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

..

batched-bench

llama : add high-throughput mode (#14363)

2025-07-16 16:35:42 +03:00

cvector-generator

llama : deprecate llama_kv_self_ API (#14030)

2025-06-06 14:11:15 +03:00

export-lora

llama : move end-user examples to tools directory (#13249)

2025-05-02 20:27:13 +02:00

gguf-split

scripts : make the shell scripts cross-platform (#14341)

2025-06-30 10:17:18 +02:00

imatrix

llama : deprecate llama_kv_self_ API (#14030)

2025-06-06 14:11:15 +03:00

llama-bench

llama-bench : add --no-warmup flag (#14224) (#14270)

2025-06-19 12:24:12 +02:00

main

main : honor --verbose-prompt on interactive prompts (#14350)

2025-06-24 09:31:00 +02:00

mtmd

ggml : implement GEGLU_ERF and GEGLU_QUICK ops (#14445)

2025-07-03 23:07:22 +02:00

perplexity

llama : deprecate llama_kv_self_ API (#14030)

2025-06-06 14:11:15 +03:00

quantize

scripts : make the shell scripts cross-platform (#14341)

2025-06-30 10:17:18 +02:00

rpc

rpc : Fix build on OpenBSD (#13541)

2025-05-25 15:35:53 +03:00

run

cmake : do not search for curl libraries by ourselves (#14613)

2025-07-10 15:29:05 +03:00

server

server : pre-calculate EOG logit biases (#14721)

2025-07-16 14:04:12 +03:00

tokenize

llama : move end-user examples to tools directory (#13249)

2025-05-02 20:27:13 +02:00

tts

sync : vendor (#13901)

2025-05-30 16:25:45 +03:00

CMakeLists.txt

mtmd : rename llava directory to mtmd (#13311)

2025-05-05 16:02:55 +02:00