Skip to content

Commit

Permalink
add skywork support for cpp
Browse files Browse the repository at this point in the history
  • Loading branch information
@yxq321
yxq321 committed Nov 8, 2023
1 parent 56a5838 commit 0e760e8
Showing 1 changed file with 100 additions and 0 deletions.
100 changes: 100 additions & 0 deletions llama.cpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -182,6 +182,7 @@ enum llm_arch {
LLM_ARCH_LLAMA,
LLM_ARCH_FALCON,
LLM_ARCH_BAICHUAN,
LLM_ARCH_SKYWORK,
LLM_ARCH_GPT2,
LLM_ARCH_GPTJ,
LLM_ARCH_GPTNEOX,
Expand All @@ -201,6 +202,7 @@ static std::map<llm_arch, std::string> LLM_ARCH_NAMES = {
{ LLM_ARCH_GPTNEOX, "gptneox" },
{ LLM_ARCH_MPT, "mpt" },
{ LLM_ARCH_BAICHUAN, "baichuan" },
{ LLM_ARCH_SKYWORK, "skywork" },
{ LLM_ARCH_STARCODER, "starcoder" },
{ LLM_ARCH_PERSIMMON, "persimmon" },
{ LLM_ARCH_REFACT, "refact" },
Expand Down Expand Up @@ -375,6 +377,25 @@ static std::map<llm_arch, std::map<llm_tensor, std::string>> LLM_TENSOR_NAMES =
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
},
},
{
LLM_ARCH_SKYWORK, // TODO by yxq.
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_OUTPUT, "output" },
{ LLM_TENSOR_ROPE_FREQS, "rope_freqs" },
{ 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_ROT_EMBD, "blk.%d.attn_rot_embd" },
{ 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_FALCON,
{
Expand Down Expand Up @@ -2160,6 +2181,14 @@ static void llm_load_hparams(
default: model.type = e_model::MODEL_UNKNOWN;
}
} break;
case LLM_ARCH_SKYWORK:
{
GGUF_GET_KEY(ctx, hparams.f_norm_rms_eps, gguf_get_val_f32, GGUF_TYPE_FLOAT32, true, kv(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS));
switch (hparams.n_layer) {
case 52: model.type = e_model::MODEL_13B; break;
default: model.type = e_model::MODEL_UNKNOWN;
}
} break;
case LLM_ARCH_STARCODER:
{
GGUF_GET_KEY(ctx, hparams.f_norm_eps, gguf_get_val_f32, GGUF_TYPE_FLOAT32, true, kv(LLM_KV_ATTENTION_LAYERNORM_EPS));
Expand Down Expand Up @@ -2684,6 +2713,72 @@ static void llm_load_tensors(

model.layers.resize(n_layer);

for (uint32_t i = 0; i < n_layer; ++i) {
const ggml_backend_type backend = int(i) < i_gpu_start ? GGML_BACKEND_CPU : llama_backend_offload; // NOLINT
const ggml_backend_type backend_split = int(i) < i_gpu_start ? GGML_BACKEND_CPU : llama_backend_offload_split; // NOLINT

auto & layer = model.layers[i];

layer.attn_norm = ml.create_tensor(ctx, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, backend);

layer.wq = ml.create_tensor(ctx, tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd}, backend_split);
layer.wk = ml.create_tensor(ctx, tn(LLM_TENSOR_ATTN_K, "weight", i), {n_embd, n_embd_gqa}, backend_split);
layer.wv = ml.create_tensor(ctx, tn(LLM_TENSOR_ATTN_V, "weight", i), {n_embd, n_embd_gqa}, backend_split);
layer.wo = ml.create_tensor(ctx, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, backend_split);

layer.ffn_norm = ml.create_tensor(ctx, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, backend);

layer.ffn_gate = ml.create_tensor(ctx, tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}, backend_split);
layer.ffn_down = ml.create_tensor(ctx, tn(LLM_TENSOR_FFN_DOWN, "weight", i), { n_ff, n_embd}, backend_split);
layer.ffn_up = ml.create_tensor(ctx, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff}, backend_split);

if (backend == GGML_BACKEND_GPU) {
vram_weights +=
ggml_nbytes(layer.attn_norm) + ggml_nbytes(layer.wq) + ggml_nbytes(layer.wk) +
ggml_nbytes(layer.wv) + ggml_nbytes(layer.wo) + ggml_nbytes(layer.ffn_norm) +
ggml_nbytes(layer.ffn_gate) + ggml_nbytes(layer.ffn_down) + ggml_nbytes(layer.ffn_up);
}
}
} break;
case LLM_ARCH_SKYWORK:
{
model.tok_embd = ml.create_tensor(ctx, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, GGML_BACKEND_CPU);
{
ggml_backend_type backend_norm;
ggml_backend_type backend_output;

if (n_gpu_layers > int(n_layer)) {
// norm is not performance relevant on its own but keeping it in VRAM reduces data copying
// on Windows however this is detrimental unless everything is on the GPU
#ifndef _WIN32
backend_norm = llama_backend_offload;
#else
backend_norm = n_gpu_layers <= (int) n_layer + 2 ? GGML_BACKEND_CPU : llama_backend_offload;
#endif // _WIN32

backend_output = llama_backend_offload_split;
} else {
backend_norm = GGML_BACKEND_CPU;
backend_output = GGML_BACKEND_CPU;
}

model.output_norm = ml.create_tensor(ctx, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, backend_norm);
model.output = ml.create_tensor(ctx, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, backend_output);

if (backend_norm == GGML_BACKEND_GPU) {
vram_weights += ggml_nbytes(model.output_norm);
}
if (backend_output == GGML_BACKEND_GPU_SPLIT) {
vram_weights += ggml_nbytes(model.output);
}
}

const uint32_t n_ff = hparams.n_ff;

const int i_gpu_start = n_layer - n_gpu_layers;

model.layers.resize(n_layer);

for (uint32_t i = 0; i < n_layer; ++i) {
const ggml_backend_type backend = int(i) < i_gpu_start ? GGML_BACKEND_CPU : llama_backend_offload; // NOLINT
const ggml_backend_type backend_split = int(i) < i_gpu_start ? GGML_BACKEND_CPU : llama_backend_offload_split; // NOLINT
Expand Down Expand Up @@ -5001,6 +5096,10 @@ static struct ggml_cgraph * llama_build_graph(
{
result = llm.build_baichuan();
} break;
case LLM_ARCH_SKYWORK: // TODO by yxq.
{
result = llm.build_baichuan();
} break;
case LLM_ARCH_FALCON:
{
result = llm.build_falcon();
Expand Down Expand Up @@ -5197,6 +5296,7 @@ static int llama_decode_internal(
const bool full_offload_supported =
model.arch == LLM_ARCH_LLAMA ||
model.arch == LLM_ARCH_BAICHUAN ||
model.arch == LLM_ARCH_SKYWORK ||
model.arch == LLM_ARCH_FALCON ||
model.arch == LLM_ARCH_REFACT ||
model.arch == LLM_ARCH_MPT ||
Expand Down

0 comments on commit 0e760e8

Please sign in to comment.