Georgi Gerganov
ba42794c9e
graph : fix equal_seq() check ( #14986 )
...
ggml-ci
2025-08-01 06:38:12 +03:00
Daniel Bevenius
ca0ef2dddb
llama : clarify comment about pp and tg graphs [no ci] ( #14895 )
...
* llama : clarify comment about pp and tg graphs [no ci]
This commit clarifies the comment in `llama-context.cpp` regarding the
prefill prompt (pp), and token generation (tg) graphs.
The motivation for this is that I've struggled to remember these and had
to look them up more than once, so I thought it would be helpful to add
a comment that makes it clear what these stand for.
* squash! llama : clarify comment about pp and tg graphs [no ci]
Change "pp" to "prompt processing".
2025-07-27 12:10:51 +02:00
Georgi Gerganov
c1dbea752a
context : restore preemptive sched reset when LLAMA_SET_ROWS=0 ( #14870 )
...
ggml-ci
2025-07-25 14:28:06 +03:00
Georgi Gerganov
e4868d16d2
context : perform output reorder lazily upon access after sync ( #14853 )
...
* context : perform output reorder after lazily upon access after sync
ggml-ci
* cont : add TODO
2025-07-24 16:31:48 +03:00
Georgi Gerganov
d498af3d5a
graph : avoid huge warm-up graphs for MoE models ( #14753 )
...
* graph : avoid huge warm-up graphs for MoE models
ggml-ci
* cont : bump max nodes to 8x model tensors
2025-07-18 14:31:15 +03:00
Georgi Gerganov
8f974bc1e9
graph : refactor context to not pass gf explicitly ( #14629 )
...
ggml-ci
2025-07-18 08:29:28 +03:00
Georgi Gerganov
01612b7409
llama : reuse compute graphs ( #14482 )
...
* llama : reuse compute graphs
ggml-ci
* llama-bench : add graph reuse parameter
ggml-ci
* cont : remove the parameter and the sched resets
ggml-ci
* graph : rename update() to can_reuse()
ggml-ci
* params : remove is_same()
ggml-ci
* graph : set res->params in llm_graph_context constructor
ggml-ci
* graph : avoid set_max_nodes in llm_graph_result
ggml-ci
* kv-cache : reuse llama_context's graph result instance
ggml-ci
* context : reset the previous graph result upon memory updates
ggml-ci
* batch : llama_ubatch now carries its data instead of pointing to balloc
ggml-ci
* merge : fix build
ggml-ci
* graph : fix can_reuse() checks when flash-attention is disabled
* graph : move llm_graph_result impl in source file + debug env
ggml-ci
2025-07-17 19:08:33 +03:00
Georgi Gerganov
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 >
2025-07-16 16:35:42 +03:00
Aman Gupta
9c9e4fc635
llama-context: add ability to get logits ( #14672 )
2025-07-14 21:01:41 +08:00
Georgi Gerganov
7b50d589a8
kv-cells : fix tracking of seq_pos ( #14339 )
...
* kv-cells : fix tracking of seq_pos during cache reuse
ggml-ci
* cont : improve error message
ggml-ci
* cont : add more comments
2025-06-23 12:27:35 +03:00
Georgi Gerganov
692e3cdd0a
memory : rename interface to llama_memory_context_i ( #14296 )
...
* memory : rename interface to llama_memory_context_i
ggml-ci
* cont : fix comments
* cont : use "mctx" for referencing a memory context
ggml-ci
2025-06-21 08:03:46 +03:00
Georgi Gerganov
4c9fdfbe15
ubatch : new splitting logic ( #14217 )
...
ggml-ci
2025-06-20 10:14:14 +03:00
Georgi Gerganov
d3e64b9f49
llama : rework embeddings logic ( #14208 )
...
* llama : rework embeddings logic
ggml-ci
* cont : fix rerank
ggml-ci
* cont : engrish [no ci]
* cont : fix rerank
ggml-ci
* server : support both embeddings and completions with single model
ggml-ci
* cont : avoid embeddings_org
ggml-ci
2025-06-16 14:14:00 +03:00
Georgi Gerganov
c311ac664d
cparams : rename LLAMA_MAX_PARALLEL_SEQUENCES to LLAMA_MAX_SEQ ( #14188 )
...
ggml-ci
2025-06-15 10:08:58 +03:00
Georgi Gerganov
b9912ac570
batch : auto-gen positions + verify multi-sequence input ( #14177 )
...
* batch : verify multi-sequence input batches
ggml-ci
* cont : auto-gen positions + verify multi-seq input
ggml-ci
* cont : first print debug info, then perform validation
ggml-ci
* cont : fix position auto-gen + add comments
ggml-ci
2025-06-15 09:18:37 +03:00
Georgi Gerganov
60c666347b
batch : rework llama_batch_allocr ( #14153 )
...
* batch : rework llama_batch_allocr
ggml-ci
* cont : move validation inside class
ggml-ci
* cont : move output counting to class
ggml-ci
* cont : minor
ggml-ci
* batch : add TODOs
ggml-ci
2025-06-13 13:47:55 +03:00
Georgi Gerganov
f6e1a7aa87
context : simplify output counting logic during decode ( #14142 )
...
* batch : remove logits_all flag
ggml-ci
* context : simplify output counting logic during decode
ggml-ci
* cont : fix comments
2025-06-12 11:50:01 +03:00
Georgi Gerganov
c3ee46fab4
batch : remove logits_all flag ( #14141 )
...
ggml-ci
2025-06-12 11:49:26 +03:00
Georgi Gerganov
9596506965
kv-cache : fix split_equal handling in unified implementation ( #14130 )
...
ggml-ci
2025-06-12 10:02:15 +03:00
compilade
a20b2b05bc
context : round n_tokens to next multiple of n_seqs when reserving ( #14140 )
...
This fixes RWKV inference which otherwise failed
when the worst case ubatch.n_seq_tokens rounded to 0.
2025-06-12 02:56:04 -04:00
Georgi Gerganov
745aa5319b
llama : deprecate llama_kv_self_ API ( #14030 )
...
* llama : deprecate llama_kv_self_ API
ggml-ci
* llama : allow llama_memory_(nullptr)
ggml-ci
* memory : add flag for optional data clear in llama_memory_clear
ggml-ci
2025-06-06 14:11:15 +03:00
Georgi Gerganov
487a5e0401
context : fix SWA-related warning for multiple sequences ( #14045 )
2025-06-06 13:29:18 +03:00
Sigbjørn Skjæret
d17a809ef0
llama : support multiple classifier outputs and labels ( #13940 )
2025-06-06 09:03:25 +02:00
Georgi Gerganov
7f37b6cf1e
memory : migrate from llama_kv_cache to more generic llama_memory ( #14006 )
...
* memory : merge llama_kv_cache into llama_memory + new `llama_memory` API
ggml-ci
* context : fix casts
ggml-ci
2025-06-05 15:29:22 +03:00
Georgi Gerganov
9e31bec4fd
context : fix pos_min initialization upon error decode ( #14008 )
...
ggml-ci
2025-06-05 09:06:29 +03:00
Georgi Gerganov
3e63a58ef7
kv-cache : refactor the update/defrag mechanism ( #13988 )
...
* kv-cache : refactor update mechanism
ggml-ci
* memory : improve status handling
* defrag : reset head + add comments
ggml-ci
* cont : minor fixes
ggml-ci
2025-06-04 18:58:20 +03:00
Georgi Gerganov
803f8baf4f
llama : deprecate explicit kv_self defrag/update calls ( #13921 )
...
ggml-ci
2025-05-31 15:58:33 +03:00
Georgi Gerganov
3600cc2886
llama : use n_swa + n_ubatch cells for SWA cache ( #13833 )
...
* llama : use n_swa + n_ubatch cells for SWA cache
ggml-ci
* llama : add warning about multi-sqeuence SWA contexts
2025-05-31 15:57:44 +03:00
Georgi Gerganov
3f55f781f1
llama : auto-batch preparation ( #13845 )
...
* llama : auto-batch
ggml-ci
* context : simplify if branching
2025-05-31 12:55:57 +03:00
Georgi Gerganov
12d0188c0d
kv-cache : refactor + add llama_memory_state_i ( #13746 )
...
* kv-cache : simplify the "struct llama_kv_cache" interface
ggml-ci
* kv-cache : revert the (n_swa + n_ubatch) change (for next PR)
ggml-ci
* kv-cache : some comments
ggml-ci
* context : fix graph reserve for multiple sequences
ggml-ci
* kv-cache : fix typo [no ci]
* kv-cache : fix find_slot() logic for free slots
ggml-ci
* llama : add TODO for deprecating the defrag API in the future
* kv-cache : improve find_slot() using min/max seq pos info
ggml-ci
* llama : handle aborts and compute errors
ggml-ci
* memory : extract state into llama_memory_state
ggml-ci
* kv-cache : add comments
ggml-ci
* server : update batching logic to reset n_batch on successful decode
* server : upon full re-processing, remove the sequence from the cache
* kv-cache : add TODO for doing split_equal when split_simple fails
ggml-ci
2025-05-31 10:24:04 +03:00
Georgi Gerganov
4f81b33e32
llama : validate seq id batch input ( #13809 )
...
* llama : validate seq id batch input
ggml-ci
* cont : fix the fix
ggml-ci
2025-05-27 09:40:59 +03:00
Georgi Gerganov
79c137f776
examples : allow extracting embeddings from decoder contexts ( #13797 )
...
ggml-ci
2025-05-26 14:03:54 +03:00
Georgi Gerganov
de2ef53a4b
kv-cache : rework kv_cell ( #13706 )
...
* kv-cache : rework kv_cell
ggml-ci
* kv-cells : use "shift" instead of "delta" consistently
ggml-ci
* llama : add llama_max_parallel_sequences()
ggml-ci
* kv-cells : update comments [no ci]
* context : fail upon construction if sequences exceed max value
ggml-ci
* kv-cells : get_pos() -> pos_get() + comments
ggml-ci
* kv-cells : fix tracking of "used" cells
ggml-ci
2025-05-25 16:34:36 +03:00
Georgi Gerganov
797f2ac062
kv-cache : simplify the interface ( #13660 )
...
* kv-cache : simplify the interface
ggml-ci
* context : revert llama_batch_allocr position change
ggml-ci
2025-05-21 15:11:13 +03:00
Georgi Gerganov
a4090d1174
llama : remove llama_kv_cache_view API + remove deprecated ( #13653 )
...
ggml-ci
2025-05-20 16:13:16 +03:00
Georgi Gerganov
e298d2fbd0
kv-cache : add SWA support ( #13194 )
...
* kv-cache : prepare for SWA
ggml-ci
* kv-cache : initial iSWA implementation
ggml-ci
* kv-cache : rework error recovery logic
ggml-ci
* models : fix Phi-3 SWA parameters
ggml-ci
* model : adjust Granite to rope factor changes
ggml-ci
* server : check if context can do shifts
ggml-ci
* iswa : for now, always enable shifts (experiment)
ggml-ci
* kv-cache : simplify SWA logic
ggml-ci
* kv-cache : apply defrag when we fail to find slots for the batch
ggml-ci
* llama : update docs about llama_decode
ggml-ci
* kv-cache : update warning logs when no space for the batch is available
ggml-ci
* llama : add llama_kv_self_seq_pos_min()
* kv-cache : keep track of partial SWA computes and print warnings
* server : disallow use cases involving partial SWA context
ggml-ci
* llama : add param to control SWA cache size
ggml-ci
* minor : clean-up
ggml-ci
2025-05-20 08:05:46 +03:00
Sigbjørn Skjæret
f5170c1d7a
editorconfig : fix trailing whitespace from #13542 ( #13546 )
2025-05-14 21:22:49 +03:00
Gilad S.
017f10b5fa
fix: crash when calling llama_state_get_size on a context without a KV cache ( #13542 )
2025-05-14 19:18:18 +03:00
Johannes Gäßler
10d2af0eaa
llama/ggml: add LLM training support ( #10544 )
...
* llama/ggml: add LLM training support
more compact progress bar
llama_save_model_to_file
llama_opt_param_filter
ggml_graph_dup force_grads
refactor ggml_opt, fix test-opt
* remove logits_all
* refactor CUDA implementation for ACC
* reset graph at beginning of opt period
2025-05-12 14:44:49 +02:00
Georgi Gerganov
064cc596ac
context : fix state io for memory-less contexts ( #13470 )
...
ggml-ci
2025-05-12 15:12:27 +03:00
David Huang
7f323a589f
Add --no-op-offload to improve -ot pp perf in MoE models like llama4 400B ( #13386 )
2025-05-11 14:18:39 +02:00
Georgi Gerganov
6562e5a4d6
context : allow cache-less context for embeddings ( #13108 )
...
* context : allow cache-less context for embeddings
ggml-ci
* context : enable reranking with encode()
ggml-ci
* context : encode() clears embd_seq
ggml-ci
* examples : use llama_encode() when appropriate
ggml-ci
* models : nomic bert moe does not require KV cache
* llama : update comments for llama_decode/llama_encode
ggml-ci
* context : update warning log [no ci]
2025-05-08 14:28:33 +03:00
Georgi Gerganov
51fb96b1ff
context : remove logits_all flag ( #13284 )
...
* context : remove logits_all flag
ggml-ci
* llama : remove logits_all flag + reorder llama_context_params
ggml-ci
2025-05-08 14:26:50 +03:00
Georgi Gerganov
a75cb30dc9
context : fix reorder logic ( #13267 )
...
ggml-ci
2025-05-02 20:54:13 +03:00
Georgi Gerganov
c642bc014c
kv-cache : separate recurrent vs non-recurrent impl ( #12799 )
...
* kv-cache : serparate recurrent vs non-recurrent impl (wip)
ggml-ci
* kv-cache : init -> contructor + add llama_memory_params
ggml-ci
* kv-cache : fix callback reference
ggml-ci
* context : llama_kv_cache -> llama_memory_i
ggml-ci
* context : move memory creation logic to model
ggml-ci
* llama : remove reference of memory during encode
ggml-ci
* kv-cache : hide padding details in the implementation
ggml-ci
* kv-cache : add ubatch_next()
ggml-ci
* context : simplify sbatch logic
ggml-ci
* kv-cache : hide defrag logic in the implementation
ggml-ci
* context : hide kv cache details in implementation
ggml-ci
* build : fix
ggml-ci
* cont : another fix
ggml-ci
* kv-cache : simplify interface (wip)
ggml-ci
* kv-cache : use separate KV cell structs for unified/recurrent
ggml-ci
* kv-cache : clean-up
ggml-ci
* model : better llama_model::create_model() signature
ggml-ci
* kv-cache : fix recurrent seq_rm()
ggml-ci
* kv-cache : replace `struct callbacks` with `llama_model &`
ggml-ci
* kv-cache : replace `struct graph_params` with `llama_context &`
ggml-ci
* kv-cache : fix offload check
ggml-ci
* context : avoid passing unique_ptr
ggml-ci
* kv-cache : avoid using the backends from the llama_context
ref #13113
ggml-ci
* kv-cache : more consistent debug logs [no ci]
* kv-cache : do not pass the full llama_context for kv graphs
ggml-ci
* kv-cache : remove comment
* kv-cache : ggml_rope_ext_inplace -> ggml_rope_ext
ggml-ci
* kv-cache : fix recurrent multi-user case
ggml-ci
* memory : remove comments [no ci]
2025-05-02 17:48:36 +03:00
ddh0
16a457facd
fix typo: n_ctx_pre_seq -> n_ctx_per_seq ( #13221 )
2025-04-30 21:28:43 +01:00
pockers21
fb0471d175
context : do not clear output buffer on reserve ( #13152 )
...
Co-authored-by: pockers21 <liyang2@uniontech.com >
2025-04-28 16:45:40 +03:00
Diego Devesa
295354ea68
llama : fix K-shift with quantized K and BLAS backend ( #13113 )
2025-04-25 19:40:11 +02:00
Georgi Gerganov
2f74c354c0
graph : make FA compatible with MLA + add initial Metal kernels ( #12953 )
...
* graph : make mla compatible with FA
* metal : add exp FA kernels for DeepSeek models
ggml-ci
* llama : minor naming updates
ggml-ci
* ggml : disable FA for DS head sizes
* tests : add FA tests for MLA shapes
ggml-ci
2025-04-17 18:16:36 +03:00
Juk Armstrong
daa422881a
llama : DeepSeek V2/V3 MLA implementation ( #12801 )
...
* Merged using squash to remove all noise commit messages
* Force flash attention off for `LLM_ARCH_DEEPSEEK2` - embedding too large
* Removed 3 conts (2x RoPE and 1x RMS-norm)
* Changed to use `<cmath>` instead of `<math.h>`
* Reverted removal of the 3 conts
* Used `reshape` in `llm_graph_context::build_attn_mha()`
* Use `k_pe = ggml_reshape`
* Removed the 3 conts again
* Removed the 3D views of `wk_b` and `wv_b`, and just save and 3D in GGUF
* Removed MQA optimisation from `build_attn_mha()` as no gains now
* Simplified `is_mla` branch in `llm_build_deepseek2()`
* Removed `build_attn_mla` and added `nullptr` to all `build_atnn` calls
* Fixed call to `build_attn` in `llm_build_t5_enc`
2025-04-15 09:49:57 +03:00
