vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

fix-review

Browse source
This commit is contained in:
simonJJJ 2024-04-16 03:36:03 +08:00
parent edd8e2e5a0
commit 4256fe6a1a
3 changed files with 16 additions and 17 deletions
Download patch file Download diff file Expand all files Collapse all files

2
ggml-backend.c
View file

@ -1006,7 +1006,6 @@ static bool ggml_is_view_op(enum ggml_op op) {
// scheduler // scheduler
#ifndef GGML_SCHED_MAX_BACKENDS #ifndef GGML_SCHED_MAX_BACKENDS
#define GGML_SCHED_MAX_BACKENDS 16 #define GGML_SCHED_MAX_BACKENDS 16
#endif #endif
@ -1023,7 +1022,6 @@ static bool ggml_is_view_op(enum ggml_op op) {
#define GGML_SCHED_MAX_COPIES 4 #define GGML_SCHED_MAX_COPIES 4
#endif #endif
struct ggml_backend_sched_split { struct ggml_backend_sched_split {
int backend_id; int backend_id;
int i_start; int i_start;

2
ggml.h
View file

@ -228,7 +228,7 @@
#define GGML_MAX_DIMS 4 #define GGML_MAX_DIMS 4
#define GGML_MAX_PARAMS 2048 #define GGML_MAX_PARAMS 2048
#define GGML_MAX_CONTEXTS 64 #define GGML_MAX_CONTEXTS 64
#define GGML_MAX_SRC 62 #define GGML_MAX_SRC 10
#ifndef GGML_MAX_NAME #ifndef GGML_MAX_NAME
#define GGML_MAX_NAME 64 #define GGML_MAX_NAME 64
#endif #endif

29
llama.cpp
View file

@ -746,19 +746,19 @@ static const std::map<llm_arch, std::map<llm_tensor, std::string>> LLM_TENSOR_NA
{ {
LLM_ARCH_QWEN2MOE, LLM_ARCH_QWEN2MOE,
{ {
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" }, { LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" }, { LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_OUTPUT, "output" }, { LLM_TENSOR_OUTPUT, "output" },
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" }, { LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" }, { LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" }, { LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" }, { LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" }, { LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" }, { LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
{ LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" }, { LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" },
{ LLM_TENSOR_FFN_GATE_EXPS, "blk.%d.ffn_gate_exps" }, { LLM_TENSOR_FFN_GATE_EXPS, "blk.%d.ffn_gate_exps" },
{ LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" }, { LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" },
{ LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" }, { LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" },
{ LLM_TENSOR_FFN_GATE_INP_SHARED_EXP, "blk.%d.ffn_gate_inp_shared_exp" }, { LLM_TENSOR_FFN_GATE_INP_SHARED_EXP, "blk.%d.ffn_gate_inp_shared_exp" },
{ LLM_TENSOR_FFN_GATE_SHARED_EXP, "blk.%d.ffn_gate_shared_exp" }, { LLM_TENSOR_FFN_GATE_SHARED_EXP, "blk.%d.ffn_gate_shared_exp" },
{ LLM_TENSOR_FFN_DOWN_SHARED_EXP, "blk.%d.ffn_down_shared_exp" }, { LLM_TENSOR_FFN_DOWN_SHARED_EXP, "blk.%d.ffn_down_shared_exp" },
@ -1751,6 +1751,7 @@ enum e_model {
MODEL_MEDIUM, MODEL_MEDIUM,
MODEL_LARGE, MODEL_LARGE,
MODEL_XL, MODEL_XL,
MODEL_A2_7B,
MODEL_8x7B, MODEL_8x7B,
MODEL_8x22B, MODEL_8x22B,
MODEL_16x12B, MODEL_16x12B,
@ -3916,7 +3917,7 @@ static void llm_load_hparams(
{ {
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
switch (hparams.n_layer) { switch (hparams.n_layer) {
case 24: model.type = e_model::MODEL_1B; break; case 24: model.type = e_model::MODEL_A2_7B; break;
default: model.type = e_model::MODEL_UNKNOWN; default: model.type = e_model::MODEL_UNKNOWN;
} }
} break; } break;