Skip to content

sync : ggml #3319

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 50 commits into from
Jul 12, 2025
Merged

sync : ggml #3319

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
50 commits
Select commit Hold shift + click to select a range
dbb85a4
ggml : add version function to get lib version (ggml/1286)
danbev Jul 2, 2025
9c20ec6
vulkan : implement ggml_roll (ggml/1290)
Acly Jul 3, 2025
ab813e8
vulkan : implement bilinear interpolation for ggml_upscale/ggml_inter…
Acly Jul 3, 2025
266de70
add GELU_ERF (llama/14455)
CISC Jul 1, 2025
c1f229a
vulkan: Split large mul_mat_id to fit in shared memory (llama/14451)
jeffbolznv Jul 1, 2025
f2609ab
CANN: update aclnnGroupedMatmulV2 to aclnnGroupedMatmulV3 (llama/14411)
noemotiovon Jul 1, 2025
4234492
ci : disable fast-math for Metal GHA CI (llama/14478)
ggerganov Jul 1, 2025
98163d4
ggml : Callback before abort (llama/14481)
ScaledLizard Jul 2, 2025
7645aee
opencl : update upscale to support align corners (llama/14488)
lhez Jul 2, 2025
801367c
opencl : skip empty nodes on cgraph compute (llama/14491)
EZForever Jul 2, 2025
50785ef
opencl : fix possible buffer overflow in dump_tensor (llama/14490)
jeffzhou2000 Jul 2, 2025
66e89a0
ggml : support bcast ggml_soft_max_ext, ggml_flash_attn_ext (llama/14…
ggerganov Jul 12, 2025
5c6866f
vulkan: support softmax/FA batch and broadcast (llama/14449)
jeffbolznv Jul 1, 2025
73ea436
CUDA: broadcasting for FlashAttention mask (llama/14500)
JohannesGaessler Jul 2, 2025
7366a28
CUDA: add softmax broadcast (llama/14475)
am17an Jul 2, 2025
ed13f9d
llama : initial Mamba-2 support (llama/9126)
compilade Jul 2, 2025
580e2ed
CUDA: add dynamic shared mem to softmax, refactor general usage (llam…
am17an Jul 2, 2025
ca2b806
ggml : fix FA mask dim 2 and 3 (llama/14505)
ggerganov Jul 3, 2025
ca9c4d2
kv-cache : use ggml_set_rows (llama/14285)
ggerganov Jul 3, 2025
417be2a
Fix conditional enabling following arch checks for ggml-sycl (llama/1…
s-Nick Jul 3, 2025
dfa89e7
ggml: backward pass for split swiglu (llama/14483)
JohannesGaessler Jul 3, 2025
e9acb01
vulkan: support mixed/deepseekR1 FA head sizes (llama/14509)
jeffbolznv Jul 3, 2025
dea036f
opencl : broadcast for soft_max (llama/14510)
lhez Jul 3, 2025
8fde826
ggml : implement GEGLU_ERF and GEGLU_QUICK ops (llama/14445)
CISC Jul 3, 2025
fda1c64
CANN: Replace aclrtMemsetSync with aclnnInplaceZero operator (llama/1…
luyhcsu Jul 4, 2025
c1ee7c4
metal : disable fast math in all quantize kernels (llama/14528)
ggerganov Jul 4, 2025
5b2b458
opencl: add GELU_ERF (llama/14476)
CISC Jul 5, 2025
3c5e104
vulkan: Handle updated FA dim2/3 definition (llama/14518)
jeffbolznv Jul 5, 2025
0a8d97e
vulkan: fix rms_norm+mul fusion (llama/14545)
jeffbolznv Jul 6, 2025
71bcb61
vulkan: increase LOAD_VEC_A to 8 (IQ1/IQ2) or 4 (IQ3) (llama/14485)
netrunnereve Jul 6, 2025
602d77e
CUDA: add bf16 and i32 to getrows (llama/14529)
am17an Jul 7, 2025
f1bab24
musa: fix build warnings (unused variable) (llama/14561)
yeahdongcn Jul 7, 2025
2324247
CUDA: add bilinear interpolation for upscale (llama/14563)
am17an Jul 8, 2025
5f4d05b
cuda : fix rope with partial rotation and non-cont src (llama/14580)
ggerganov Jul 8, 2025
d5684df
vulkan : fix rope with partial rotation and non-cont src (llama/14582)
jeffbolznv Jul 8, 2025
c896a4f
vulkan: optimize flash attention split_k_reduce (llama/14554)
jeffbolznv Jul 8, 2025
6cfb733
ggml : prevent integer overflow in gguf tensor size calculation (llam…
Yuuoniy Jul 9, 2025
adfb259
ggml : add ggml_scale_bias (llama/14417)
ngxson Jul 9, 2025
c161799
cuda : support Falcon-H1 state size for SSM_SCAN (llama/14602)
compilade Jul 10, 2025
20c8613
SYCL: Initial set_rows kernel implementation (llama/14562)
qnixsynapse Jul 10, 2025
f5449b1
opencl: add `set_rows` for `f16` and `f32` (llama/14547)
lhez Jul 10, 2025
445bd88
opencl: add tiled mul_mat_f16_f32 (llama/14535)
rmatif Jul 10, 2025
56b0b32
HIP : Add HIP 7.0+ compatibility for hipBLAS compute types (llama/14634)
slojosic-amd Jul 11, 2025
948eb50
model : support LiquidAI LFM2 hybrid family (llama/14620)
tdakhran Jul 11, 2025
a267512
vulkan: optimizations for deepseek prompt processing (llama/14555)
jeffbolznv Jul 12, 2025
06669e7
vulkan: support SET_ROWS (llama/14587)
jeffbolznv Jul 12, 2025
d79d3c2
sync : resolve conflicts (ggml/0)
ggerganov Jul 12, 2025
4bd83ef
sync : ggml
ggerganov Jul 12, 2025
ee9f540
talk-llama : sync llama.cpp
ggerganov Jul 12, 2025
98509f6
sync : resolve conflicts (#0)
ggerganov Jul 12, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
195 changes: 192 additions & 3 deletions examples/talk-llama/llama-arch.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -45,6 +45,9 @@ static const std::map<llm_arch, const char *> LLM_ARCH_NAMES = {
{ LLM_ARCH_GEMMA3N, "gemma3n" },
{ LLM_ARCH_STARCODER2, "starcoder2" },
{ LLM_ARCH_MAMBA, "mamba" },
{ LLM_ARCH_MAMBA2, "mamba2" },
{ LLM_ARCH_JAMBA, "jamba" },
{ LLM_ARCH_FALCON_H1, "falcon-h1" },
{ LLM_ARCH_XVERSE, "xverse" },
{ LLM_ARCH_COMMAND_R, "command-r" },
{ LLM_ARCH_COHERE2, "cohere2" },
Expand All @@ -70,13 +73,17 @@ static const std::map<llm_arch, const char *> LLM_ARCH_NAMES = {
{ LLM_ARCH_ARWKV7, "arwkv7" },
{ LLM_ARCH_GRANITE, "granite" },
{ LLM_ARCH_GRANITE_MOE, "granitemoe" },
{ LLM_ARCH_GRANITE_HYBRID, "granitehybrid" },
{ LLM_ARCH_CHAMELEON, "chameleon" },
{ LLM_ARCH_WAVTOKENIZER_DEC, "wavtokenizer-dec" },
{ LLM_ARCH_PLM, "plm" },
{ LLM_ARCH_BAILINGMOE, "bailingmoe" },
{ LLM_ARCH_DOTS1, "dots1" },
{ LLM_ARCH_ARCEE, "arcee" },
{ LLM_ARCH_ERNIE4_5, "ernie4_5" },
{ LLM_ARCH_HUNYUAN_MOE, "hunyuan-moe" },
{ LLM_ARCH_SMOLLM3, "smollm3" },
{ LLM_ARCH_LFM2, "lfm2" },
{ LLM_ARCH_UNKNOWN, "(unknown)" },
};

Expand Down Expand Up @@ -149,7 +156,6 @@ static const std::map<llm_kv, const char *> LLM_KV_NAMES = {
{ LLM_KV_ATTENTION_SCALE, "%s.attention.scale" },
{ LLM_KV_ATTENTION_KEY_LENGTH_MLA, "%s.attention.key_length_mla" },
{ LLM_KV_ATTENTION_VALUE_LENGTH_MLA, "%s.attention.value_length_mla" },
{ LLM_KV_ATTENTION_LAYER_INDICES, "%s.attention.layer_indices" },

{ LLM_KV_ROPE_DIMENSION_COUNT, "%s.rope.dimension_count" },
{ LLM_KV_ROPE_DIMENSION_SECTIONS, "%s.rope.dimension_sections" },
Expand All @@ -170,6 +176,7 @@ static const std::map<llm_kv, const char *> LLM_KV_NAMES = {
{ LLM_KV_SSM_INNER_SIZE, "%s.ssm.inner_size" },
{ LLM_KV_SSM_STATE_SIZE, "%s.ssm.state_size" },
{ LLM_KV_SSM_TIME_STEP_RANK, "%s.ssm.time_step_rank" },
{ LLM_KV_SSM_GROUP_COUNT, "%s.ssm.group_count" },
{ LLM_KV_SSM_DT_B_C_RMS, "%s.ssm.dt_b_c_rms" },

{ LLM_KV_WKV_HEAD_SIZE, "%s.wkv.head_size" },
Expand All @@ -182,6 +189,8 @@ static const std::map<llm_kv, const char *> LLM_KV_NAMES = {

{ LLM_KV_CLASSIFIER_OUTPUT_LABELS, "%s.classifier.output_labels" },

{ LLM_KV_SHORTCONV_L_CACHE, "%s.shortconv.l_cache" },

{ LLM_KV_TOKENIZER_MODEL, "tokenizer.ggml.model" },
{ LLM_KV_TOKENIZER_PRE, "tokenizer.ggml.pre" },
{ LLM_KV_TOKENIZER_LIST, "tokenizer.ggml.tokens" },
Expand Down Expand Up @@ -1004,6 +1013,77 @@ static const std::map<llm_arch, std::map<llm_tensor, const char *>> LLM_TENSOR_N
{ LLM_TENSOR_SSM_OUT, "blk.%d.ssm_out" },
},
},
{
LLM_ARCH_MAMBA2,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_OUTPUT, "output" },
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_SSM_IN, "blk.%d.ssm_in" },
{ LLM_TENSOR_SSM_CONV1D, "blk.%d.ssm_conv1d" },
{ LLM_TENSOR_SSM_DT, "blk.%d.ssm_dt" },
{ LLM_TENSOR_SSM_A, "blk.%d.ssm_a" },
{ LLM_TENSOR_SSM_D, "blk.%d.ssm_d" },
{ LLM_TENSOR_SSM_NORM, "blk.%d.ssm_norm" },
{ LLM_TENSOR_SSM_OUT, "blk.%d.ssm_out" },
},
},
{
LLM_ARCH_JAMBA,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_OUTPUT, "output" },
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_SSM_IN, "blk.%d.ssm_in" },
{ LLM_TENSOR_SSM_CONV1D, "blk.%d.ssm_conv1d" },
{ LLM_TENSOR_SSM_X, "blk.%d.ssm_x" },
{ LLM_TENSOR_SSM_DT, "blk.%d.ssm_dt" },
{ LLM_TENSOR_SSM_DT_NORM, "blk.%d.ssm_dt_norm" },
{ LLM_TENSOR_SSM_A, "blk.%d.ssm_a" },
{ LLM_TENSOR_SSM_B_NORM, "blk.%d.ssm_b_norm" },
{ LLM_TENSOR_SSM_C_NORM, "blk.%d.ssm_c_norm" },
{ LLM_TENSOR_SSM_D, "blk.%d.ssm_d" },
{ LLM_TENSOR_SSM_OUT, "blk.%d.ssm_out" },
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" },
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
{ LLM_TENSOR_FFN_GATE_EXPS, "blk.%d.ffn_gate_exps" },
{ LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" },
{ LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" },
},
},
{
LLM_ARCH_FALCON_H1,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_OUTPUT, "output" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_SSM_IN, "blk.%d.ssm_in" },
{ LLM_TENSOR_SSM_CONV1D, "blk.%d.ssm_conv1d" },
{ LLM_TENSOR_SSM_DT, "blk.%d.ssm_dt" },
{ LLM_TENSOR_SSM_A, "blk.%d.ssm_a" },
{ LLM_TENSOR_SSM_D, "blk.%d.ssm_d" },
{ LLM_TENSOR_SSM_NORM, "blk.%d.ssm_norm" },
{ LLM_TENSOR_SSM_OUT, "blk.%d.ssm_out" },
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
},
},
{
LLM_ARCH_XVERSE,
{
Expand Down Expand Up @@ -1564,6 +1644,43 @@ static const std::map<llm_arch, std::map<llm_tensor, const char *>> LLM_TENSOR_N
{ LLM_TENSOR_FFN_UP_SHEXP, "blk.%d.ffn_up_shexp" },
},
},
{
LLM_ARCH_GRANITE_HYBRID,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_OUTPUT, "output" },
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
// mamba(2) ssm layers
{ LLM_TENSOR_SSM_IN, "blk.%d.ssm_in" },
{ LLM_TENSOR_SSM_CONV1D, "blk.%d.ssm_conv1d" },
{ LLM_TENSOR_SSM_DT, "blk.%d.ssm_dt" },
{ LLM_TENSOR_SSM_A, "blk.%d.ssm_a" },
{ LLM_TENSOR_SSM_D, "blk.%d.ssm_d" },
{ LLM_TENSOR_SSM_NORM, "blk.%d.ssm_norm" },
{ LLM_TENSOR_SSM_OUT, "blk.%d.ssm_out" },
// attention layers
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
// dense FFN
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
// moe FFN
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
{ LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" },
{ LLM_TENSOR_FFN_GATE_EXPS, "blk.%d.ffn_gate_exps" },
{ LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" },
{ LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" },
// shared expert
{ LLM_TENSOR_FFN_GATE_SHEXP, "blk.%d.ffn_gate_shexp" },
{ LLM_TENSOR_FFN_DOWN_SHEXP, "blk.%d.ffn_down_shexp" },
{ LLM_TENSOR_FFN_UP_SHEXP, "blk.%d.ffn_up_shexp" },
},
},
{
LLM_ARCH_CHAMELEON,
{
Expand Down Expand Up @@ -1676,6 +1793,67 @@ static const std::map<llm_arch, std::map<llm_tensor, const char *>> LLM_TENSOR_N
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
},
},
{
LLM_ARCH_HUNYUAN_MOE,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_OUTPUT, "output" },
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
{ LLM_TENSOR_ATTN_Q_NORM, "blk.%d.attn_q_norm" },
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
{ LLM_TENSOR_ATTN_K_NORM, "blk.%d.attn_k_norm" },
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" },
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
{ LLM_TENSOR_FFN_GATE_SHEXP, "blk.%d.ffn_gate_shexp" },
{ LLM_TENSOR_FFN_DOWN_SHEXP, "blk.%d.ffn_down_shexp" },
{ LLM_TENSOR_FFN_UP_SHEXP, "blk.%d.ffn_up_shexp" },
{ LLM_TENSOR_FFN_GATE_EXPS, "blk.%d.ffn_gate_exps" },
{ LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" },
{ LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" },
},
},
{
LLM_ARCH_SMOLLM3,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_OUTPUT, "output" },
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
},
},
{
LLM_ARCH_LFM2,
{
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_ATTN_K_NORM, "blk.%d.attn_k_norm" },
{ LLM_TENSOR_ATTN_Q_NORM, "blk.%d.attn_q_norm" },
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
{ LLM_TENSOR_SHORTCONV_CONV, "blk.%d.shortconv.conv" },
{ LLM_TENSOR_SHORTCONV_INPROJ, "blk.%d.shortconv.in_proj" },
{ LLM_TENSOR_SHORTCONV_OUTPROJ, "blk.%d.shortconv.out_proj" },
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_TOKEN_EMBD_NORM, "token_embd_norm" },
}
},
{
LLM_ARCH_UNKNOWN,
{
Expand Down Expand Up @@ -1760,7 +1938,11 @@ static const std::map<llm_tensor, llm_tensor_info> LLM_TENSOR_INFOS = {
{LLM_TENSOR_FFN_ACT, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_DIV}},
{LLM_TENSOR_SSM_CONV1D, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_SSM_CONV}},
{LLM_TENSOR_SSM_A, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_SSM_SCAN}},
{LLM_TENSOR_SSM_DT_NORM, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
{LLM_TENSOR_SSM_B_NORM, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
{LLM_TENSOR_SSM_C_NORM, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
{LLM_TENSOR_SSM_D, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
{LLM_TENSOR_SSM_NORM, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
{LLM_TENSOR_TIME_MIX_LERP_X, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
{LLM_TENSOR_TIME_MIX_LN, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
{LLM_TENSOR_CHANNEL_MIX_LERP_K, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
Expand Down Expand Up @@ -1839,6 +2021,9 @@ static const std::map<llm_tensor, llm_tensor_info> LLM_TENSOR_INFOS = {
{LLM_TENSOR_CONVNEXT_PW1, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
{LLM_TENSOR_CONVNEXT_PW2, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
{LLM_TENSOR_CONVNEXT_GAMMA, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
{LLM_TENSOR_SHORTCONV_CONV, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_SSM_CONV}},
{LLM_TENSOR_SHORTCONV_INPROJ, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
{LLM_TENSOR_SHORTCONV_OUTPROJ, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
};

LLM_KV::LLM_KV(llm_arch arch, const char * suffix) : arch(arch), suffix(suffix) {}
Expand Down Expand Up @@ -1894,6 +2079,7 @@ const llm_tensor_info & llm_tensor_info_for(llm_tensor tensor) {
bool llm_arch_is_recurrent(const llm_arch & arch) {
switch (arch) {
case LLM_ARCH_MAMBA:
case LLM_ARCH_MAMBA2:
case LLM_ARCH_RWKV6:
case LLM_ARCH_RWKV6QWEN2:
case LLM_ARCH_RWKV7:
Expand All @@ -1905,9 +2091,12 @@ bool llm_arch_is_recurrent(const llm_arch & arch) {
}

bool llm_arch_is_hybrid(const llm_arch & arch) {
// TODO: There are currently no hybrid models! Once there are, this will be
// the place to identify them
switch (arch) {
case LLM_ARCH_JAMBA:
case LLM_ARCH_FALCON_H1:
case LLM_ARCH_GRANITE_HYBRID:
case LLM_ARCH_LFM2:
return true;
default:
return false;
}
Expand Down
18 changes: 17 additions & 1 deletion examples/talk-llama/llama-arch.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -49,6 +49,9 @@ enum llm_arch {
LLM_ARCH_GEMMA3N,
LLM_ARCH_STARCODER2,
LLM_ARCH_MAMBA,
LLM_ARCH_MAMBA2,
LLM_ARCH_JAMBA,
LLM_ARCH_FALCON_H1,
LLM_ARCH_XVERSE,
LLM_ARCH_COMMAND_R,
LLM_ARCH_COHERE2,
Expand All @@ -74,13 +77,17 @@ enum llm_arch {
LLM_ARCH_ARWKV7,
LLM_ARCH_GRANITE,
LLM_ARCH_GRANITE_MOE,
LLM_ARCH_GRANITE_HYBRID,
LLM_ARCH_CHAMELEON,
LLM_ARCH_WAVTOKENIZER_DEC,
LLM_ARCH_PLM,
LLM_ARCH_BAILINGMOE,
LLM_ARCH_DOTS1,
LLM_ARCH_ARCEE,
LLM_ARCH_ERNIE4_5,
LLM_ARCH_HUNYUAN_MOE,
LLM_ARCH_SMOLLM3,
LLM_ARCH_LFM2,
LLM_ARCH_UNKNOWN,
};

Expand Down Expand Up @@ -153,7 +160,6 @@ enum llm_kv {
LLM_KV_ATTENTION_SCALE,
LLM_KV_ATTENTION_KEY_LENGTH_MLA,
LLM_KV_ATTENTION_VALUE_LENGTH_MLA,
LLM_KV_ATTENTION_LAYER_INDICES,

LLM_KV_ROPE_DIMENSION_COUNT,
LLM_KV_ROPE_DIMENSION_SECTIONS,
Expand All @@ -174,6 +180,7 @@ enum llm_kv {
LLM_KV_SSM_CONV_KERNEL,
LLM_KV_SSM_STATE_SIZE,
LLM_KV_SSM_TIME_STEP_RANK,
LLM_KV_SSM_GROUP_COUNT,
LLM_KV_SSM_DT_B_C_RMS,

LLM_KV_WKV_HEAD_SIZE,
Expand Down Expand Up @@ -221,6 +228,8 @@ enum llm_kv {

LLM_KV_CLASSIFIER_OUTPUT_LABELS,

LLM_KV_SHORTCONV_L_CACHE,

// deprecated:
LLM_KV_TOKENIZER_PREFIX_ID,
LLM_KV_TOKENIZER_SUFFIX_ID,
Expand Down Expand Up @@ -291,8 +300,12 @@ enum llm_tensor {
LLM_TENSOR_SSM_CONV1D,
LLM_TENSOR_SSM_X,
LLM_TENSOR_SSM_DT,
LLM_TENSOR_SSM_DT_NORM,
LLM_TENSOR_SSM_A,
LLM_TENSOR_SSM_B_NORM,
LLM_TENSOR_SSM_C_NORM,
LLM_TENSOR_SSM_D,
LLM_TENSOR_SSM_NORM,
LLM_TENSOR_SSM_OUT,
LLM_TENSOR_TIME_MIX_W0,
LLM_TENSOR_TIME_MIX_W1,
Expand Down Expand Up @@ -386,6 +399,9 @@ enum llm_tensor {
LLM_TENSOR_POS_NET_ATTN_K,
LLM_TENSOR_POS_NET_ATTN_V,
LLM_TENSOR_POS_NET_ATTN_OUT,
LLM_TENSOR_SHORTCONV_CONV,
LLM_TENSOR_SHORTCONV_INPROJ,
LLM_TENSOR_SHORTCONV_OUTPROJ,
};

enum llm_tensor_layer {
Expand Down
Loading
Loading