owns its own cells, and a cell index means a different
token in each stream, so a single shared mapping is wrong.
- cell_blk, blk_cells and bias gain a stream dimension. At n_stream == 1
these collapse to the shapes they had, so the unified path is unchanged.
- Scoring is now batched over streams. ggml_mul_mat matches ne[2] on both
operands, so stream s's queries only ever meet stream s's blocks; without
this sequences would score against each other's context.
- set_input_qsa loops per stream and resolves cells through
v_cells[seq_to_stream[seq_id]], following set_input_kq_mask_impl, instead
of hardcoding v_cells[0].
- llama_kv_cache_context::get_n_stream() is added, mirroring the ns that
get_k and get_v already derive from the slot info.
build_attn_mask_top_k needed no change: it already expects
[n_top_k, n_batch, 1, n_stream], so the top-k result is reshaped to meet it.
set_input_qsa has exactly one caller, so the blast radius is qwen4exp only.
Validation, UD-Q4_K_XL on one B200:
- unified cache unchanged within noise: 1802.9/68.85 -> 1807.2/69.11 t/s at
batch 1, 2262.5/192.43 -> 2270.1/193.75 at batch 4.
- non-unified now runs at npl 1, 4, 16 where it previously aborted, and is
22% faster than the -kvu workaround at batch 16 (1205 vs 984 t/s total),
since per-stream cells avoid the cross-sequence masking a unified cache
pays for.
- no cross-stream contamination: four concurrent sequences each carrying a
distinct secret all recall their own and no other, on both cache modes.
- test-llama-archs green on qwen4exp, deepseek2, gemma3n, qwen3next, llama.
Note on testing: comparing concurrent output against solo output exactly is
not a valid check. It failed 0/4 with no bug present, and the unified-cache
control failed the same way, because batch composition changes the
floating-point reduction order and near-tied tokens flip. The contamination
test above is what the exit code gates on.
* llama: keep the qwen4exp top-k attention mask arch-local
The QSA graph needed a build_attn that attends only to the cells named by a
top_k tensor, and the first version got it by adding a llm_graph_input_attn_kv
overload to llm_graph_context and factoring the mask construction out of the
existing MLA sparse path into a shared build_attn_mask_top_k.
That put a new arch on the shared attention path and made the deepseek32 and
glm-dsa attention build depend on a helper introduced for qwen4exp. Build the
mask in src/models/qwen4exp.cpp instead and leave llama-graph.{h,cpp} exactly as
they were: the MLA path keeps its own copy of the same node sequence.
The nodes emitted are unchanged, so this is bit-identical.
* llama: hold the qwen4exp indexer cache in a new llama_memory_hybrid_idx
The indexer key cache was added by extending llama_memory_hybrid with an
optional third cache, and the host-side cell/block mapping that drives QSA was
added as set_input_qsa on llama_kv_cache. Both are shared classes that every
hybrid and every attention model goes through.
Move both into a new memory type, llama_memory_hybrid_idx, following
llama_kv_cache_msa: the indexer cache and the pos<->cell translation live with
the sparse-attention memory rather than in the classes that serve every other
architecture. llama-kv-cache.{h,cpp} and llama-memory-hybrid.{h,cpp} are
restored to their unmodified state.
init_batch is repeated from llama_memory_hybrid because the indexer cache has to
be handed the attention cache's slot infos, and those are not reachable through
the context the base returns. Allocating them separately lets the two caches
drift, which is what pointed QSA's top-k at the wrong cells before.
The context derives from llama_memory_hybrid_context so build_inp_mem_hybrid
keeps working unchanged, and get_n_stream is computed from the slot infos
exactly as llama_kv_cache_context did.
Behaviour is unchanged: logits over an 8192-token sequence are bit-identical to
the previous implementation, sparse and dense alike.
* llama: save and restore the qwen4exp indexer KV cache
llama_memory_hybrid_idx forwarded clear, seq_rm, seq_cp, seq_keep, seq_add and
seq_div to the indexer cache but not state_write / state_read, so a saved
session dropped the indexer keys and a restored one selected QSA top-k against
an empty cache. The effect is invisible until the context passes
indexer_top_k + compress_ratio - 1 cells, because QSA is exactly dense below
that and the indexer contents cannot change the result.
The indexer section is written last rather than next to the attention cache it
mirrors. As a suffix, a reader that does not expect it stops early and the
trailing bytes are caught by the size check in state_load_file; placed between
the attention and recurrent sections it would instead be parsed as recurrent
state, which can succeed and restore silent garbage. It follows the same
LLAMA_STATE_SEQ_FLAGS_PARTIAL_ONLY gate as the attention cache, since a partial
checkpoint deliberately skips the token-level attention caches.
The indexer restores its own cells instead of taking the attention cache's
restored slots. The two caches share size, padding and every sequence
operation, and init_batch hands the indexer the attention cache's slot infos,
so both state_read_meta calls run find_slot over identical occupancy and land
on identical cells.
The overrides live on llama_memory_hybrid_idx, the only memory type that owns
an indexer cache, so llama_memory_hybrid and every architecture that uses it
write and read exactly the bytes they did before.
The session and sequence state versions are bumped because the qwen4exp state
layout changed. The session path already rejects a short read via its size
check, but llama_state_seq_load_file accepts one silently, so only the version
check stops a pre-fix blob from being half-restored by a fixed build.
(cherry picked from commit 2721542354f8e158c3217625f4e2e7b83e51e3fe)
* llama: make the qwen4exp PLE n-gram history per context and serialise it
The PLE hash of a token mixes in the ple_ngram_size - 1 tokens before it, which
a decode ubatch does not carry, so they were remembered in a map on
llama_model_qwen4exp. That is the wrong owner twice over.
A llama_model is shared by every context that loads it, and the map was keyed
only by llama_seq_id, so two contexts running the same sequence id - two server
instances on one model, or a draft/target pair - overwrote each other's window.
The next_pos guard turned that into EOS padding instead of a crash, so it
degraded quality silently.
The map was also in no state blob: grep found ple_hist in neither
llama-kv-cache.cpp nor llama-memory-*.cpp nor llama-context.cpp. A restored
context therefore failed the next_pos check on its first ubatch and hashed the
first tokens after the restore against EOS padding. This is why a session blob
round-tripped byte for byte while the restored context computed different
logits: the state was never in the bytes.
It moves to llama_memory_hybrid_idx, which is per context, is the memory type
qwen4exp always builds, and already does the per-sequence bookkeeping this
needs. Every sequence operation now carries the window with it:
seq_rm a rewind (p1 < 0) truncates the window to the surviving prefix and
moves next_pos to p0, so a rollback keeps exact context; a hole
punched in the middle leaves the window non-contiguous, so it is
dropped
seq_cp the destination inherits the source's window, truncated to the
copied position range - a copied sequence continues with the same
n-grams the source would have used
seq_keep every other sequence's window is dropped, like its cells
seq_add a shift that moves the whole window keeps it and moves next_pos with
it, which is the context-shift case; one that cuts through it drops
it
seq_div positions stop being consecutive, so an overlapping window is
dropped
clear everything is dropped
Dropping means next_pos = -1, which set_input turns into full EOS padding: the
same thing a fresh sequence gets, and the same thing this code did before it
followed the sequence operations at all, so no case is worse than before.
The state payload is a self-delimiting list, u32 count then per entry
{ i32 seq_id, i32 next_pos, u32 n_toks, i32 toks[n_toks] }, so a whole-context
save and a single-sequence save share one format and a single-sequence restore
can retarget the window at its destination seq_id. It is written after the
indexer section, last, for the same reason that one is: as a pure suffix an
older reader stops early instead of parsing these bytes as something else.
Unlike the indexer section it is not under LLAMA_STATE_SEQ_FLAGS_PARTIAL_ONLY.
The window is recurrent state - it is the input the PLE convolution's own
recurrent state is derived from - and the recurrent cache beside it is written
for partial checkpoints too. Gating it would leave the server's speculative
decoding checkpoints restoring the conv state without the window that produced
it.
No further version bump: LLAMA_SESSION_VERSION 10 and LLAMA_STATE_SEQ_VERSION 3
were introduced for the indexer section in the same unreleased series, and both
changes are qwen4exp-only additions to the same blob layout.
Also fixes the padding of a short window. set_input pads a window shorter than
ngram_size - 1 up to that length, but prev() indexes the snapshot with the most
recent token last, and resize() pads at the back, so the filler EOS landed where
the immediately preceding token belongs. It now pads at the front. A window is
short at a sequence start after a one-token prefill, and after a seq_rm rewind,
which the new bookkeeping makes common.
Every architecture other than qwen4exp builds llama_memory_hybrid rather than
llama_memory_hybrid_idx, has no PLE table and never asks for a history, so
nothing about its graph, its sequence operations or its state bytes changes.
(cherry picked from commit de170364c052c68fcf63285cc0028095edb9f23c)
* qwen4exp: tidy comments and simplify image token read
Rewrite the comments this series adds to the AGENTS.md rules: one or two lines,
no prose hard-wrapped mid-sentence, no narrative or history, and no comment that
only restates the code. Net 146 fewer comment lines, no code change.
Correct the PLE image comment: mtmd does not consume the placeholder ids. An
image is decoded as an embeddings-only batch, so ubatch->token is null and the
per-position ids never exist here. gemma3n and gemma4 hit the same case and
stand in row 0 of per_layer_token_embd; qwen4exp stands in the configured image
token id instead.
Read image_token_id straight from self.hparams in the converter. base.py merges
text_config into the root of hparams, and the key sits at the root of
config.json, so the config.json re-read was redundant.
(cherry picked from commit 205840c12169057da3e8d2f65ec4ceec3e18b980)
* qwen4exp: support a quantized KV cache in the QSA attention path
(cherry picked from commit 4c30574f81dc1115d08078c47b6cf8c789c0a842)
* llama: give qwen4exp a large-graph node budget
(cherry picked from commit 37c8c194e6a30e4c46ac29bee3fb264f091596ef)
* qwen4exp: drop an unused variable that breaks -Werror builds
(cherry picked from commit 528d032b51fa3cf935ed3ef6e0fb1c7401df53b5)
* quantize: dequantize and quantize large tensors in row bands
f32_conv_buf held the whole dequantized tensor, which is 204.8 GB for
per_layer_token_embd alone and dies with std::bad_alloc long before the
work buffer is reached. Dequantize and quantize in bands of whole rows
instead, capping the f32 staging at 1 GiB per band.
Rows are independent and the imatrix is indexed by column, so band
boundaries cannot change any output byte. Bands nest inside the existing
per-expert loop so each expert slice keeps its own imatrix, and a band is
kept to at least one quantization chunk per worker thread so the existing
multithreading still has work. F32 sources still stage nothing and are
banded by pointer arithmetic into the tensor.
llama_tensor_dequantize_impl now takes a first element offset; the single
caller is updated.
(cherry picked from commit 658c22549613555dbce57a772be4de8509eba3ee)
* llama: segment the qwen4exp fused QKV for tensor split
qwen4exp was missing from the gated delta net branch of get_split_segments,
so its attn_qkv.weight, shaped {n_embd, 2*key_dim + value_dim}, fell through
to the generic fused QKV rule and tripped
GGML_ASSERT(tensor->ne[axis] == n_embd + 2*n_embd_gqa) while loading with
--split-mode tensor. --split-mode layer was unaffected.
qwen4exp broadcasts K to the V heads by tiling, k_conv is grown with a plain
ggml_repeat_4d over the head axis so that v head j pairs with k head
j % n_k_heads. That is the Qwen 3.5 pattern, not the repeat interleave that
Qwen 3 Next builds explicitly, so qwen4exp takes the else branch and its V is
segmented on the scale of K.
Reported by benklop.
(cherry picked from commit 353d753f595dc81634ae6130188b31f06018f5ae)
* llama: fix the qwen4exp PLE history seq_rm(-1) iterator invalidation and the fatal-warning build
ple_hist_rm recursed over ple_hist with a range-based for and the recursive call
erases the entry it is iterating when the whole sequence is removed (p0 <= 0,
p1 < 0), so the loop then increments an invalidated iterator. It is unreachable
today only because llama_memory_recurrent::seq_rm rejects seq_id < 0 before
llama_memory_hybrid_idx::seq_rm reaches the history, which is a guard in another
class. Advance past the entry before recursing.
Two smaller things in the same area:
- the n_toks sanity bound in ple_hist_state_read was the literal 64, which is
the value of LLAMA_MAX_PLE_HEADS, not of the quantity being checked. The
window is at most ple_ngram_size - 1 tokens, so the bound is
LLAMA_MAX_PLE_NGRAM - 1, eight times tighter.
- build_conv_state_at left mem_size unused, so -DLLAMA_FATAL_WARNINGS=ON does
not compile. Predates this series; drop the line.
(cherry picked from commit 6eba44a89d5f328eb4859b844e1d28fb564cbe3e)
* qwen4exp: include llama-impl.h explicitly for llama_mul_mat_hadamard
(cherry picked from commit b634fd4d250d181ef82bf78bd00c1ae3b96a7af6)
* convert: fix the qwen4exp lint and type-check failures
flake8 flagged an unused MmprojModel import, and ty flagged seven errors in
the PLE streaming path: eos_token_id can be absent, and _ple_map, _ple_path,
_ple_row_dim and _ple_rows_per_shard are all Optional at the declaration but
were dereferenced without narrowing.
The map is opened and the stride fixed before the first shard is written, and
_finish_ple_table only runs once every shard has landed, so the invariants
hold. Assert them so the checker can see it. A missing eos_token_id now raises
with the reason instead of a TypeError from int(None).
* llama: give the qwen4exp indexer cache its own tensor names
The indexer KV cache and the attention KV cache both named their tensors
cache_k_l%d, so the Meta backend matched the indexer cache against the
attention split pattern and aborted in handle_set_rows. Tag the names
instead, and mirror the indexer cache: it has one key head and its
projections are mirrored.
(cherry picked from commit a1cdc8181134659766763a17762545a1f0e5db7b)
* qwen4exp: double the Q split granularity for tensor parallelism
qwen4exp fuses the attention gate into attn_q.weight the same way qwen3next
and qwen 3.5 do, so a device boundary must fall on a whole q+gate pair or the
Q heads stop lining up with the K/V heads and attn_output rows.
(cherry picked from commit 6c9a592f0a425a459ab6efae3b897cf68460e244)
* qwen4exp: keep the indexer cache in step across server slots
The QSA indexer keeps a side cache addressed by the cells of the attention
cache, so cell j has to hold the same token in both: the top-k indices it
produces are applied to the attention KQ mask. init_batch already hands the
indexer the attention cache's slot layout rather than letting it look for its
own, but the restore path did not. state_read called llama_kv_cache::state_read
on the two caches in turn and each ran its own find_slot over its own occupancy.
That agrees only for as long as nothing has already pushed the two caches apart,
which is the property a restore is supposed to re-establish rather than one it
can lean on.
The failure path was the worse half, and it is reachable from the public API
with nothing more than a short buffer. Truncating a good blob at 35 offsets and
feeding it to llama_state_seq_set_data left the two caches disagreeing at 5 of
them, and every one of 23 truncations of a whole-context blob did. Four of those
five land inside the attention section, so the attention cache drops the
sequence and the indexer keeps it; only the cut that lands in the indexer
section gives the opposite direction. llama_kv_cache::state_read cleans up its
own cache and rethrows, so whichever way it falls, nothing is left to bring the
two back together. The server papers over this by clearing the slot when a
prompt cache load fails; a caller of llama_state_seq_set_data that does not is
left with an indexer addressing cells that no longer mean what it thinks.
llama_kv_cache::state_read_sinfo reports the cells a restore landed in, or takes
a copy of them, and state_read_meta uses a supplied layout in place of find_slot
once it has checked that those cells are free here too. The indexer now adopts
the attention cache's restored layout by construction instead of reproducing it
by coincidence, and a layout that does not fit fails the read rather than being
applied over cells that already drifted. The hybrid restore is wrapped so that
any failure drops the sequence, or for a whole-context restore the context, from
all three caches at once, which is a state they do agree on.
* kv-cache: clear the cache once when restoring a whole context
state_read walks the streams of the cache in turn, and for a whole-context restore
each stream went through state_read_meta, which starts by calling clear(). clear()
resets every stream at once, so each stream after the first threw away the streams
already restored, and the K/V buffers with them. A non-unified cache holds one
stream per sequence, so a context saved with N sequences in it came back with only
the sequence in the last stream that carried any cells - the highest sequence id.
A unified cache has one stream and never showed it.
The cache is now emptied once, before the loop, which is what a whole-context
restore means. A blob whose streams are all empty now empties the cache as well,
where before it left the old contents in place.
* kv-cache: check the mirrored slot layout on a whole-context restore too
state_read_meta only looked at the layout it was given on the single-sequence path.
A whole-context restore lays the cells out from 0 in both caches, so they agree as
long as they restore the same number of cells, but nothing checked that they did: an
indexer section belonging to some other context was read over cells the attention
cache had filled from a different one, which is the state the indexer must never be
left in.
* qwen4exp: give the PLE conv history its own mirrored recurrent row
n_embd_r() reserved n_conv + ple_conv_state() so that one cache_r_l row could
carry both the delta-net conv state and the PLE dilated conv history, but the
QWEN4EXP arm of get_split_segments only described n_conv. Under -sm tensor the
segment sum came up short by ple_conv_state() and llama_memory_recurrent
construction aborted in ggml_backend_meta_alloc_ctx_tensors_from_buft.
Widening the segment list is not the fix. The Meta backend propagates a view's
split descriptor from its parent unchanged, so a view of one sub-range of a
split axis is sized as the whole row on every device; declaring the PLE tail as
a second segment merely moves the abort to "shape mismatch for VIEW" at graph
allocation. The two histories also want opposite policies: the delta-net state
is split by head to match wqkv and ssm_conv1d, while per_layer_tok_embd,
ple_conv1d and ple_norm_conv are all mirrored, so every device computes the
whole dilated conv and needs the whole history. One tensor cannot be both, and
the split state has no per-segment mirroring.
Move the PLE history into its own cache_ple_r_l%d row, mark it MIRRORED, and
return n_embd_r() to n_conv. The row is allocated only on layers where is_ple
holds, so mirroring one 92160-element row per device replaces a 92160-element
tail on all 36 recurrent rows: the recurrent R footprint drops rather than
grows. build_conv_state_at now takes its width from the tensor it was handed
and keys its gather on that tensor, which also drops a cont of a strided view.
* no more ple_hist (use master version)
* llama: give the qwen4exp full memory context its indexer cache
graph_reserve() walks a full memory context, and qwen4exp builds its
sparse attention only when the context exposes an indexer cache. the
full-context constructor left ctx_idx null, so the reserved worst case
was the dense fallback: a smaller graph than the one decode executes.
ggml-alloc then had to grow the compute buffer on the first decode,
past the size reported at load.
with -np 4 -c 32768 -fa on -ctk q8_0 -ctv q8_0 on an IQ1_S qwen4exp,
the reserved CUDA0 buffer was 217.00 MiB against 275.71 MiB actually
used, and CUDA_Host 42.31 MiB against 191.14 MiB. reserving the sparse
graph makes both match exactly, in unified and non-unified cache mode.
Co-authored-by: Pascal <admin@serveurperso.com>
Assisted-by: Claude
* qwen4exp: shrink the PLE hparams storage
llama_hparams is held by value inside llm_graph_params and every llm_graph_input_*,
and llm_graph_params is a stack local in graph_reserve and process_ubatch, so its
width is paid on every worker thread stack.
is_ple_impl spent 2048 bytes carrying 512 bits. It is the one per-layer flag that is
not moved through the loader's uint32 array templates, so a bitset costs nothing in
call sites and also removes the uninitialized read that non-qwen4exp archs had, since
nothing filled the array for them.
The PLE head offsets and vocab sizes are token-space indices; the gather that consumes
them already truncates to int32, so 64-bit storage was never reachable. The gguf arrays
stay uint64 for file compatibility and are narrowed on load.
sizeof(llama_hparams) 34440 -> 31944, sizeof(llm_graph_params) 34872 -> 32376.
* llama: opt-in random-access mmap advice for host-resident gather tables
qwen4exp keeps per_layer_token_embd on the host: 26.8 GiB at IQ4_NL, read
by ggml_get_rows as 16 gathers of ~90-170 bytes per token, spread across
16 head regions ~20M rows apart. Measured over 4.75M gathers, no two
consecutive gathers land on the same 4 KiB page, so the readahead the
loader asks for buys nothing here and the whole table ends up cached to
serve about 4% of itself.
llama_mmap applies POSIX_FADV_SEQUENTIAL, MAP_POPULATE and a whole-file
POSIX_MADV_WILLNEED unconditionally. Those are right for streaming the
file once into buffers and wrong for whatever stays mapped afterwards.
Under LLAMA_MMAP_RANDOM the eager pull-in is skipped and the mapping is
advised random once every tensor has been read, so the load itself keeps
its sequential readahead. That alone drops the table to 4.4% resident but
serializes one NVMe latency per gather.
The second half is what pays for it: the PLE input already computes every
row index for the ubatch before the graph runs, so the pages those rows
fall on are handed to the kernel in one batch and the reads overlap.
POSIX_MADV_WILLNEED on POSIX, PrefetchVirtualMemory on Windows, which
takes the discontiguous ranges in a single call.
Off by default and off for every other model: the batched prefetch keys
off "this mapping was advised random", which nothing sets unless the user
opts in.
-c 512 --chunks 60, cold, IQ1_S, mean of 3:
default 35.3 s 26.82 GiB resident (100%)
advice only 104.5 s 1.19 GiB resident (4.4%)
advice + prefetch 34.2 s 1.19 GiB resident (4.4%)
PPL 4.2346 +/- 0.07862 in all three. IQ1_S KLD is unchanged in every
field, including Mean KLD 0.396070 +/- 0.001931 and Same top p 77.325%.
* llama: narrow the random-access mmap advice to the gather table
The advice was applied per mapping: every mapping the model kept got
POSIX_MADV_RANDOM plus a whole-file POSIX_FADV_RANDOM, and the eager
pull-in was skipped for every file. On qwen4exp that also hit
token_embd.weight, which sits 0.33 GiB past the PLE table in the same
shard and is read densely, not by sparse gathers. Measured over
-c 512 --chunks 60 on IQ1_S it fell to 8.45% resident, against 100% with
the feature off.
A model now nominates its gather tables (qwen4exp: per_layer_tok_embd)
and only those byte ranges are advised. The range is rounded out to
whole pages, which on this model takes in 832 bytes before and 192
after. token_embd goes back to 86.55% resident and the PLE table still
drops to 4.44%; smaps shows one VM_RAND_READ VMA of exactly the table
instead of one over all 27.16 GiB that stays mapped.
posix_fadvise is dropped from the narrowed path. POSIX_FADV_RANDOM
ignores its offset and length and marks the whole open file, and the
FMODE_RANDOM it sets is only read by page_cache_sync_ra() on the read()
path, which a fault on a MADV_RANDOM vma never reaches. POSIX_FADV_
DONTNEED does take a range, so the drop mode keeps it.
The eager pull-in is now skipped only for the files holding a nominated
table, and re-issued as WILLNEED over the rest of such a file, so other
shards load exactly as before.
prefetch_rows() keys off the tensor being nominated rather than off a
mapping-level flag, so the batched readahead lands only where the advice
did.
-c 512 --chunks 60, cold, IQ1_S, mean of 3, total wall:
default 32.50 s
whole mapping 30.05 s
narrowed 30.35 s
PPL 4.2061 in all three. IQ1_S KLD is bit-identical with the feature on
and off, including Mean KLD 0.396070 +/- 0.001931 and Same top p
77.325%. tg128 73.65 +/- 0.33 narrowed against 73.49 +/- 0.34 whole.
Assisted-by: Claude
* llama: fold the random-access prefetch into its own feature flag
LLAMA_MMAP_RANDOM_PREFETCH existed to measure the two halves of the feature
apart, and the measurement is done: on a cold cache over the same wikitext
run, MADV_RANDOM without the batched readahead takes 94.4 s against 36.7 s
for an untouched mapping, while the pair together take 34.1 s. Suppressing
the kernel's readahead only pays if we replace it, so the split let a user
select a 2.6x regression through a documented switch.
Keep the accessor, since the call site reads better than a mode comparison,
but derive it from the mode alone.
* FACP (Fewer Acronym Classes Please)
* qwen4exp: bias the QSA selection per block, not per cell
The QSA bias is a graph input, so it is pinned on the host and uploaded every
decode, and at -c 32768 -np 4 its twelve copies were 768 of the 815 MiB of
reserved host compute buffer.
Only one half of it needs a cell: whether the cell sits in the always-visible
tail, and whether its block was pooled. Both are properties of the block. The
other half - empty, other sequence, or in the future - is the plain visible/not
test the attention mask already carries over the same cells, so add that mask
instead of repeating it. The bias then holds one value per block.
A block sits wholly inside or wholly outside the tail because the tail starts on
a block boundary, so one value per block is exact. Cells no block covers keep
their -inf from the mask.
The mask is F16 and the bias F32, and a mixed ggml_add reinterprets the F16
buffer as float rather than converting it, so the cast is required.
reserved host compute buffer at -c 32768 -np 4:
--kv-unified 814.86 -> 238.86 MiB, CUDA0 721.07 -> 421.07 MiB
--no-kv-unified 214.86 -> 70.86 MiB, CUDA0 317.07 -> 265.07 MiB
Selection is unchanged: over 8192 tokens, four times the budget, every QSA
layer returns identical top-k indices and the logprobs are bitwise equal.
Two things a reviewer should know. A cell whose position divides past the last
block is guarded by an assert rather than handled, because no run reached it.
And the mask's same-position M-RoPE rule cannot fire for text and was never
exercised for images, so the 2D case is unverified.
* clean up code comments
* clean up new comments
* revert LLAMA_MMAP_RANDOM
* nits
* replace some changes with #27795
* improve the m-rope image for get_prev_tokens
* LazyChunkedTensor
* fix lint
* add some validations
* reduce input nodes
* trim output tokens
* nits
* some more sanity checks
* fix llm_graph_input_ple reuse
* exclude from webgpu test
---------
Co-authored-by: danielhanchen <danielhanchen@users.noreply.github.com>
Co-authored-by: danielhanchen <unslothshared@gmail.com>
Co-authored-by: Xuan Son Nguyen <son@huggingface.co>
Co-authored-by: Pascal <admin@serveurperso.com>
Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@huggingface.co>
|
||
|
|
7221e24f57 |
model : GraniteSWAForCausalLM / GraniteMoeSWAForCausalLM (#25505)
* feat(convert): Add conversion for GraniteSWAForCausalLM Branch: GraniteSWAForCausalLM AI-usage: full (Bob, OpenCode + Qwen3.6-35b) Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * feat(llama): Add granite_swa support Branch: GraniteSWAForCausalLM AI-usage: full (Bob, OpenCode + Qwen3.6-35b) Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * feat(conversion): Add conversion infra for rope_pattern array NOTE: There is other work also targeting this, so this may be removed depending on merge order. Branch: GraniteSWAForCausalLM AI-usage: full (Bob) Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * fix(conversion): Fix SWA pattern logic and support for non-rope layers Branch: GraniteSWAForCausalLM AI-usage: full (Bob) Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * feat(conversion): Add support for GraniteMoeSWA Branch: GraniteSWAForCausalLM AI-usage: full (Bob) Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * feat: Add llama_hparams::has_rope and arch constants NOTE: This shadows the work done for Granite Speech https://github.com/ggml-org/llama.cpp/pull/25107 Branch: GraniteSWAForCausalLM AI-usage: full (Bob) Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * feat: Add support for per-layer rope determination Branch: GraniteSWAForCausalLM AI-usage: full (Bob) Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * style: Fix failing flake8 for extra newlines Branch: GraniteSWAForCausalLM AI-usage: none Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * test: Write out SLIDING_WINDOW_PATTERN in llama-model-saver Branch: GraniteSWAForCausalLM AI-usage: full (OpenCode + Qwen3.6-35b) Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * fix(convert): Fix missing registration for GraniteMoeSWAForCausalLM Branch: GraniteSWAForCausalLM AI-usage: none Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * fix: Load MoE params as optional Branch: GraniteSWAForCausalLM AI-usage: draft (OpenCode + Qwen3.6-35b) Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * feat: Handle MoE params in conversion branch: GraniteSWAForCausalLM AI-usage: full (OpenCode + Qwen3.6-35b) Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * style: Remove unnecessary newline AI-usage: none Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * fix: Remove unnecessary tensor additions to GRANITE architecture Branch: GraniteSWAForCausalLM AI-usage: none Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * fix: Correctly handle naming for ffn gate inp Branch: GraniteSWAForCausalLM AI-usage: none Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * fix: Always default hparams.rope_pattern to 1s This isn't strictly necessary, but it will allow other models to rely on hparams.has_rope(il) without needting to prepopulate. Branch: GraniteSWAForCausalLM AI-usage: none Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * feat: Move to has_rope for all granite model architectures Now that we have a proper hparam for this, it's better to use it and not require a hacky fallback in the hparam method itself. Branch: GraniteSWAForCausalLM AI-usage: none Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * feat: No hacky rope_finetuned fallback in has_rope Branch: GraniteSWAForCausalLM AI-usage: none Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * fix: Fully remove rope hparam filling in granitemoe There are no granitemoe models that use NoPE (it's not actually used in the layer building below), so this was just dead code. Branch: GraniteSWAForCausalLM AI-usage: none Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * fix: Save out rope_pattern in model-saver Branch: GraniteSWAForCausalLM AI-usage: none Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * fix: Set hparams.rope_finetuned for round trip Since the value is _read_ from rope_finetuned, we need to persist it when the model is saved with the saver. Branch: GraniteSWAForCausalLM AI-usage: none Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * fix: Code review cleanup Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@huggingface.co> Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@huggingface.co> * refactor: Keep gate/up fused for MoE path Branch: GraniteSWAForCausalLM AI-usage: full (Claude + Sonnet 5) Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * fix: Skip GRANITE_SWA in model saver https://github.com/ggml-org/llama.cpp/pull/25505#discussion_r3773175651 Keeping is_swa_impl in the saver can break other models. Branch: GraniteSWAForCausalLM AI-usage: none Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * add sliding window pattern for model in test * style: Fix indentation Branch: GraniteSWAForCausalLM AI-usage: none Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * fix: Fix \r\n Thanks Claude! Branch: GraniteSWAForCausalLM AI-usage: none Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * feat: Keep shared expert fused Branch: GraniteSWAForCausalLM AI-usage: full (Claude + Sonnet 5) Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * style: More indentation fixes Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@huggingface.co> --------- Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@huggingface.co> |
||
|
|
16d222fc5e |
model : add support for MiniMaxText01ForCausalLM and MiniMaxM1ForCausalLM (#27018)
* llama : support for MiniMax-Text-01 model * chore : renames to match the other MiniMax models * model : add logits mask as MiniMax-Text-01 embeddings tensor has zero-valued embeddings for tokens >= 200032 that produce zero logits disrupting the token sampling process * llama : replace hardcoded conditions with hparams.is_recr() * model : used build_rs() for recurrent state management * chore : code cleanup * model : optimized MiniMax-Text-01 by removing the state tranpose operations * chore : removed unnecessary ggml_cont() in MiniMax-Text-01 implementation * llama : add generic logits mask graph input * model : permuted diag_decay dimensions to avoid doing it inside MiniMax-Text-01 graph * chore : code cleanup * chore : code cleanup * model : use token positions when calculating MiniMax-Text-01 decay tensors * convert : add support for MiniMaxM1ForCausalLM as it seems to be the same as MiniMaxText01ForCausalLM * chat : add jinja template for MiniMax-M1 Co-authored-by: QscQ <qscqesze@gmail.com> * chore : code cleanup * tests : MINIMAX_01-related fixes * chore : silence Python lint errors * vocab : remove unnecessary vocab type * convert : update MiniMaxText01Model conversion to use yield when modifying tensors * convert : suppress tokens with zero-valued embeddings during MiniMax-Text-01 conversion * llama : removed logits mask - no longer necessary as token suppression is used instead * model : use common functions to make MiniMax-Text-01 implementation more concise Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@huggingface.co> * model : use common functions to make MiniMax-Text-01 implementation more concise Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@huggingface.co> * convert : override non-working built-in chat template during conversion * tests : skip arch MINIMAX_01 tests for WebGPU backend (it breaks again) --------- Co-authored-by: Stanisław Szymczyk <sszymczy@gmail.com> Co-authored-by: QscQ <qscqesze@gmail.com> Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@huggingface.co> |
||
|
Bar HaimandGitHub
|
157b81fe6d |
model : Granite-Switch Architecture (#25107)
* granite-switch: add llama.cpp backend (POC, CPU)
New "granite-switch" architecture: a dense, all-attention Granite-4.1
model with N embedded LoRA adapters selected per-token by control tokens.
- gguf-py schema (arch, KV keys, stacked LoRA tensor names) + writer helpers
- conversion/granite.py: GraniteSwitchModel converter (stacks N adapters +
zero base slot into per-projection A/B tensors; emits switch metadata)
- C++ arch registration (llama-arch.{h,cpp}, llama-model.{h,cpp})
- src/models/granite_switch.cpp: load + per-token switched-LoRA graph via
ggml_mul_mat_id over stacked tensors; sticky per-token index + control-token
substitution in llm_graph_input_switch::set_input
- llm_graph_input_switch in src/models/models.h
Runs end-to-end on CPU: convert 3b checkpoint (842 tensors, stacked dim 13)
and generate on both base and control-token paths. Sticky switch state is
single-sequence (POC); full multi-sequence machinery is a follow-up.
* granite-switch: add Mac (Metal) build + mid-sequence switch demo script
Self-contained script to build llama.cpp on Apple Silicon (Metal),
convert the composed 3b checkpoint, and run the crisp mid-sequence
adapter-switch demos verified on Vela:
- answerability: <|answerability|> mid-seq -> "unanswerable"
- query_rewrite: <|query_rewrite|> mid-seq -> {"rewritten_question": ...}
Each demo runs the same prompt twice, differing only by a control token
placed before the assistant turn, so the per-token switch is visible.
* granite-switch mac demo: add -no-cnv so each run is one-shot
The composed model ships a chat template, so llama-completion auto-enables
interactive conversation mode and halts at a `>` prompt after generating,
stalling the script. -no-cnv disables conversation mode: generate once from
the raw prompt and exit (also prints special tokens, making the switch visible).
* granite-switch: replace global sticky index with in-graph router attention
The POC computed the per-token adapter index on the CPU and carried it
across ubatches in ONE global `mutable int32_t poc_sticky_index`, reset
only when a ubatch contained sequence position 0. That global had two
problems:
1. Concurrency: with multiple sequences in a batch it was last-writer-
wins — one sequence's adapter leaked into the others.
2. Multi-turn: an interactive `ollama run` chat continues one KV cache,
so turn 2 never saw position 0 and the index never reset — the
adapter stayed stuck on across turns.
Port the vLLM/HF backend mechanism faithfully: a single-head causal
"router" attention recovers the adapter index in-graph. Per token, only
dim 0 carries signal — Q[0]=1, K[0]=+gain for a control token / -gain
otherwise, V[0]=adapter slot / 0 — and the causal softmax over the single
visible control token recovers that adapter's slot (readback =
clamp(round(V[0]), 0, n_adapters)). gain=15 matches config.py and is
F16-safe (no F32 cache).
The router's K/V live in the model KV cache at an extra layer
R == hparams.router_layer (== n_layer). We bump n_layer_all to n_real+1
so the cache allocator gives the router its own per-sequence slot, and
set n_layer_nextn=1 so n_layer() stays n_real — the decoder loop and
tensor loading are untouched and never reference layer R. The router K is
exempted from the k-shift RoPE loop (its dim-0 value is a literal
magnitude, not a rotation).
Because the selection now lives in the per-sequence KV cache, CONCURRENT
requests are isolated for free (problem 1 fixed; verified by
scratch/concurrent_switch_test.cpp). set_input becomes stateless pure
per-token maps; the global is gone.
Single-switch contract / known limitation, identical to vLLM & HF: the
gain is flat (no recency), so within one sequence there is no mechanism to
revert to base mid-sequence — once an adapter fires it stays on until that
sequence ends (problem 2 is therefore NOT fixed by a faithful copy; vLLM/HF
avoid it only because each served request is a fresh sequence). A client
continuing one KV cache across turns must start a fresh sequence per turn,
or opt into a recency-biased router (a deliberate divergence, not done
here). Documented in granite_switch.cpp and asserted by
scratch/multiturn_leak_test.cpp.
Verified (CPU): both demos unchanged (answerability -> "unanswerable",
query_rewrite -> rewritten query); concurrent two-sequence isolation
passes; multi-turn carry-over matches the vLLM/HF contract.
* granite-switch: drop scratch tests and mac demo for upstream PR
Remove the local-only development artifacts that should not ship in the
upstream PR:
- granite-switch-mac-demo.sh (local Metal build + demo driver)
- scratch/concurrent_switch_test.cpp
- scratch/multiturn_leak_test.cpp
Also drop the now-dangling reference to the scratch tests from the
granite_switch.cpp header comment. Leaves only the core architecture
support (conversion, gguf constants, llama-arch/model/kv-cache, and the
granite_switch graph).
* granite-switch: trim comments to match native llama.cpp style
* granite-switch: trim conversion comments to match native style
* granite-switch: drop unused adapter_ranks metadata
* granite-switch: rename arch to graniteswitch and drop obid alias
* granite-switch: fix non-ASCII comments and document router gain assumption
* granite-switch: drop section comments from constants.py to match native style
* granite-switch: add functional tensor block comments matching Granite4 Vision style
* granite-switch: clarify n_expert_used comment
State the actual constraint: mul_mat_id needs n_expert_used == 1, and
since the GGUF carries expert_count = 0 the generic loader's
n_expert == 0 => n_expert_used == 0 assertion has already passed by the
time load_arch_hparams runs, so it is forced to 1 here.
* granite-switch: note n_layer_nextn reuse has no MTP
The router carving reuses n_layer_nextn, normally the MTP/next-token
count. Clarify in the comment that it is borrowed here purely as the
trailing-layers lever and that there is no MTP head, to spare readers
the double-take.
* granite-switch: rename source file and apply review nits
* granite-switch: don't force LoRA tensors to F16, follow --outtype instead
* granite-switch: drop redundant _permute_qk wrapper, call LlamaModel.permute directly
* granite-switch: read router gain from GGUF (control_token_gain) instead of hardcoding 15.0
* granite-switch: derive n_slots()
* granite-switch: move llm_graph_input_switch into granite-switch.cpp
* granite-switch: cut AI-style narration comments
* granite-switch: collapse multi-line comments
* granite-switch: rename control_token_* maps to adapter_token_*
* granite-switch: cut noise comments
* granite-switch: rename embedded LoRA tensors to <base>.lora_a/lora_b
* granite-switch: GGML_ASSERT token input to avoid UB on embeddings
* granite-switch: TODO for raw embedding input support
* granite-switch: collapse LoRA tensor constants to .lora_a/.lora_b suffix
* granite-switch: drop n_expert_used hack, guard mul_mat_id buft probe
* granite-switch: stop forcing dense expert counts, read from config
* granite-switch: renamed control_token_gain metadata key to router_gain
* granite-switch: trim header comments to match native style
* granite-switch: collapse LoRA tensors to base name + suffix
* granite-switch: inline suffix checks in tensor op resolution
* granite-switch: drop switch-lora struct comment
* granite-switch: guard router layer index and inline n_slots
* granite-switch: group adapter metadata under {arch}.adapters.* namespace
* granite-switch: add hparams.has_rope(il) for KV-shift rope skipping
* granite-switch: skip arch in test-llama-archs (adapter fixture missing, TODO)
* granite-switch: Keys.Adapters namespace + simplify n_slots
* granite-switch: validate substitute token ids against n_vocab
* granite-switch: bound adapter count and lora rank from GGUF
* granite-switch: reject MTP context type when router_layer is set
* granite-switch: throw on bad adapter metadata instead of GGML_ASSERT
* granite-switch: use ASCII +/- in router K signal comment
* granite-switch: document n_layer_nextn repurpose and its leak points
* granite-switch: gate lora_a/lora_b op mapping on router_layer
* granite-switch: label all three preview model sizes
|
||
|
|
67d5978bb1 |
model: M3: Move MSA into a new memory implementation (#26338)
* Move MSA logic from llama-kv-cache into llama-kv-cache-msa * cont : minor * cont : ws fix --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> |
||
|
|
b1d4c65524 |
model: Add MiniMax-M3 (MSA: MiniMax Sparse Attention) support (#24908)
* Add preliminary MiniMax-M3 support Text-only port that re-uses existing components: MiniMax-M2 style GQA with per-head QK-norm and partial rotary, DeepSeek-V3 style leading-dense and routed/shared experts, and swigluoai activation. Sparse attention is not yet supported (dense fallback); vision tower and MTP heads are dropped. * MiniMax-M3 vision tower (mmproj + clip graph) * Delete m3_vision_ref.py * Update clip.cpp * MSA * Update constants.py * Update minimax.py * Cache creation. Working withotu flash attention * Added flash attention for sparse layers * Decomposed slow cpu OP into GPU + CPU ops. Massive speedup over long ctx * Rewrote indexer op to be cuda native. Modified flash attention to match per group block picking * Implement sparse attention calc out of stock ops. * Fix a cache allocation and cont issue * Fixed -fa auto crash, flagged debug spots * Delete vocab.json * Delete model.safetensors.index.json * Delete generation_config.json * Delete Minimax directory * Handled multi stream case to fall back on Dense Attention * Development scaffolding cleanup. No functional change to the decode or 4-way paths. Full debug harness remains at <8136a9c68ed7a5eb009aa67bba3fda8062f4648f> for reproducing the selection-parity validation. * Remove redundant comment from minimax-m3.cpp * Changed 3 Gelu Ops for vision into Gelu_erf ops * Assert that n_kv is multiple of 128 * Rename MSA index tensors to indexer convention Note: All GGUFs generated before this change will need to be regenerated. * Fix incorrect Assert * Review driven changes (#3) * Remove comment from conversion minimax.py Co-authored-by: Sigbjørn Skjæret <1629204+CISC@users.noreply.github.com> * Remove whitespaces from constants.py Co-authored-by: Sigbjørn Skjæret <1629204+CISC@users.noreply.github.com> * Tighten comment in minimax.py Co-authored-by: Sigbjørn Skjæret <1629204+CISC@users.noreply.github.com> * inherit MiniMax-M3 from MiniMax-M2 * drop dead text_config fallbacks * Add indexer writer methods * Reuse LLM_FFN_SWIGLU_OAI_MOE * Remove duplicate indexer setters, add only block_size/local_blocks, follow value naming convention * Fix conversion error /gguf_writer.py Co-authored-by: Sigbjørn Skjæret <1629204+CISC@users.noreply.github.com> * Update gguf-py/gguf/gguf_writer.py Co-authored-by: Sigbjørn Skjæret <1629204+CISC@users.noreply.github.com> * Update gguf-py/gguf/tensor_mapping.py Co-authored-by: Sigbjørn Skjæret <1629204+CISC@users.noreply.github.com> * Update conversion/minimax.py Co-authored-by: Sigbjørn Skjæret <1629204+CISC@users.noreply.github.com> * Update conversion/minimax.py Co-authored-by: Sigbjørn Skjæret <1629204+CISC@users.noreply.github.com> * Remove whitespace in src/llama-kv-cache.cpp Co-authored-by: Sigbjørn Skjæret <1629204+CISC@users.noreply.github.com> * Remove Whitespace in Update src/llama-model.h Co-authored-by: Sigbjørn Skjæret <1629204+CISC@users.noreply.github.com> * Remove whitespace in src/llama-hparams.h Co-authored-by: Sigbjørn Skjæret <1629204+CISC@users.noreply.github.com> * remove multimodal code upon maintainer request. Will be made as a separate PR * Whitespace clean in tensor_mapping.py * Log cache size on launch, block ctx shift, support prompt caching Log indexer cache size on launch Disallow ctx shift Support prompt caching * Update minimax-m3.cpp * Optimize implementation, add multi stream support. Fully rewrote minimax-m3.cpp for speed and buffer size gains: Unified the 4-way + decode, 1 FA call per layer instead of 4, with the groups mapped onto ne[3] Custom CPU op now emits block-level mask, expanded on GPU, which causes CPU to GPU transfer to shrinks at prefill Decode: ~25 nodes/layer vs ~50, no per-group concats/conts Unified selection semantics, so both regimes rank bs + local bias (position-anchored local force), which means prefill/decode can no longer disagree on selection can_reuse on the MSA bias input. Graph reuse at decode restored (was rebuilding the full graph every token) In-place mask adds, shrinking compute buffer ~6.8 to ~4.2 GiB at ub2048/62k Multi-stream: MSA now runs with -np N when kv_unified=false. Decode stays batched across streams (still 1 FA call), prefill loops per stream. dense fallback only for --kv-unified + multi-seq Measured effect on expert offload bound setup: decode 6.2(4WAY)–7.15(MSA_decode) -> 7.7~7.8 t/s, flat from 5k to 60k+. prefill around 10% faster. buffer about 20% smaller, multi-user support. * set default cache type to F32 * Fix potential DSA double indexer cache allocation bug, only allocate in-cache k_idx for archs that opt in * remove F16 downcasts in MSA attention, force F32 indexer score accum * Add Minimax eos to llama vocab * Guard edge case where idx cache can become stale after a tail trim * Update llama-kv-cache.h * Update llama-kv-cache.cpp * Update llama-kv-cache.cpp * Update llama-kv-cache.h * Update llama-kv-cache.cpp * Review driven changes * style fix * indexer hparams are required * fix tests * fix lint --------- Co-authored-by: Daniel Han <danielhanchen@gmail.com> Co-authored-by: Sigbjørn Skjæret <1629204+CISC@users.noreply.github.com> Co-authored-by: Xuan Son Nguyen <son@huggingface.co> |
||
|
|
88bfee1429 |
model: add GLM 5.2 Indexer support (#25407)
* Start building graph - reuse deepseek32 * Enable kv cache and rotation for glm_dsa architecture Just follow Deepseek 3.2 for now. * Reuse prev_top_k for "shared" indexer layers * GLM 5.2 uses LLAMA_ROPE_TYPE_NORM for the indexer. This is transformers' `apply_rotary_pos_emb_interleave` * Default indexer types to GLM pattern Previous converted GGUFs like https://huggingface.co/unsloth/GLM-5.2-GGUF write indexer weights to _all_ layers, even if they are only required for "full" types. This PR relies on a new key "%s.attention.indexer.types"; if absent, it will use the default GLM 5.2 schedule as defined in https://huggingface.co/zai-org/GLM-5.2/blob/main/config.json#L26. Note that conversion is not saving this key yet. * Save indexer types to gguf, restore on load * Use ggml_lightning_indexer when cparams.fused_lid Co-authored-by: fairydreaming <166155368+fairydreaming@users.noreply.github.com> * GLM 5 and 5.1 use full indexers Co-authored-by: fairydreaming <166155368+fairydreaming@users.noreply.github.com> * Fix indentation * Ensure array is zero-filled * Prefer explicit std::fill * Assert prev_top_k exists for shared indexer --------- Co-authored-by: fairydreaming <166155368+fairydreaming@users.noreply.github.com> |
||
|
Ruixiang WangandGitHub
|
1a2dea29b9 | spec: fix segfault error on long prompts for eagle3 (#24707) | ||
|
Aman GuptaandGitHub
|
04eb4c446d | llama : add Gemma4 MTP (#23398) | ||
|
|
7acb4e8cd2 |
hparams : refactor hparams.n_layer (#24060)
* hparams : refactor hparams.n_layer * cont : remove `n_layer_kv()`, use n_layer_all instead * cont : type consistency * pi : update SYSTEM.md * models : fix Step3.5 MTP * cont : remove duplicate switch cases * cont : explicitly set `false` to extra layers for `is_swa` and `is_recr` * cont : fix nextn layer count handling Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com> --------- Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com> |
||
|
Georgi GerganovandGitHub
|
06938ac129 |
tests : add support for qwen3 SSM archs (#24031)
* tests : add support for qwen3 SSM archs * arch : add LLM_KV_ATTENTION_RECURRENT_LAYERS * cont : naming + TODOs |
||
|
|
255582687b |
llama + spec: MTP Support (#22673)
* 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> |
||
|
Xuan-Son NguyenandGitHub
|
59db9a357d |
llama: dynamic head_dim and n_rot for SWA (#20301)
* llama: dynamic head_dim and n_rot for SWA * also add gguf_writer wrappers * fix build * build_rope_shift arg reorder |
||
|
|
3688c4f504 |
Kimi-Linear support (backend agnostic + MLA KV cache) (#18755)
* kimi linear model implementation * kimi linear convert_hf_to_gguf * kimi linear constants.py tensor_mapping.py * Kimi Linear ggml.h * kimi linear ggml-cpu * Kimi Linear ggml-cuda * Kimi Linear ggml.c * kimi linear src/llama * remove "const int64_t n_seq_tokens = q->ne[2];" to get rid of unused variable warning * remove type mismatch warning * read MoE params * removed some hard coded code * removed all hard code * use DeepseekV2 tokenizer * removed unnecessary internal methods called by the old set_vocab of KimiLinear * rewrite get_vocab for KimiLinear. Removed all kda_scan code * removed all traces of kda_scan * reduce OP count by 1 due to removal of kda_scan * Move KIMI_LINEAR to llm_arch_is_hybrid to enable KV cache * set n_embd_head_k/v to ensure kv cache works * don't quantize conv1d of Kimi Linear * Kimi Linear backend agnostic * removed LOG_INFO * naive chunking form implemented * fixed some comments * add Kimi-K2 specific tokens to be recognized as EOG * build_kda_autoregressive is implemented to replace build_kda_recurrent for faster inference. sync'd to b7682 * replaced Akk and Aqk with mul_mat and clamp * no clamp version * Moved Aqk computation out of the loop * fixed typo and split wkv_b into wk_b and wv_b * MLA KV cache support * fix trailing spaces * moved const llama_model & model; around to follow qwen3next format and see if it cna pass the -Wunused-private-field error * fix trailing whitespace * removed traling whitespaces in empty line + make sure indentation is multiple of 4 * try to make lint happy * remove blank lines to make lint happy * removed at least blank line containing white space * fixed flake8 complaints locally * return ggml_tensor * pair in kda_autoregressive and kda_chunking as in ngxson's Qwen3Next improvement * removed Kimi-Linear specific change that causes failure at server-windows * removed private: from kimi_linear to make build checks happy * removed unnecessary ggml_cont before ggml_reshape * created static function causal_conv1d to abtract similar code for q/k/v * merged dt_bias to SSM_DT. Do -exp(log_A) in convert_hf_to_gguf.py. * reverted to original * fixed find_hparam calls. Fixed e_score_correction_bias to use bias instead of weight. Removed all ssm_conv bias terms. * remove DT_B from constants.py. remove one comment line in llama-model.cpp * new class llm_graph_input_mem_hybrid_k to get around the new MLA change. switch the concat order of ggml_concat calls in kimi-linear.cpp to accommodate MLA changes. Removed support for exp_probs_b.weight * remove ssm_o_norm_b * remove ssm_o_norm_b * changed hparams.kda_head_dim to hparams.n_embd_head_kda. added TODO comment for class llama_graph_mem_hybrid_k * removed all ggml_cont b4 ggml_reshape_4d * Whitespace * replaced all hparams.get with find_hparams * added new names for n_experts, n_experts_used and score_func in TextModel and removed their code in KimiLinear in convert_hf_to_gguf.py. Removed unnecessary ggml_cont and GGML_ASSERT in kimi-linear.cpp * use is_mla to switch between different mem_hybrid types * fixed logical errors in convert_hf_to_gguf.py pointed out by CISC * removed if else for required parameters kv_lora_rank and qk_rope_head_dim * add back ggml_cont for Vcur * minor changes * removed extra line in llama-vocab.cpp. Added back the comment in llama-graph.cpp * f16 gguf cannot run without context length * made a mistake of adding back n_ctx parsing --------- Co-authored-by: Piotr Wilkin (ilintar) <piotr.wilkin@syndatis.com> |
||
|
Georgi GerganovandGitHub
|
d9c6ce46f7 |
kv-cache : support V-less cache (#19067)
* kv-cache : support V-less cache * cuda : better check for V_is_K_view * cuda : improve V_is_K_view check * graph : add comments * hparams : refactor |
||
|
Georgi GerganovandGitHub
|
2fbde785bc |
kv-cache : optimize KQ mask construction (#18842)
* kv-cache : optimize KQ mask construction * cont : add explanation + improve * cont : fix |
||
|
Tarek DakhranandGitHub
|
73d284a250 |
model : add LFM2-ColBert-350M (#18607)
* model : add LFM2-ColBert-350M * llama_model_n_embd_out() - returns `hparams.n_embd_out` if set and fallbacks to `hparams.n_embd` |
||
|
|
3d86c6c2b5 |
model: support GLM4V vision encoder (#18042)
* convert ok * no deepstack * less new tensors * cgraph ok * add mrope for text model * faster patch merger * add GGML_ROPE_TYPE_MRNORM * add support for metal * move glm4v do dedicated graph * convert: add norm_embd * clip: add debugging fn * working correctly * fix style * use bicubic * fix mrope metal * improve cpu * convert to neox ordering on conversion * revert backend changes * force stop if using old weight * support moe variant * fix conversion * fix convert (2) * Update tools/mtmd/clip-graph.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * process mrope_section on TextModel base class * resolve conflict merge --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> |
||
|
Chris PetersonandGitHub
|
2aa45ef9e3 | llama: Include algorithm header needed for C++23 (#18078) | ||
|
Georgi GerganovandGitHub
|
609a2d0268 |
models : fix YaRN regression + consolidate logic (#18006)
* models : fix YaRN regression + consolidate logic * cont : fix the fix * cont : remove header * cont : add header |
||
|
Georgi GerganovandGitHub
|
7bed317f53 |
models : fix the attn_factor for mistral3 graphs + improve consistency (#17945)
* models : fix the attn_factor for mistral3 graphs * cont : rework attn_factor correction logic * cont : make deepseek2 consistent * cont : add TODO * cont : special-case DSv2 * cont : revert Mistral 3 Large changes * cont : fix DS2 to use the original attn_factor * cont : minor comments |
||
|
Sigbjørn SkjæretandGitHub
|
9008027aa3 |
hparams : add n_embd_inp() to support extended embed (#16928)
* add n_embd_full to support extended embed * don't change output * rename to n_embd_inp * restore n_embd where applicable |
||
|
|
d261223d24 |
model: add support for qwen3vl series (#16780)
* support qwen3vl series.
Co-authored-by: Thireus ☠ <Thireus@users.noreply.github.com>
Co-authored-by: yairpatch <yairpatch@users.noreply.github.com>
Co-authored-by: LETS-BEE <LETS-BEE@users.noreply.github.com>
* bugfix: fix the arch check for qwen3vl-moe.
* use build_ffn
* optimize deepstack structure
* optimize deepstack feature saving
* Revert "optimize deepstack feature saving" for temporal fix
This reverts commit
|
||
|
Daniel BeveniusandGitHub
|
a2fba89a42 |
hparams : add check for layer index in is_recurrent (#16511)
* hparams : add check for layer index in is_recurrent This commit adds a check in the is_recurrent method to ensure that the provided layer index is within the valid range. The motivation for this change is to prevent potential out-of-bounds and also be consistent with other methods in the class that perform similar checks, like is_swa. |
||
|
Georgi GerganovandGitHub
|
c610b6c11b |
kv-cache : fix SWA checks + disable cacheless iSWA (#15811)
ggml-ci |
||
|
Daniel BeveniusandGitHub
|
fb15d649ed |
llama : add support for EmbeddingGemma 300m (#15798)
This commit add support for the EmbeddingGemma 300m. This model supports sliding window attention (SWA) and a new swq_type is introduced to support symmetric SWA masking. This commit also extracts the code from the function llama_is_masked_swa in llama-impl.h, so that the logic can be shared by both llm_graph_input_attn_no_cache::set_input and llama_kv_cache::set_input_kq_mask. With this commit the EmbeddingGemma 300m model can be converted to to GGUF and used with llama.cpp. Once the model has been uploaded to HuggingFace it can be used like this: ```console ./build/bin/llama-cli -hf ggml-org/embeddinggemma-300m-GGUF:Q8_0 ``` |
||
|
Georgi GerganovandGitHub
|
b730706a49 |
kv-cache : support layer reuse (#15504)
* kv-cache : support layer reuse ggml-ci * cont : update comments [no ci] |
||
|
|
6c6e397aff |
model : add support for SmallThinker series (#14898)
* support smallthinker * support 20b softmax, 4b no sliding window * new build_moe_ffn_from_probs, and can run 4b * fix 4b rope bug * fix python type check * remove is_moe judge * remove set_dense_start_swa_pattern function and modify set_swa_pattern function * trim trailing whitespace * remove get_vocab_base of SmallThinkerModel in convert_hf_to_gguf.py Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com> * better whitespace Apply suggestions from code review Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com> * use GGML_ASSERT for expert count validation Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com> * Improve null pointer check for probs Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com> * use template parameter for SWA attention logic * better whitespace Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * move the creation of inp_out_ids before the layer loop * remove redundant judge for probs --------- Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> |
||
|
|
225e7a1438 |
llama : add high-throughput mode (#14363)
* kv-cache : prepare K/V buffers for separation ggml-ci * batched-bench : fix oob write ggml-ci * llama : add "virtual sequences" ggml-ci * llama : use "stream" vs "virtual sequence" ggml-ci * graph : fix stream splitting when KV cache is not used ggml-ci * kv-cache : add multi-stream save/load support ggml-ci * llama : add "--attn-streams" flag ggml-ci * kv-cache : fix handling when find_slot fails ggml-ci * kv-cache : restore find_slot impl ggml-ci * kv-cache : add comments * kv-cache : add bounds checks for sequence id ggml-ci * cont : add n_seq_max to batch allocr ggml-ci * kv-cache : perform stream copies lazily after llama_synchronize ggml-ci * kv-cache : avoid throwing exceptions across the C boundary ggml-ci * CUDA: 4D FlashAttention support (#14628) * CUDA: 4D FlashAttention support * CUDA: fix WMMA FA kernel * llama : rename attn_streams -> kv_unified ggml-ci * common : rename kv_split -> kv_unified ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> |
||
|
Tarek DakhranandGitHub
|
f5e96b368f |
model : support LiquidAI LFM2 hybrid family (#14620)
**Important** LFM2 was [merged ](https://github.com/huggingface/transformers/pull/39340)into transformers, but has not yet been released. To convert into gguf, install transformers from source ```shell pip install "transformers @ git+https://github.com/huggingface/transformers.git@main" ``` |
||
|
compiladeandGitHub
|
5d46babdc2 |
llama : initial Mamba-2 support (#9126)
* llama : initial Mamba-2 support * ggml : SIMD ggml_ssm_scan for Mamba-2 * ggml : improve ggml_mul speed when masking recurrent states * llama : support running Mamba-Codestral-7B-v0.1 * llama : fix Mamba-2 conv state saving * ggml : make the ggml_mul fast broadcast path more consistently formatted * llama : remove unused variable * llama : add missing break * convert_hf : prefer SentencePiece tokenizer for Mamba-2 when present The tokenzier.json of Mamba-Codestral-7B-v0.1 otherwise requires workarounds to work correctly. * llama : avoid redundant state copy for Mamba 1 and 2 * metal : attempt to adapt SSM_SCAN for Mamba-2 * metal : fix SSM_SCAN pipeline scope * metal : use log and exp instead of log1pf and expf in SSM_SCAN * metal : remove unused arguments for SSM_SCAN The max index is 31, so trimming the arguments is necessary. * metal : add back n_seqs to SSM_SCAN args Whoops, this is needed for the offset in the concatenated output. * metal : fix SSM_SCAN state head offset * metal : fix wrong number of tokens per sequence in SSM_SCAN * ggml : remove unused fast broadcast path in GGML_MUL This was initially added because states were masked with ggml_mul, but this is no longer done and so this "optimisation" is no longer necessary, or at least not worth the additional code complexity. * ggml : avoid multiply by D in GGML_OP_SSM_SCAN This makes the weight buft detection in src/llama.cpp simpler. * convert : transpose Mamba-2 A, D and reshape SSM_NORM This breaks existing conversions of Mamba-2 models to avoid some reshapes. Not sure if it's a good idea, but it makes the graph slightly cleaner. * llama : more appropriate SSM_SCAN and SSM_CONV buft support checks * convert : fix flake8 lint * metal : fix confusion between ; and , * metal : add missing args for nb references in ssm_scan_f32_group * metal : single-user mamba2 inference works * kv-cache : remove const_cast when setting inputs for s_copy And also fix multi-user inference for recurrent models by using cell_id instead of i as the kv cell index when populating s_copy. * convert : avoid AutoConfig for Mamba and Mamba2 hparams * kv-cache : allow context shift for recurrent models * graph : fix recurrent state copies when avoiding copies Works, but using lambda functions might not be that clean. * ggml : fix mamba2 ssm scan when compiled with SVE * ggml-cpu : reorder SVE FMA for consistency with other SIMD arches * cuda : implement ssm scan for Mamba2 There is still room for improvement, but it works! * cuda : adapt Mamba1 ssm scan to shape changes from Mamba2 * mamba : fix mismatched new and delete size for llm_build_mamba Subclasses of llm_graph_context cannot have extra fields, because the called destructor is not the one from the subclass. This otherwise would cause problems when runnning Mamba-(1|2) inference when compiled -DGGML_SANITIZE_ADDRESS=ON * cuda : graceful fallback for Mamba-1 models with weird embd size |
||
|
Georgi GerganovandGitHub
|
4c9fdfbe15 |
ubatch : new splitting logic (#14217)
ggml-ci |
||
|
|
edc4a29eff |
memory : Hybrid recurrent cache (#13979)
* feat: Add llama_model_is_hybrid API call Also, split llama_model_is_recurrent into llm_arch_is_recurrent in llama-arch with llama_model_is_recurrent delegating to llm_arch_is_recurrent. The same split is done for hybird. This is needed because there are places where the llama_model has not yet been initialized but we need to check if the model is recurrent (specifically for the per-layer recurrent check array in hparams). Branch: GraniteFour Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * feat: Add c++ side constants for attention layer indices hparam Branch: GraniteFour * feat: Add support for distinguishing recurrent vs non-recurrent layers in hparams Branch: GraniteFour Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * feat: Auto-fill hparams.recurrent_layer_arr based on whether the model is recurrent Branch: GraniteFour Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * refactor: rename *_is_hybrid -> *_is_hybrid_recurrent The implementation of the hybrid cache intentionally does not specify the types of the child caches, so there was a naming mismatch with these predicate functions that used "hybrid" to imply "hybrid recurrent." Branch: HybridCache Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * feat: Add layer filter to recurrent cache Branch: HybridCache Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * fix: Use per-layer sizing everywhere in kv caches Branch: GraniteFour Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * feat: First pass at llama_kv_cache_hybrid_recurrent This follows the pattern in iswa where the two child caches are held explicitly to support the case where a model requires a single attention cache and a single recurrent cache where each layer uses exactly one of the caches. This is a rewrite of the more generic approach in the original hybrid cache PR: https://github.com/ggml-org/llama.cpp/pull/13276 Branch: HybridRecurrentCache Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * feat: Construct hybrid recurrent cache for hybrid recurrent models This includes a refactor of the create_memory logic to avoid needing to use the arch enum explicitly unless a model needs explicit cache instantiation logic beyond the standard logic for recurrent, hybrid, unified, and iswa. Branch: HybridRecurrentCache Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * fix: Fix wrong bool condition for split equal in hybrid cache Branch: HybridRecurrentCache Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * fix: Fix shift logic to defer to unified cache Branch: HybridRecurrentCache Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * feat: Support hybrid recurrent in llama-graph NOTE: I intentionally did not add support for s_mask since it will be going away soon Branch: HybridRecurrentCache Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * fix: Fix logic for initializing inputs and attn layers for hybrid caches Branch: GraniteFour Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * fix: Update recurrent cache for changes to remove intermediate kv_cache interface Branch: HybridRecurrentCache Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * fix: Fix status for init_update sig for recurrent cache state Branch: GraniteFour Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * fix: Add missing padding to n_ctx for hybrid cache construction Branch: GraniteFour Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * fix: Update clear signature for data argument after rebase Branch: HybridRecurrentCache Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * fix: Remove errant virtual destructor leftover from previous impl attempt Branch: HybridRecurrentCache Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * fix: Use per-layer n_embd_k/v_s calls for mamba (1) layers Branch: HybridRecurrentCache Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * refactor: Remove n_embd_k/v_s from unified cache No longer needed now that unified isn't also supporting recurrent https://github.com/ggml-org/llama.cpp/pull/13979#discussion_r2140761069 Branch: HybridRecurrentCache * refactor: Remove layer index from n_embd_k/v_s Now that it's not used at all in the unified cache, we don't need to use the layer index to zero it out for attention layers. Branch: HybridRecurrentCache Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * refactor: Remove n_embd_k/v_gqa from recurrent cache This is no longer needed now that there are separate implementations https://github.com/ggml-org/llama.cpp/pull/13979#discussion_r2140825128 Branch: HybridRecurrentCache Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * feat: Allow custom layer filters for hybrid recurrent This should help support architectures like Falcon H1 where there is overlap between layers that need attention and recurrent caches. https://github.com/ggml-org/llama.cpp/pull/13979#discussion_r2140748922 Branch: HybridRecurrentCache Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * fix: Remove logits_all after rebase Branch: HybridRecurrentCache Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * fix: Remove llama_model_is_hybrid_Recurrent public API https://github.com/ggml-org/llama.cpp/pull/13979#discussion_r2141728423 Branch: HybridRecurrentCache Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * refactor: Use llama_memory_state_ptr for child states in hybrid memory state Branch: HybridRecurrentCache Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * feat: Overhaul build_recurrent_state / build_inp_s_copy to match attention pattern https://github.com/ggml-org/llama.cpp/pull/13979/files#r2141701738 This is a big overhaul to bring consistency between how inputs and per- layer components are created for attention layers and recurrent layers. The main changes are: - Rename class llm_graph_input_s_copy -> llm_graph_input_rs - Add a corresponding llm_graph_input_rs_hybrid_recurrent - Rename build_inp_s_copy -> build_rs_inp_recurrent - Add a corresponding build_rs_inp_hybrid_recurrent - Rename build_recurrent_state -> build_rs to match build_attn w/ llm_graph_input_rs android-build AUTHORS bamba-9b-2.2T.gguf bamba-9b-2.2T.q4_k_m.gguf broken.log build build-rel build-xcframework.sh build.android build.android.bak ci cmake CMakeLists.txt CMakePresets.json CODEOWNERS common common.o CONTRIBUTING.md convert_hf_to_gguf_update.py convert_hf_to_gguf.py convert_llama_ggml_to_gguf.py convert_lora_to_gguf.py debug.log docs examples flake.lock flake.nix ggml ggml-alloc.o ggml-backend.o ggml-metal.o ggml-model-BF16.gguf ggml-model-Q4_K_M.gguf ggml-quants.o ggml.o gguf-py grammar-parser.o grammars include LICENSE licenses llama.log llama.o llamacpp_trace.log main.log Makefile media models mypy.ini pocs poetry.lock prompts pyproject.toml pyrightconfig.json q4_k_m_boot.log q8_0_boot.log quant.log quant2.log README.md requirements requirements.txt sampling.o scripts SECURITY.md src test-grammar-output.tmp test-json-schema-input.tmp tests tools vendor working.log as the first input - Add a corresponding overload of build_rs w/ llm_graph_input_rs_hybrid_recurrent android-build AUTHORS bamba-9b-2.2T.gguf bamba-9b-2.2T.q4_k_m.gguf broken.log build build-rel build-xcframework.sh build.android build.android.bak ci cmake CMakeLists.txt CMakePresets.json CODEOWNERS common common.o CONTRIBUTING.md convert_hf_to_gguf_update.py convert_hf_to_gguf.py convert_llama_ggml_to_gguf.py convert_lora_to_gguf.py debug.log docs examples flake.lock flake.nix ggml ggml-alloc.o ggml-backend.o ggml-metal.o ggml-model-BF16.gguf ggml-model-Q4_K_M.gguf ggml-quants.o ggml.o gguf-py grammar-parser.o grammars include LICENSE licenses llama.log llama.o llamacpp_trace.log main.log Makefile media models mypy.ini pocs poetry.lock prompts pyproject.toml pyrightconfig.json q4_k_m_boot.log q8_0_boot.log quant.log quant2.log README.md requirements requirements.txt sampling.o scripts SECURITY.md src test-grammar-output.tmp test-json-schema-input.tmp tests tools vendor working.log as the first input - Add a llm_graph_input_attn_kv_hybrid_recurrent analogous to llm_graph_input_attn_kv_unified - Add a build_attn override that takes llm_graph_input_attn_kv_hybrid_recurrent android-build AUTHORS bamba-9b-2.2T.gguf bamba-9b-2.2T.q4_k_m.gguf broken.log build build-rel build-xcframework.sh build.android build.android.bak ci cmake CMakeLists.txt CMakePresets.json CODEOWNERS common common.o CONTRIBUTING.md convert_hf_to_gguf_update.py convert_hf_to_gguf.py convert_llama_ggml_to_gguf.py convert_lora_to_gguf.py debug.log docs examples flake.lock flake.nix ggml ggml-alloc.o ggml-backend.o ggml-metal.o ggml-model-BF16.gguf ggml-model-Q4_K_M.gguf ggml-quants.o ggml.o gguf-py grammar-parser.o grammars include LICENSE licenses llama.log llama.o llamacpp_trace.log main.log Makefile media models mypy.ini pocs poetry.lock prompts pyproject.toml pyrightconfig.json q4_k_m_boot.log q8_0_boot.log quant.log quant2.log README.md requirements requirements.txt sampling.o scripts SECURITY.md src test-grammar-output.tmp test-json-schema-input.tmp tests tools vendor working.log as the first input This makes the two paradigms fully consistent. The main drawback is the code duplication in the build_attn and build_rs implementations where the only difference between implementations is how they cast the memory state. Branch: HybridRecurrentCache Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * fix: Fix resize vs reserve and skip null tensors in size computation https://github.com/ggml-org/llama.cpp/pull/13979/files#r2149469788 Branch: HybridRecurrentCache Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> Co-Authored-By: @younesbelkada * fix: Fix initialization of child states Since initially writing this PR, the logic in the child state types changed such that using the "init full" signature and keeping the ubatches on the parent struct no longer worked. Branch: HybridRecurrentCache Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * refactor: Use a common build_recurrent_state method that is cache-agnostic This reduces the code duplication between the different build_rs impls and also retains a similar signature to the previous build_recurrent_state method while standardizing on the input-dispatched build_rs implementation. Branch: HybridRecurrentCache Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * recurrent : rework graph inputs + add TODOs ggml-ci * refactor: Make status and child states const in hybrid and iswa Branch: HybridRecurrentCache Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * refactor: Rename llama_kv_cache_[recurrent|hybrid_recurrent] to remove kv cache This removes the notion of "kv" from the interface names for these memory types. There are still many references to kv in the implementation of the recurrent memory which will need further adjustment. Branch: HybridRecurrentCache Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * refactor!: Rename all k/v related values for recurrent/hybrid to r/s Anywhere that "kv_<state|cell|size|etc>" is used, I've used the more generic "mem_" prefix. The specifics of "k" (key) translate to "r" (recurrent state) and "v" (value) translate to "s" (state-space embedding states). Branch: HybridRecurrentCache Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * refacor: _recurrent -> _recr for brevity It just _happens_ to have the same number of letters as _attn! Branch: HybridRecurrentCache Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * style: Fix spacing for ref Branch: HybridRecurrentCache Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * refactor: recurrent_layer() -> is_recurrent() Branch: HybridRecurrentCache Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * style: Fix spacing for size_s_bytes declaration Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> |
||
|
Georgi GerganovandGitHub
|
d13d0f6135 |
hparams : initialize arrays (#13728)
ggml-ci |
||
|
Xuan-Son NguyenandGitHub
|
8a2afb7520 | llama : allow custom list of swa_layers (#13726) | ||
|
Georgi GerganovandGitHub
|
8e186ef0e7 |
hparams : support models for which all layers use SWA (#13682)
ggml-ci |
||
|
Georgi GerganovandGitHub
|
081bee8c64 |
hparams : add SWA rope parameters (#12374)
ggml-ci |
||
|
Georgi GerganovandGitHub
|
84d5475541 |
llama : fix Gemma3 SWA KV cache shift (#12373)
* llama : fix Gemma3 SWA KV cache shift ggml-ci * hparams : add comment [no ci] |
||
|
|
ee7136c6d1 |
llama: add support for QRWKV6 model architecture (#11001)
llama: add support for QRWKV6 model architecture (#11001) * WIP: Add support for RWKV6Qwen2 Signed-off-by: Molly Sophia <mollysophia379@gmail.com> * RWKV: Some graph simplification Signed-off-by: Molly Sophia <mollysophia379@gmail.com> * Add support for RWKV6Qwen2 with cpu and cuda GLA Signed-off-by: Molly Sophia <mollysophia379@gmail.com> * RWKV6[QWEN2]: Concat lerp weights together to reduce cpu overhead Signed-off-by: Molly Sophia <mollysophia379@gmail.com> * Fix some typos Signed-off-by: Molly Sophia <mollysophia379@gmail.com> * code format changes Signed-off-by: Molly Sophia <mollysophia379@gmail.com> * Fix wkv test & add gla test Signed-off-by: Molly Sophia <mollysophia379@gmail.com> * Fix cuda warning Signed-off-by: Molly Sophia <mollysophia379@gmail.com> * Update README.md Signed-off-by: Molly Sophia <mollysophia379@gmail.com> * Update ggml/src/ggml-cuda/gla.cu Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix fused lerp weights loading with RWKV6 Signed-off-by: Molly Sophia <mollysophia379@gmail.com> * better sanity check skipping for QRWKV6 in llama-quant thanks @compilade Signed-off-by: Molly Sophia <mollysophia379@gmail.com> Co-authored-by: compilade <git@compilade.net> --------- Signed-off-by: Molly Sophia <mollysophia379@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: compilade <git@compilade.net> |
||
|
Georgi GerganovandGitHub
|
f66f582927 |
llama : refactor src/llama.cpp (#10902)
* llama : scatter llama.cpp into multiple modules (wip) * llama : control-vector -> adapter * llama : arch * llama : mmap ggml-ci * ci : remove BUILD_SHARED_LIBS=OFF ggml-ci * llama : arch (cont) ggml-ci * llama : chat ggml-ci * llama : model ggml-ci * llama : hparams ggml-ci * llama : adapter ggml-ci * examples : fix ggml-ci * rebase ggml-ci * minor * llama : kv cache ggml-ci * llama : impl ggml-ci * llama : batch ggml-ci * cont ggml-ci * llama : context ggml-ci * minor * llama : context (cont) ggml-ci * llama : model loader ggml-ci * common : update lora ggml-ci * llama : quant ggml-ci * llama : quant (cont) ggml-ci * minor [no ci] |