llama-swap SIGTERMs a model when the owner switches models (only one large
model fits in RAM), and switching back re-prefills the whole conversation:
about 150 s for a 30k-token chat on the 35B. With --cache-disk-path DIR,
each slot with a prompt is saved to DIR on shutdown. After the restart, a
request whose prompt starts with a saved prompt is restored from the file,
so only the suffix is processed. Off by default.
Save, in server_context::start_loop() after the loop returns:
- disk_cache_collect() copies the state of each slot with >= 256 text tokens
to host memory (llama_state_seq_get_data_ext, flags NONE) while the device
buffers still exist. A cold server reads its released state through
llama_context_restore_memory_host(), with no device use and no extra copy.
- vram_share_exit() then frees the VRAM and releases the GPU token as before,
so only the D2H copy delays the token release, never the disk IO.
- disk_cache_flush() writes the files. A slot is skipped if its KV does not
cover its token list (seq_pos_max + 1 != n_tokens: the last batch was not
decoded) or if the same token list is already on disk.
File (server_prompt_disk_cache, server-task.cpp):
- A 48-byte header (magic, version, key hash, sizes, payload hash), then a
payload of 8-byte padded sections: tokens, checkpoint table, state, the
checkpoint states. The hash is FNV-1a over the 64-bit words of the payload.
- Plain fwrite to <name>.tmp, fdatasync, POSIX_FADV_DONTNEED, rename. The read
is a plain fread plus DONTNEED. There is no mmap: upstream #28092 livelocked
the kernel on AMD + Vulkan by DMA-ing into a MAP_SHARED file.
- The index is built at startup from the headers. A bad magic, a size that
does not match the header, or an inconsistent table means the file is
removed. A hash mismatch at load also removes it, and nothing is restored.
Other keys and other versions are kept. A .tmp older than 1 h is removed.
- --cache-disk-size N MiB (default 4096, 0 = no limit): before a write, the
least recently used *.lpcd files of every model in DIR are removed until the
new file fits; a restore updates the file mtime. Entries the new prompt
extends are dropped after the write and do not count against the limit.
A write also needs its size + 1 GiB free on the file system. A write error
or a full disk is logged; the file is not saved and the exit stays clean.
Key: the build commit, the model path + size + mtime, LoRA and control
vectors, n_ctx, n_ctx_seq, n_seq_max, n_ubatch, kv_unified, swa_full, the
K/V types, the flash-attn type, and the RoPE/YaRN parameters.
Restore, in get_available_slot() after the RAM cache:
- Only while warm (holding the GPU token), with cache_prompt on and no media
tokens. The file must reuse at least 256 more tokens than the slot holds.
- Reusable means one of two things. Either the saved token list is an exact
prefix of the new prompt, or (for a recurrent/SWA model) a saved checkpoint
at or below the common prefix can be used. A plain-KV model can be cut back
to any prefix. Exact prefix alone is not enough for OpenWebUI on Qwen: the
template rewrites the previous assistant turn (the reasoning is dropped), so
the saved list is almost never a prefix of the next turn. The checkpoints
are exact snapshots, and update_slots() rolls back to one exactly as it does
for the RAM cache. They are restored with id_task -1, so one that fails to
load means a re-prefill, not an abort.
- The slot's current prompt goes to the RAM cache first. A set_data failure
clears the slot and removes the file.
Not supported with speculative decoding (warning, disabled): the draft state
is not saved. Only the slots are saved, not the RAM cache entries.
Measured on the pod, CPU only, Qwen3.5-0.8B (hybrid, 6 of 24 layers attention):
- 573-token chat: 64.5 MiB file with 2 checkpoints, copy 18 ms, write 171 ms,
exit 314 ms (125 ms without the cache). After the restart: read 166 ms,
restore 3 ms, turn 2 prompt_n 58 of 596 in 462 ms against a 2913 ms full
prefill. The answer is identical to a server that never restarted.
- 12k tokens: 200.7 MiB, copy 110 ms, write 350 ms, exit 641 ms, peak RSS
+162 MiB during the exit save. After the restart: read 307 ms, restore 18 ms,
turn 2 prompt_n 35 of 12188 in 1.2 s against a 72 s prefill.
- The /root PVC writes at about 0.66 GB/s with fdatasync. /tmp (the node
overlay) stalled once: 5.8 s for 64.5 MiB.
llama-swap v239 waits for the old process to exit before it starts the next
model, and SIGKILLs a process that is still there after 30 s. So the exit
save adds to the model switch time: about 1 s per GB on the PVC.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Comments only, plus one log string:
- the arbiter block comment said the KV cache is never evicted; with
LLAMA_SLEEP_EVICT_KV it is (and the scheduler and scratch with it)
- vram_go_cold said vram_ensure_warm() rebuilds the encoder; only
encode_mtmd_chunk() restores it, and only for an encode
- vram_ensure_warm and the mtmd init block named process_mtmd_chunk; the
restore lives in encode_mtmd_chunk
- handle_sleeping_state said waking happens lazily before decode; it
restores eagerly on the wake transition, before update_slots()
- the vram_cold ordering note now says why it still holds with the inline
want-lock probe
- the sleep log line said "keeping KV cache" even when LLAMA_SLEEP_EVICT_KV
evicts it; it now names what actually happens (the prefix
"server entering sleeping state" that fork-validate.sh counts is kept)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Two ways a shutting-down server kept the VRAM arbiter's token, stalling every
other model:
- SIGTERM with a request in flight: terminate() stopped only the task queue.
The HTTP handler kept waiting for results the loop would never send, the main
thread blocked joining it, and the token was released only in the destructor
after that join - i.e. never, until the client went away. Terminate the
result queue too (readers now get "server is shutting down": a 503, or an SSE
error event), and let wait_until_no_sleep give up once stopping.
- SIGTERM while waiting for the token: the loop thread sat in a blocking flock
inside the wake callback, holding mutex_tasks, so terminate() waited for the
current holder to finish and then for a pointless restore of our weights.
Poll the lock non-blocking and abort the wait on shutdown; the aborted wake
stays cold, skips the restore, and the loop exits before update_slots.
Also release the token right after start_loop() returns rather than in the
destructor, and only after the device buffers are actually freed: the old
order unlocked before destroy(), so the next model could start uploading into
VRAM that was still occupied. Freeing at exit uses a new
llama_context_discard_device(), which drops weights, KV, MoE cache and scratch
without capturing host shadows first.
Off production: SIGTERM mid-generation or mid-wait now exits in ~0.2 s with
exit 0 and frees the token within 5-70 ms (previously never / after the
holder's whole generation). Handoff latency and greedy output are unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A SIGTERM that lands while the start_loop() thread is inside a sleeping
callback deadlocks the server. The callbacks run under
server_queue::mutex_tasks, and the kernel delivers the signal to that same
thread; the handler calls terminate(), which locks mutex_tasks again, and the
thread waits on itself forever.
With the VRAM arbiter that window is wide - a whole device release, or a
blocking wait for the GPU token - and llama-swap sends SIGTERM to evict
models, so it hit production: an evicted model wedged mid-release, never
released token.lock, and every other model queued behind it. SIGTERM could
not stop it either; only SIGKILL did.
Make the handler only write the signal number to a self-pipe, which is
async-signal-safe, and run shutdown_handler on a dedicated thread. terminate()
there simply waits until the loop thread drops mutex_tasks, after which the
loop exits normally. The thread is joined before the objects the handler uses
are destroyed. If pipe() fails the old direct call is kept.
Reproduced 6/6 off production (SIGTERM during release and during the token
wait); with the fix 8/8 clean exits, and handoffs, image requests and greedy
output are unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The doorbell was edge-triggered: a waiter wrote <arena>/doorbell/<pid> and the
holder's warden turned the inotify event into a queue flag. A ring carried no
notion of "still wanted", so it was lost whenever the holder was not already warm
and listening - during its model load (the watch does not exist yet, and the
kernel does not queue events for a watch that is not there) or inside
restore_device - and any ring that survived past the point its sender had been
satisfied caused a spurious release. Lost rings wedged the waiter permanently,
because flock(LOCK_EX) never times out and, with idle-sleep disabled, the holder
had no other reason to release. The wait blocks under mutex_tasks, so the whole
server stopped answering while /health still returned 200.
Express the request as kernel state instead. A waiter holds <arena>/want.lock
shared while it waits and drops it once it owns the token; the sleep decision
probes that lock non-blocking and releases the GPU while anyone is waiting. The
probe needs its own fd - flock treats two open file descriptions of one file
independently, so probing on the waiter's fd would convert our own lock rather
than conflict with it. Nothing can be missed, nothing goes stale, and a waiter
that dies is cleaned up by the kernel.
Probe from should_sleep() on the loop thread rather than from a warden thread.
Routing it through a flag is what made the first attempts fail: start_loop()
holds mutex_tasks from should_sleep() through the callbacks to the wait, so a
warden's request_yield() blocks on that mutex and is admitted only after the flag
has been consumed, latching a release for the next wake. Reading live state where
the decision is made has no edge to latch, and drops the warden's poll latency.
Two sleep-path bugs this exposed: wait_until_no_sleep() waited on !sleeping but
the loop clears req_stop_sleeping on the way in, so a loop that slept again
before the waiter ran stranded it forever - re-ask on every wake. And a task
queued after the waiter saw us awake could not wake us by itself, so sleep now
also breaks on a non-empty queue. Hold the GPU for 100 ms after a wake: the
request that woke us is not queued yet, and yielding at once only sends it round
again.
Measured on the RX 580 pod, two servers contending, 60 alternating handoffs:
0 stranded, median 0.309 s against the doorbell's 0.314 s.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PZz44SLQvTXMyWGio6t9DZ
The vision encoder sits idle in VRAM on every text request. With
LLAMA_MMPROJ_ONDEMAND=1 the server keeps the encoder in RAM (released
after load) and brings it into VRAM just-in-time before mtmd_batch_encode,
releasing it again afterwards - so the encoder's VRAM is free for KV /
expert cache on the common text-only path.
When the encoder does not fit (that VRAM has been claimed), evict the LLM
backbone weights first: they are not needed while the encoder runs (the
decode that uses them happens afterwards), their host shadow is read-only
(cheap free, no D2H), and the KV / prompt cache is left untouched. Order
on the way back matters: free the encoder before re-uploading the weights
so the peak stays within VRAM.
Also wire the encoder release/restore into the VRAM arbiter: vram_go_cold
releases it; the just-in-time encode-path restore replaces the previous
restore in vram_ensure_warm.
Assisted-by: Claude
With LLAMA_SLEEP_EVICT_KV the KV cache device buffers are freed when a model
goes cold. update_slots() touches the KV before the decode-time vram_ensure_warm
(e.g. SWA models create a checkpoint that reads the KV via ggml_backend_tensor_get),
so the KV must already be resident by then. Move the restore to the sleep-wake
handler (handle_sleeping_state(false)), which runs before update_slots, fixing a
GGML_ASSERT(buffer) crash on iSWA models (gemma) when KV eviction is enabled.
Assisted-by: Claude
Before uploading a model's weights, acquire the shared VRAM token (ring the
doorbell so any resident model releases first). Previously load uploaded weights
to VRAM before the arbiter was active, so loading a large model while another
(e.g. the warm 4B task model) held VRAM could exceed the budget and OOM.
- add vram_arena_open() (idempotent flock/doorbell setup) and
vram_acquire_for_load(), called from load_model() before
common_init_from_params().
- a coordinated load now stays warm holding the token and serves its first
request without a re-warm (drop the init-time go_cold cold-start).
Assisted-by: Claude
The request handler only calls wait_until_no_sleep() (which wakes a server out
of its sleeping state) when sleep_idle_seconds >= 0. But the VRAM arbiter's
cross-process doorbell can put a server to sleep even when idle-sleep is
disabled, so without this a doorbell-slept server would never wake and requests
to it would hang until timeout.
Do not bypass the wake path when LLAMA_SLEEP_VRAM_ONLY is set, so the arbiter no
longer depends on --sleep-idle-seconds being configured.
Assisted-by: Claude
Extend on-demand device residency to the KV cache so that when a model's KV
plus another model would not fit in VRAM, the KV can also be evicted to a host
shadow (D2H on release, H2D on restore) instead of only the weights.
- llama_memory_i: add release_device_buffers()/restore_device_buffers()
(default no-op). Implemented in llama_kv_cache (D2H shadow of the live
ctxs_bufs, freed and reallocated like the weights); llama_memory_hybrid and
llama_kv_cache_iswa delegate to their child caches.
- llama_context::release_device(evict_kv): also evict the memory's device
buffers when requested; restore_device() rebuilds them. Public API
llama_context_release_device gains an evict_kv flag.
- server: LLAMA_SLEEP_EVICT_KV=1 enables it. Off by default (weights-only),
since the KV shadow adds a D2H/H2D copy of the live cache each cycle.
Validated on RX 580 (Vulkan), 4B @ 32k ctx: weights-only cold VRAM 1750 MB
(KV stays); weights+KV cold VRAM 726 MB (KV freed, ~1 GB reclaimed). KV
survives the round-trip: prompt cache reused after the cycle (prompt_n 4 vs
42), correct output.
Assisted-by: Claude
Add release/restore of a model's GPU weight buffers (keeping a host shadow
and the KV cache) so several always-loaded llama-server processes can
time-share a single GPU without reloading or losing the prompt cache.
- llama-model: release_device_weights()/restore_device_weights() capture a
compact host shadow (stable iteration order, view-skipping) and free then
realloc the device weight buffers; weights_resident() query.
- llama-context: release_device()/restore_device() wrappers; decode() auto-
restores; public C API llama_context_release_device/restore_device.
- server: LLAMA_SLEEP_VRAM_ONLY makes idle-sleep release only the VRAM weights
(not a full unload/reload). A cross-process flock token in LLAMA_VRAM_ARENA
enforces "resident iff holds token"; an inotify doorbell forces the holder
to release on contention. The warden thread only touches the task queue, so
releases run on the loop thread and never race a decode.
Validated on RX 580 (Vulkan): two models share 8GB, never both resident,
correct output under contention, KV cache preserved (no re-prefill).
Assisted-by: Claude
* mtmd: support hibiki realtime duplex
* fix and clean up
* fix long session quality
* fix ci
* tests : add the moshi fusion baseline for Metal
The Metal fusion check found 5 RMS_NORM+MUL fusions in the new moshi
arch and no baseline for them.
---------
Co-authored-by: Pascal <admin@serveurperso.com>
* convert: qwen3-asr: put "language ..." output into thinking block
* convert: qwen3-asr: read the chat template from models/templates
The converter embedded its own copy of Qwen3-ASR.jinja. Read the
bundled file instead, as the Kimi-K3 and Kokoro converters do, so
the GGUF and test-chat share a single template.
---------
Co-authored-by: Pascal <admin@serveurperso.com>
* cuda: add scatter for batched tensor writes
* update test case to mirror get_n test case
* save restore timing logs
* refactor code into batch-copy.cu file
* llama-bench : size the batch to the prompt for models without memory
Assisted-by: Claude Opus 5.5
* server : read up to 16 lfm3-d1 questions in one pass
Assisted-by: Claude Opus 5.5
* server: fix lora and recurrent
* tests: run the hybrid SWA checkpoint test with one token per ubatch
With and without the cache, the prompt was not split into the same
ubatches, and the generated tokens are decoded one at a time while the
uncached run processes them in a batch. On CUDA this moves the top
logprobs of the tiny model by up to 0.03 and swaps near ties, so the
test failed while the cache behaved as expected.
With one token per ubatch both runs go through the same kernels and
the output is compared exactly.
---------
Co-authored-by: Pascal <admin@serveurperso.com>
* jinja: impl scope-chain key lookup
* add test for call_statement lambda
* chat: add tts_dict for kokoro G2P template
* more testing
* value_float_t range check
* no json-in-meta, add llama_model_meta_arr_str/_n
* mtmd: add kokoro tts support
* detailed guide about G2P and text norm port
* refine G2P
* clean up
* migrate batch api
* nits
* Added support for Qwen3-TTS ICL/reference mode
Assisted-By: Claude Opus 5.5
* fix correctness
* attach transcription to bitmap
* nits
---------
Co-authored-by: Xuan Son Nguyen <son@huggingface.co>
* sampling : fix crash when both ignore_eos and grammar are on
* server : rework, throw 400 when grammar and ignore_eos are on
* sampling : move the check into the sampler (review suggestion)
* tests : remove redundant tests
* ui: replace deprecated packages and fix npm audit findings
Swap the mdast stub for @types/mdast, move the katex override to 0.19.0,
and override glob, source-map and html-encoding-sniffer under
workbox-build and http-server so installs no longer pull deprecated
packages. Apply the non-breaking npm audit fixes to the lockfile.
Assisted-by: Claude (Opus 5.5 Medium)
* ui: migrate Llama Chat to SvelteKit 3 and its built-in service worker
Move to SvelteKit 3, Vite 8, TypeScript 6 and Storybook 10.6. The
SvelteKit options move from svelte.config.js into sveltekit() in
vite.config.ts, $lib imports become #lib subpath imports from
package.json, and $app/environment becomes $app/env. Modules named
server.* or living under a server/ folder are server-only in SvelteKit 3,
so the llama-server components, store and enums move to llama-server
names.
@vite-pwa/sveltekit and Workbox give way to SvelteKit's own service
worker in src/service-worker, registered by usePwa(), which precaches the
app files and leaves API requests to the network. A small plugin emits
manifest.webmanifest. static/sw.js removes the worker of older builds, and
llama-server serves a self-removing service-worker.js as well as sw.js
when the built-in UI is off or replaced.
Assisted-by: Claude (Opus 5.5 Medium)
* ui: use SvelteKit 3's update detection and Svelte 5's current patterns
A page the service worker doesn't control now learns of a new build
from SvelteKit itself: `updated` from $app/state checks
_app/version.json on navigation, on focus and every hour. It replaces
the version Llama Chat kept in localStorage and compared by hand.
Reload on the update prompt pulls in the new worker and activates it
before reloading, so one click lands on the new build, whether or not
a worker controlled the page.
Following Svelte's current guidance:
- usePwa keeps its state in $state rather than a store, and no effect
checks `browser`
- nearViewport, autoResizeTextarea and the custom CSS injection are
attachments ({@attach}) rather than actions (use:)
- class lists use arrays rather than class: directives
- the chat contexts come from createContext, typed, in place of
setContext/getContext with string keys
- ActionIconCopyToClipboard takes $props() rather than `export let`
Assisted-by: Claude (Opus 5.5 Medium)
* ui: move to ESLint 10
ESLint 9 is out of support. Every plugin the app uses supports 10:
typescript-eslint, eslint-plugin-svelte, perfectionist,
simple-import-sort, eslint-plugin-storybook and eslint-config-prettier.
@eslint/js and @eslint/compat follow.
ESLint 10's recommended rules ask for two things the code now does:
errors rethrown from a catch keep the caught error as their `cause`,
and no variable starts with a value that every path overwrites.
Assisted-by: Claude (Opus 5.5 Medium)
* ui: override minimatch to clear npm audit's last finding
ESLint 10.14, @eslint/config-array 0.23.5 and typescript-eslint 8.72
depend on minimatch 10, which GHSA-g3hv-75wg-m249 (denial of service
via globparts amplification) affects below 11.0.1. Their fixed releases
move to minimatch 11 but are all younger than the repo's 7-day
min-release-age, so an override pins minimatch 11.0.5, the newest the
rule allows. The three only call new Minimatch(), match() and
minimatch(), which minimatch 11 keeps. npm audit now finds nothing.
Assisted-by: Claude (Opus 5.5 Medium)
* mtmd: support YuE2-3B
* improve speed
* add CFG
* fix streaming
* add cfg control from api
* check ctx size
* fix n_predict
* add flow_steps config
* nits
* fix batch api
* nits
* adapt
* mtmd: yue2 samples from its own saved logits
The context logits belong to the last decode of any slot, so with
several server slots the score and the unguided codes were sampled
from another sequence. Keep the logits of each own decode in
save_logits(), which also does the guidance mix.
* mtmd: let the yue2 snake chain fuse
The beta reshape landed between sqr and mul, so the mul, sin, sqr,
mul, add chain was not consecutive and the backends never fused it.
Expand both reshapes first, like qwen3tts does.
---------
Co-authored-by: Pascal <admin@serveurperso.com>
* ui: move tools/ui to ui/apps/chat
The chat app moves to its own root folder, ui/apps/chat, as the first step of splitting the UI into one library
(ui/lib) and the apps built on it (ui/apps). The code is unchanged; every path that pointed at tools/ui now points at
ui/apps/chat: CMake adds it from tools/, and the Dockerfiles, the Nix package, the workflows, the labeler, CODEOWNERS,
.editorconfig, the git hooks, the dev and demo scripts and the docs follow it.
Assisted-by: Claude (Opus 5.5 Medium)
* ui: name Llama Chat's package @ggml-org/llama-chat
It was still called llama-ui, the name of the library that now sits beside it
in ui/lib (@ggml-org/llama-ui). The package stays private; the app's display name
and the llama-ui.zip build artifact are unchanged.
Assisted-by: Claude (Opus 5.5 Medium)
* ui: call the built-in web UI Llama Chat in docs, help text and comments
The app in tools/ui is Llama Chat; Llama UI is the component library it is
built from. Prose that said "Web UI", "web UI" or "WebUI" for the app now
names it: the READMEs, SECURITY.md, the --ui help text, the CMake option
description, one server log line, code comments and the Storybook intro.
Names that others depend on are unchanged: the deprecated --webui flag
aliases, the LLAMA_*_WEBUI CMake and environment names, the webui_settings
field, the storage migration keys and the Nix attributes.
Assisted-by: Claude (Opus 5.5 Medium)
* ui: name tools/ui as a former path of Llama Chat in the code-review skill
Changes under the old tools/ui path, as well as tools/server/webui, still
get the Llama Chat checklist.
Assisted-by: Claude (Opus 5.5 Medium)
* ui: name Llama Chat's path in the security policy's exclusions
Llama Chat no longer lives under tools/server/, so the exclusion listed
under it now names ui/apps/chat. What is covered stays the same.
Assisted-by: Claude (Opus 5.5 Medium)
The reasoning effort probe already renders every candidate level, but
it dropped the ones the template raises on, so a request with such a
level (e.g. max on Qwen3.8) reached the template and came back as a
500 on /v1/chat/completions, /v1/responses and /v1/messages.
Keep the raising levels in reasoning_effort_rejected and reject them
in oaicompat_chat_params_parse with an invalid_argument, which the
server answers with a 400 listing the supported levels. Aliases and
levels a template ignores are not rejected, so high on Qwen3.8
still renders as xhigh.
Also trim the output_config.effort test to a single level.
* mtmd : allow custom video decoder
* also adapt to wllama requirements
* clean up
* add test
---------
Co-authored-by: Xuan Son Nguyen <son@huggingface.co>
* dflash : move the DSpark markov head and backbone graphs to dspark.cpp
Move-only: build_dspark_markov_head, graph_dsv4, graph_dsv41 and
graph_mistral_large4 move verbatim. build_inp_target_features and
build_dspark_markov_head become static members of llama_model_dflash so
both files can use them.
* dflash : detect the head and the backbone once, split the loaders per file
load_arch_hparams sets llama_model_dflash::variant (DFlash, DFlash2,
DSpark, LiliCorr) and ::backbone (base, DSV4, DSV4.1, Mistral Large 4).
The head, backbone and base layer loaders become member helpers in
dflash.cpp, dflash2.cpp, lilicorr.cpp and dspark.cpp. build_arch_graph
dispatches on the backbone.
The DFlash2 and LiliCorr loaders create the DFlash2 conv tensors through
a helper in dflash2.cpp, right after their head tensors instead of in
the base layer loop; the other tensors keep their creation order. The
DFlash2 loader drops its copy of the conv group size check, which master
never reached.
* dflash : share the injection input, token embeddings and lm head lookups
build_inject_inp (target features + fc + encoder norm), get_tok_embd and
get_lm_head replace the copies in the base, DSV4, DSV4.1 and Mistral
Large 4 graphs. build_inp_target_features is file-static again.
* dflash : split the DFlash decoder into a graph per head
graph_base creates the shared inputs and provides build_inject,
build_inp_tokens, build_layers and build_logits. graph_dflash
(dflash.cpp), graph_dflash2 (dflash2.cpp), graph_dspark (dspark.cpp) and
graph_lilicorr (lilicorr.cpp) assemble them and add their head; the
selector, scorer and markov head builders move to the file of their
graph, and the DFlash2 local conv run by build_layers to dflash2.cpp.
build_arch_graph dispatches on the backbone, then on the head.
The op order and the tensor names are unchanged.
* speculative : split the DFlash driver into a class per head
common_speculative_impl_draft_dflash stays the DFlash implementation and
becomes the base of draft_dspark, draft_dflash2 and draft_lilicorr. Its
protected constructor only reads the draft metadata; init() then sets up
the contexts with the options of the draft (draft length, backend
sampling, device features, masked h_nextn). Each derived class overrides
draft() with the path of its head from the former draft(). DFlash2 also
overrides begin() to reseed its generator per request, and LiliCorr
overrides process() to store the context features of each injected
chunk, which its noise block carries as its state.
The factory keeps the current routing: draft-dspark picks DSpark,
draft-dflash picks DFlash2 or LiliCorr from the draft metadata and
DFlash otherwise. draft-dspark with a DFlash2 or LiliCorr draft now runs
the DSpark reader instead of a mix of both. sample_from_anchor is logged
by DSpark only.
* speculative : add the draft-dflash2 and draft-lilicorr types
draft-dflash2 and draft-lilicorr name the DFlash2 and LiliCorr drafts,
which draft-dflash keeps running too: it picks them from the draft
metadata. draft-dflash2 and draft-lilicorr must match the draft model,
draft-dspark rejects a DFlash2 or LiliCorr draft, a model that is not a
DFlash draft is rejected, and a run uses at most one DFlash family type;
these errors leave speculative decoding uninitialized.
The type detected from the draft GGUF is now the concrete one, and
draft-dflash2 and draft-lilicorr download the dflash sidecar.
* docs : describe the DFlash2 and LiliCorr drafts
Add the DFlash2 and LiliCorr sections and the new types, and explain
that draft-dflash also runs DFlash2 and LiliCorr drafts.
* speculative : drop the unused DFlash is_mrope
Nothing reads it since the migration to common_batch (#29385), which
picks the number of M-RoPE position rows from the rope type of the draft.
* doc: update spec dec
* address review comments
* requirements : support Python 3.15
torch 2.11.0, numpy 2.4, matplotlib 3.10, pandas 2.3, pillow 11.3 and
cffi 1.16 have no Python 3.15 wheels, so the CI runner on 3.15 cannot
install the requirements.
Python 3.15 gets the first releases that ship 3.15 wheels through
version markers, and every version below 3.15 stays as is.
pyproject.toml now allows 3.15 and picks torch 2.15.0 there.
* requirements : apply review suggestions
numpy 2.5.4 from Python 3.14, matplotlib 3.11.2 and pandas 3.0.6 from
3.11, cffi 2.1.1 and pillow 12.3.0 for every version. cffi 1.16.0 has
no 3.14 wheel and fails to build there, so this also fixes the
tokenizer test requirements on 3.14.
* server: read a small temp file in the read_file tool test
test_tools_builtin_read_file read its own file, which grows with every
added test. On Windows the CRLF checkout puts it at 16664 bytes, over
the 16384 byte read limit, and the tool refuses it.
The test now writes a small file of its own and checks the exact
content read back.
* server: use a temp cwd in the x-tool-cwd tool test
test_tools_builtin_cwd_header read test_tools_builtin.py through the
x-tool-cwd header, which hits the same 16384 byte read limit on the
Windows CRLF checkout, and wrote its shell marker into the source tree.
The test now uses a temp directory as cwd, reads a small file from it
and checks the shell marker lands there.
The vision tests download their images from huggingface.co, which
resolves to a private address on the runners inside the HF network.
The media URL check refuses it with "host is not a public address"
and the server-cuda job fails.
The vision fixture enables media_allow_private_urls. The private
address check keeps its own coverage in
test_vision_image_url_private_address.
The same-origin tests over the LAN address block forever on the Metal
runners: the request to the host's own LAN address never returns and
has no timeout, so the job runs until the 6h limit.
get_lan_ip() now probes /health on that address with a short timeout
and returns None when the host cannot reach it, which skips the test.
The test requests themselves use the default request timeout so a
hang fails the test instead of stalling the job.