ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-04-30 19:31:48 +00:00
Code Issues Packages Projects Releases Wiki Activity

WIP: allocate graph

Browse Source
This commit is contained in:
Kawrakow
2025-12-03 07:54:53 +00:00
parent b415e734e5
commit dde8028336
2 changed files with 202 additions and 18 deletions
Download Patch File Download Diff File Expand all files Collapse all files

208
src/llama-build-context.cpp
View File

@@ -61,6 +61,110 @@ llm_build_context::llm_build_context(
// all initializations should be done in init() // all initializations should be done in init()
} }
struct GraphAllocator {
size_t cur_offset = 0;
size_t max_offset = 0;
std::unordered_map<ggml_tensor *, size_t> tensors;
std::vector<std::pair<std::size_t,std::size_t>> gaps;
bool debug = false;
inline void add(ggml_tensor * t, bool in_place = false) {
if (t->view_src) {
auto it = tensors.find(t->view_src);
GGML_ASSERT(it != tensors.end());
auto status = tensors.emplace(std::make_pair(t, it->second + t->view_offs));
GGML_ASSERT(status.second);
return;
}
auto offset = cur_offset;
auto nbytes = ggml_nbytes(t);
if (in_place) {
auto it = tensors.find(t->src[0]);
GGML_ASSERT(it != tensors.end());
offset = it->second;
} else {
for (auto it = gaps.begin(); it != gaps.end(); ++it) {
if (nbytes == it->second) {
offset = it->first;
gaps.erase(it);
break;
}
}
}
auto status = tensors.emplace(std::make_pair(t, offset));
GGML_ASSERT(status.second);
if (debug) {
printf("Added tensor %s at offset %zu. cur_offset = %zu, max_offset = %zu\n", t->name, offset, cur_offset, max_offset);
}
if (offset == cur_offset) {
cur_offset += ggml_nbytes(t);
max_offset = std::max(max_offset, cur_offset);
}
}
inline void add(ggml_tensor * t, ggml_tensor * prev) {
GGML_ASSERT(ggml_nbytes(t) == ggml_nbytes(prev));
auto it = tensors.find(t);
GGML_ASSERT(it != tensors.end());
auto status = tensors.emplace(std::make_pair(t, it->second));
GGML_ASSERT(status.second);
}
inline void remove(ggml_tensor * t) {
if (t->view_src) return;
auto it = tensors.find(t);
GGML_ASSERT(it != tensors.end());
auto nbytes = ggml_nbytes(t);
if (it->second + nbytes == cur_offset) {
cur_offset -= nbytes;
if (debug) {
printf("Removed tensor %s at end, cur_offset is now %zu\n", t->name, cur_offset);
}
return;
}
for (int i = 0; i < int(gaps.size()); ++i) {
auto & gap = gaps[i];
if (gap.first + gap.second == it->second) {
gap.second += nbytes;
if (debug) printf("Added gap for removed tensor %s to existing gap %zu, %zu\n", t->name, gap.first, gap.second);
if (i + 1 < int(gaps.size()) && gap.first + gap.second == gaps[i+1].first) {
if (debug) printf(" merged gaps %d, %d\n", i, i+1);
gaps.erase(gaps.begin() + i + 1);
}
return;
}
}
gaps.push_back({it->second, nbytes});
if (debug) {
auto & gap = gaps.back();
printf("Created new gap %d: %zu,%zu after removing tensor %s\n", int(gaps.size()), gap.first, gap.second, t->name);
}
}
inline void remove_from(ggml_tensor * t, bool after) {
auto it = tensors.find(t);
GGML_ASSERT(it != tensors.end());
cur_offset = it->second;
if (after) cur_offset += ggml_nbytes(t);
for (auto gap = gaps.begin(); gap != gaps.end(); ) {
if (gap->first >= cur_offset) {
gap = gaps.erase(gap);
} else {
++gap;
}
}
}
inline size_t offset(ggml_tensor * t) const {
auto it = tensors.find(t);
GGML_ASSERT(it != tensors.end());
return it->second;
}
};
struct llm_build_context_data {
std::vector<GraphAllocator> alloc;
};
llm_build_context::~llm_build_context() = default;
void llm_build_context::init() { void llm_build_context::init() {
struct ggml_init_params params = { struct ggml_init_params params = {
/*.mem_size =*/ buf_compute_meta.size(), /*.mem_size =*/ buf_compute_meta.size(),
@@ -85,6 +189,12 @@ void llm_build_context::init() {
lctx.inp_pos_bucket = nullptr; lctx.inp_pos_bucket = nullptr;
lctx.inp_embd_enc = nullptr; lctx.inp_embd_enc = nullptr;
lctx.inp_KQ_mask_cross = nullptr; lctx.inp_KQ_mask_cross = nullptr;
auto & model = lctx.model;
if (model.split_mode == LLAMA_SPLIT_MODE_GRAPH && model.splits.size() > 1) {
data = std::make_unique<llm_build_context_data>();
data->alloc.resize(model.splits.size());
}
} }
void llm_build_context::free() { void llm_build_context::free() {
@@ -1502,7 +1612,7 @@ std::tuple<ggml_tensor*, ggml_tensor*, ggml_tensor*> llm_build_context::llm_buil
ggml_tensor * wq, ggml_tensor * bq, ggml_tensor * wq, ggml_tensor * bq,
ggml_tensor * wk, ggml_tensor * bk, ggml_tensor * wk, ggml_tensor * bk,
ggml_tensor * wv, ggml_tensor * bv, ggml_tensor * wv, ggml_tensor * bv,
float attention_scale, int il) const { float attention_scale, int il, GraphAllocator * alloc) const {
auto Qcur = llm_build_lora_mm(lctx, ctx0, wq, cur); auto Qcur = llm_build_lora_mm(lctx, ctx0, wq, cur);
cb(Qcur, "Qcur", il); cb(Qcur, "Qcur", il);
auto Kcur = llm_build_lora_mm(lctx, ctx0, wk, cur); auto Kcur = llm_build_lora_mm(lctx, ctx0, wk, cur);
@@ -1512,25 +1622,34 @@ std::tuple<ggml_tensor*, ggml_tensor*, ggml_tensor*> llm_build_context::llm_buil
ggml_build_forward_expand(gf, Qcur); ggml_build_forward_expand(gf, Qcur);
ggml_build_forward_expand(gf, Kcur); ggml_build_forward_expand(gf, Kcur);
ggml_build_forward_expand(gf, Vcur); ggml_build_forward_expand(gf, Vcur);
if (alloc) {
alloc->add(Qcur);
alloc->add(Kcur);
alloc->add(Vcur);
}
if (attention_scale != 0) { if (attention_scale != 0) {
Qcur = ggml_scale(ctx0, Qcur, attention_scale); Qcur = ggml_scale(ctx0, Qcur, attention_scale);
cb(Qcur, "Qcur", il); cb(Qcur, "Qcur", il);
if (alloc) alloc->add(Qcur, true);
} }
if (bq) { if (bq) {
Qcur = ggml_add(ctx0, Qcur, bq); Qcur = ggml_add(ctx0, Qcur, bq);
cb(Qcur, "Qcur", il); cb(Qcur, "Qcur", il);
ggml_build_forward_expand(gf, Qcur); ggml_build_forward_expand(gf, Qcur);
if (alloc) alloc->add(Qcur, true);
} }
if (bk) { if (bk) {
Kcur = ggml_add(ctx0, Kcur, bk); Kcur = ggml_add(ctx0, Kcur, bk);
cb(Kcur, "Kcur", il); cb(Kcur, "Kcur", il);
ggml_build_forward_expand(gf, Kcur); ggml_build_forward_expand(gf, Kcur);
if (alloc) alloc->add(Kcur, true);
} }
if (bv) { if (bv) {
Vcur = ggml_add(ctx0, Vcur, bv); Vcur = ggml_add(ctx0, Vcur, bv);
cb(Vcur, "Vcur", il); cb(Vcur, "Vcur", il);
ggml_build_forward_expand(gf, Vcur); ggml_build_forward_expand(gf, Vcur);
if (alloc) alloc->add(Vcur, true);
} }
return {Qcur, Kcur, Vcur}; return {Qcur, Kcur, Vcur};
} }
@@ -1541,15 +1660,17 @@ std::tuple<ggml_tensor*, ggml_tensor*, ggml_tensor*> llm_build_context::llm_buil
ggml_tensor * wq, ggml_tensor * bq, ggml_tensor * wq, ggml_tensor * bq,
ggml_tensor * wk, ggml_tensor * bk, ggml_tensor * wk, ggml_tensor * bk,
ggml_tensor * wv, ggml_tensor * bv, ggml_tensor * wv, ggml_tensor * bv,
ggml_tensor * q_norm, ggml_tensor * k_norm, float attention_scale, int il) const { ggml_tensor * q_norm, ggml_tensor * k_norm, float attention_scale, int il, GraphAllocator * alloc) const {
const int64_t n_embd_head = hparams.n_embd_head_v; const int64_t n_embd_head = hparams.n_embd_head_v;
const int64_t n_embd_gqa = hparams.n_embd_v_gqa(); const int64_t n_embd_gqa = hparams.n_embd_v_gqa();
if (wqkv) { if (wqkv) {
auto qkv = llm_build_lora_mm(lctx, ctx0, wqkv, cur); auto qkv = llm_build_lora_mm(lctx, ctx0, wqkv, cur);
if (alloc) alloc->add(qkv);
cb(qkv, "qkv", il); cb(qkv, "qkv", il);
if (bqkv) { if (bqkv) {
qkv = ggml_add(ctx0, qkv, bqkv); qkv = ggml_add(ctx0, qkv, bqkv);
cb(qkv, "qkv_b", il); cb(qkv, "qkv_b", il);
if (alloc) alloc->add(qkv, true);
} }
auto Qcur = ggml_view_3d(ctx0, qkv, n_embd_head, n_head, n_tokens, n_embd_head*sizeof(float), qkv->nb[1], 0*sizeof(float)*(n_embd)); auto Qcur = ggml_view_3d(ctx0, qkv, n_embd_head, n_head, n_tokens, n_embd_head*sizeof(float), qkv->nb[1], 0*sizeof(float)*(n_embd));
auto Kcur = ggml_view_3d(ctx0, qkv, n_embd_head, n_head_kv, n_tokens, n_embd_head*sizeof(float), qkv->nb[1], 1*sizeof(float)*Qcur->ne[0]*Qcur->ne[1]); auto Kcur = ggml_view_3d(ctx0, qkv, n_embd_head, n_head_kv, n_tokens, n_embd_head*sizeof(float), qkv->nb[1], 1*sizeof(float)*Qcur->ne[0]*Qcur->ne[1]);
@@ -1557,15 +1678,22 @@ std::tuple<ggml_tensor*, ggml_tensor*, ggml_tensor*> llm_build_context::llm_buil
cb(Qcur, "Qcur", il); cb(Qcur, "Qcur", il);
cb(Kcur, "Kcur", il); cb(Kcur, "Kcur", il);
cb(Vcur, "Vcur", il); cb(Vcur, "Vcur", il);
if (alloc) {
alloc->add(Qcur);
alloc->add(Kcur);
alloc->add(Vcur);
}
if (q_norm) { if (q_norm) {
Qcur = llm_build_norm(ctx0, Qcur, hparams, q_norm, NULL, LLM_NORM_RMS, cb, il); Qcur = llm_build_norm(ctx0, Qcur, hparams, q_norm, NULL, LLM_NORM_RMS, cb, il);
cb(Qcur, "Qcur_normed", il); cb(Qcur, "Qcur_normed", il);
ggml_build_forward_expand(gf, Qcur); ggml_build_forward_expand(gf, Qcur);
if (alloc) alloc->add(Qcur, true);
} }
if (k_norm) { if (k_norm) {
Kcur = llm_build_norm(ctx0, Kcur, hparams, k_norm, NULL, LLM_NORM_RMS, cb, il); Kcur = llm_build_norm(ctx0, Kcur, hparams, k_norm, NULL, LLM_NORM_RMS, cb, il);
cb(Kcur, "Kcur_normed", il); cb(Kcur, "Kcur_normed", il);
ggml_build_forward_expand(gf, Kcur); ggml_build_forward_expand(gf, Kcur);
if (alloc) alloc->add(Kcur, true);
} }
return {Qcur, Kcur, Vcur}; return {Qcur, Kcur, Vcur};
@@ -1577,49 +1705,63 @@ std::tuple<ggml_tensor*, ggml_tensor*, ggml_tensor*> llm_build_context::llm_buil
if (wqk) { if (wqk) {
auto qk = llm_build_lora_mm(lctx, ctx0, wqk, cur); auto qk = llm_build_lora_mm(lctx, ctx0, wqk, cur);
if (alloc) alloc->add(qk);
cb(qk, "qkv", il); cb(qk, "qkv", il);
if (bqk) { if (bqk) {
qk = ggml_add(ctx0, qk, bqk); qk = ggml_add(ctx0, qk, bqk);
cb(qk, "qkv_b", il); cb(qk, "qkv_b", il);
if (alloc) alloc->add(qk, true);
} }
auto Vcur = llm_build_lora_mm(lctx, ctx0, wv, cur); auto Vcur = llm_build_lora_mm(lctx, ctx0, wv, cur);
cb(Vcur, "Vcur", il); cb(Vcur, "Vcur", il);
if (alloc) alloc->add(Vcur);
if (bv) { if (bv) {
Vcur = ggml_add(ctx0, Vcur, bv); Vcur = ggml_add(ctx0, Vcur, bv);
cb(Vcur, "Vcur", il); cb(Vcur, "Vcur", il);
if (alloc) alloc->add(Vcur, true);
} }
ggml_build_forward_expand(gf, qk); ggml_build_forward_expand(gf, qk);
ggml_build_forward_expand(gf, Vcur); ggml_build_forward_expand(gf, Vcur);
auto Qcur = ggml_view_3d(ctx0, qk, n_embd_head, n_head, n_tokens, n_embd_head*sizeof(float), qk->nb[1], 0*sizeof(float)*(n_embd)); auto Qcur = ggml_view_3d(ctx0, qk, n_embd_head, n_head, n_tokens, n_embd_head*sizeof(float), qk->nb[1], 0*sizeof(float)*(n_embd));
auto Kcur = ggml_view_3d(ctx0, qk, n_embd_head, n_head_kv, n_tokens, n_embd_head*sizeof(float), qk->nb[1], 1*sizeof(float)*Qcur->ne[0]*Qcur->ne[1]); auto Kcur = ggml_view_3d(ctx0, qk, n_embd_head, n_head_kv, n_tokens, n_embd_head*sizeof(float), qk->nb[1], 1*sizeof(float)*Qcur->ne[0]*Qcur->ne[1]);
if (alloc) {
alloc->add(Qcur);
alloc->add(Vcur);
}
cb(Qcur, "Qcur", il); cb(Qcur, "Qcur", il);
cb(Kcur, "Kcur", il); cb(Kcur, "Kcur", il);
if (q_norm) { if (q_norm) {
Qcur = llm_build_norm(ctx0, Qcur, hparams, q_norm, NULL, LLM_NORM_RMS, cb, il); Qcur = llm_build_norm(ctx0, Qcur, hparams, q_norm, NULL, LLM_NORM_RMS, cb, il);
cb(Qcur, "Qcur_normed", il); cb(Qcur, "Qcur_normed", il);
ggml_build_forward_expand(gf, Qcur); ggml_build_forward_expand(gf, Qcur);
if (alloc) alloc->add(Qcur, true);
} }
if (k_norm) { if (k_norm) {
Kcur = llm_build_norm(ctx0, Kcur, hparams, k_norm, NULL, LLM_NORM_RMS, cb, il); Kcur = llm_build_norm(ctx0, Kcur, hparams, k_norm, NULL, LLM_NORM_RMS, cb, il);
cb(Kcur, "Kcur_normed", il); cb(Kcur, "Kcur_normed", il);
ggml_build_forward_expand(gf, Kcur); ggml_build_forward_expand(gf, Kcur);
if (alloc) alloc->add(Kcur, true);
} }
return {Qcur, Kcur, Vcur}; return {Qcur, Kcur, Vcur};
} }
auto [Q, K, V] = llm_build_mul_mat_qkv(gf, cur, wq, bq, wk, bk, wv, bv, attention_scale, il); auto [Q, K, V] = llm_build_mul_mat_qkv(gf, cur, wq, bq, wk, bk, wv, bv, attention_scale, il, alloc);
auto Qcur = ggml_reshape_3d(ctx0, Q, n_embd_head, Q->ne[0]/n_embd_head, n_tokens); auto Qcur = ggml_reshape_3d(ctx0, Q, n_embd_head, Q->ne[0]/n_embd_head, n_tokens);
if (alloc) alloc->add(Qcur);
if (q_norm) { if (q_norm) {
Qcur = llm_build_norm(ctx0, Qcur, hparams, q_norm, NULL, LLM_NORM_RMS, cb, il); Qcur = llm_build_norm(ctx0, Qcur, hparams, q_norm, NULL, LLM_NORM_RMS, cb, il);
cb(Qcur, "Qcur_normed", il); cb(Qcur, "Qcur_normed", il);
if (alloc) alloc->add(Qcur, true);
} }
auto Kcur = ggml_reshape_3d(ctx0, K, n_embd_head, K->ne[0]/n_embd_head, n_tokens); auto Kcur = ggml_reshape_3d(ctx0, K, n_embd_head, K->ne[0]/n_embd_head, n_tokens);
if (alloc) alloc->add(Kcur);
if (k_norm) { if (k_norm) {
Kcur = llm_build_norm(ctx0, Kcur, hparams, k_norm, NULL, LLM_NORM_RMS, cb, il); Kcur = llm_build_norm(ctx0, Kcur, hparams, k_norm, NULL, LLM_NORM_RMS, cb, il);
cb(Kcur, "Kcur_normed", il); cb(Kcur, "Kcur_normed", il);
if (alloc) alloc->add(Kcur, true);
} }
auto Vcur = V; auto Vcur = V;
return {Qcur, Kcur, Vcur}; return {Qcur, Kcur, Vcur};
@@ -9238,6 +9380,8 @@ ggml_tensor * llm_build_context::build_std_attention(ggml_cgraph * gf, ggml_tens
} }
std::vector<ggml_tensor*> attn; attn.reserve(wq->n_device); std::vector<ggml_tensor*> attn; attn.reserve(wq->n_device);
for (int id = 0; id < wq->n_device; ++id) { for (int id = 0; id < wq->n_device; ++id) {
auto alloc = &data->alloc[id];
alloc->debug = true;
int il_cb = 1000*(id+1) + il; int il_cb = 1000*(id+1) + il;
auto split_wq = wq->splits[id]; auto split_wq = wq->splits[id];
auto split_wk = wk->splits[id]; auto split_wk = wk->splits[id];
@@ -9248,14 +9392,20 @@ ggml_tensor * llm_build_context::build_std_attention(ggml_cgraph * gf, ggml_tens
GGML_ASSERT((!split_wq && !split_wk && !split_wv && !split_wo && !split_kl && !split_vl) || GGML_ASSERT((!split_wq && !split_wk && !split_wv && !split_wo && !split_kl && !split_vl) ||
(split_wq && split_wk && split_wv && split_wo && split_kl && split_vl)); (split_wq && split_wk && split_wv && split_wo && split_kl && split_vl));
if (!split_wq) continue; if (!split_wq) continue;
alloc->add(input);
auto cur = input; auto cur = input;
auto offset = alloc->offset(input) + ggml_nbytes(input);
if (attn_norm) { if (attn_norm) {
auto split_norm = attn_norm->splits[id]; auto split_norm = attn_norm->splits[id];
cur = llm_build_norm(ctx0, cur, hparams, split_norm, NULL, LLM_NORM_RMS, cb, il); cur = llm_build_norm(ctx0, cur, hparams, split_norm, NULL, LLM_NORM_RMS, cb, il);
cb(cur, "attn_norm", il_cb); cb(cur, "attn_norm", il_cb);
alloc->add(cur);
offset = alloc->offset(cur);
} }
else if (cur->type != GGML_TYPE_F32) { else if (cur->type != GGML_TYPE_F32) {
cur = ggml_cast(ctx0, cur, GGML_TYPE_F32); cur = ggml_cast(ctx0, cur, GGML_TYPE_F32);
alloc->add(cur);
offset = alloc->offset(cur);
} }
auto the_q_norm = model.layers[il].attn_q_norm ? model.layers[il].attn_q_norm->extra ? auto the_q_norm = model.layers[il].attn_q_norm ? model.layers[il].attn_q_norm->extra ?
((ggml_split_tensor_t *)model.layers[il].attn_q_norm->extra)->splits[id] : model.layers[il].attn_q_norm : nullptr; ((ggml_split_tensor_t *)model.layers[il].attn_q_norm->extra)->splits[id] : model.layers[il].attn_q_norm : nullptr;
@@ -9265,7 +9415,7 @@ ggml_tensor * llm_build_context::build_std_attention(ggml_cgraph * gf, ggml_tens
split_wq, bq ? bq->splits[id] : nullptr, split_wq, bq ? bq->splits[id] : nullptr,
split_wk, bk ? bk->splits[id] : nullptr, split_wk, bk ? bk->splits[id] : nullptr,
split_wv, bv ? bv->splits[id] : nullptr, split_wv, bv ? bv->splits[id] : nullptr,
the_q_norm, the_k_norm, f_attn_scale, il_cb); the_q_norm, the_k_norm, f_attn_scale, il_cb, alloc);
auto rope_factors = rope_factors_in; auto rope_factors = rope_factors_in;
if (!rope_factors && model.layers[il].rope_freqs && model.layers[il].rope_freqs->extra) { if (!rope_factors && model.layers[il].rope_freqs && model.layers[il].rope_freqs->extra) {
auto extra = (ggml_split_tensor_t *)model.layers[il].rope_freqs->extra; auto extra = (ggml_split_tensor_t *)model.layers[il].rope_freqs->extra;
@@ -9273,8 +9423,10 @@ ggml_tensor * llm_build_context::build_std_attention(ggml_cgraph * gf, ggml_tens
} }
Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, rope_factors, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale, Qcur = ggml_rope_ext(ctx0, Qcur, inp_pos, rope_factors, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
ext_factor, attn_factor, beta_fast, beta_slow); ext_factor, attn_factor, beta_fast, beta_slow);
alloc->add(Qcur, true);
Kcur = ggml_rope_ext(ctx0, Kcur, inp_pos, rope_factors, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale, Kcur = ggml_rope_ext(ctx0, Kcur, inp_pos, rope_factors, n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
ext_factor, attn_factor, beta_fast, beta_slow); ext_factor, attn_factor, beta_fast, beta_slow);
alloc->add(Kcur, true);
cb(Qcur, "Qcur", il_cb); cb(Qcur, "Qcur", il_cb);
cb(Kcur, "Kcur", il_cb); cb(Kcur, "Kcur", il_cb);
ggml_build_forward_expand(gf, Qcur); ggml_build_forward_expand(gf, Qcur);
@@ -9300,19 +9452,26 @@ ggml_tensor * llm_build_context::build_std_attention(ggml_cgraph * gf, ggml_tens
struct ggml_tensor * v_cache_view = nullptr; struct ggml_tensor * v_cache_view = nullptr;
if (cparams.flash_attn) { GGML_ASSERT(cparams.flash_attn);
v_cache_view = ggml_view_1d(ctx0, split_vl, n_tokens*split_wv->ne[1], v_cache_view = ggml_view_1d(ctx0, split_vl, n_tokens*split_wv->ne[1],
kv_head*ggml_row_size(split_vl->type, split_wv->ne[1])); kv_head*ggml_row_size(split_vl->type, split_wv->ne[1]));
lctx.cache_copies[idx+1].step = ggml_row_size(split_vl->type, split_wv->ne[1]); lctx.cache_copies[idx+1].step = ggml_row_size(split_vl->type, split_wv->ne[1]);
} else { alloc->remove(Vcur);
// note: the V cache is transposed when not using flash attention alloc->remove(Kcur);
v_cache_view = ggml_view_2d(ctx0, split_vl, n_tokens, split_wv->ne[1],
( n_ctx)*ggml_element_size(split_vl),
(kv_head)*ggml_element_size(split_vl));
lctx.cache_copies[idx+1].step = ggml_element_size(split_vl);
Vcur = ggml_transpose(ctx0, Vcur); //if (cparams.flash_attn) {
} // v_cache_view = ggml_view_1d(ctx0, split_vl, n_tokens*split_wv->ne[1],
// kv_head*ggml_row_size(split_vl->type, split_wv->ne[1]));
// lctx.cache_copies[idx+1].step = ggml_row_size(split_vl->type, split_wv->ne[1]);
//} else {
// // note: the V cache is transposed when not using flash attention
// v_cache_view = ggml_view_2d(ctx0, split_vl, n_tokens, split_wv->ne[1],
// ( n_ctx)*ggml_element_size(split_vl),
// (kv_head)*ggml_element_size(split_vl));
// lctx.cache_copies[idx+1].step = ggml_element_size(split_vl);
// Vcur = ggml_transpose(ctx0, Vcur);
//}
cb(v_cache_view, "v_cache_view", il_cb); cb(v_cache_view, "v_cache_view", il_cb);
lctx.cache_copies[idx+1].cpy = ggml_cpy(ctx0, Vcur, v_cache_view); lctx.cache_copies[idx+1].cpy = ggml_cpy(ctx0, Vcur, v_cache_view);
@@ -9320,6 +9479,7 @@ ggml_tensor * llm_build_context::build_std_attention(ggml_cgraph * gf, ggml_tens
auto q = ggml_permute(ctx0, Qcur, 0, 2, 1, 3); auto q = ggml_permute(ctx0, Qcur, 0, 2, 1, 3);
cb(q, "q", il_cb); cb(q, "q", il_cb);
alloc->add(q);
auto k = ggml_view_3d(ctx0, split_kl, n_embd_head_k, n_kv, n_head_kv, auto k = ggml_view_3d(ctx0, split_kl, n_embd_head_k, n_kv, n_head_kv,
ggml_row_size(split_kl->type, n_embd_head_k)*n_head_kv, //n_embd_k_gqa), ggml_row_size(split_kl->type, n_embd_head_k)*n_head_kv, //n_embd_k_gqa),
@@ -9343,6 +9503,8 @@ ggml_tensor * llm_build_context::build_std_attention(ggml_cgraph * gf, ggml_tens
if (n_swa > 0) { if (n_swa > 0) {
((int32_t *)cur->op_params)[4] = n_swa; ((int32_t *)cur->op_params)[4] = n_swa;
} }
alloc->add(cur);
alloc->remove(Qcur);
// Some models produced NaNs/gibberish when FA is computed with f16 precision on CUDA // Some models produced NaNs/gibberish when FA is computed with f16 precision on CUDA
if (use_f32_precision || model.arch == LLM_ARCH_PHI2 || model.arch == LLM_ARCH_PHI3 || model.arch == LLM_ARCH_GPTNEOX || if (use_f32_precision || model.arch == LLM_ARCH_PHI2 || model.arch == LLM_ARCH_PHI3 || model.arch == LLM_ARCH_GPTNEOX ||
@@ -9353,8 +9515,10 @@ ggml_tensor * llm_build_context::build_std_attention(ggml_cgraph * gf, ggml_tens
cur = ggml_reshape_2d(ctx0, cur, split_wo->ne[0], n_tokens); cur = ggml_reshape_2d(ctx0, cur, split_wo->ne[0], n_tokens);
cb(cur, "flash_attn_reshaped", il_cb); cb(cur, "flash_attn_reshaped", il_cb);
alloc->add(cur);
cur = llm_build_lora_mm(lctx, ctx0, split_wo, cur); cur = llm_build_lora_mm(lctx, ctx0, split_wo, cur);
alloc->add(cur);
if (lctx.model.arch == LLM_ARCH_GLM4 || lctx.model.arch == LLM_ARCH_GLM4_MOE) { if (lctx.model.arch == LLM_ARCH_GLM4 || lctx.model.arch == LLM_ARCH_GLM4_MOE) {
// GLM4 and GLM4_MOE seem to have numerical issues with half-precision accumulators // GLM4 and GLM4_MOE seem to have numerical issues with half-precision accumulators
ggml_mul_mat_set_prec(cur, GGML_PREC_F32); ggml_mul_mat_set_prec(cur, GGML_PREC_F32);
@@ -9363,22 +9527,34 @@ ggml_tensor * llm_build_context::build_std_attention(ggml_cgraph * gf, ggml_tens
if (bo) { if (bo) {
cur = ggml_add(ctx0, cur, bo->splits[id]); cur = ggml_add(ctx0, cur, bo->splits[id]);
cb(cur, "kqv_wo_biased", il_cb); cb(cur, "kqv_wo_biased", il_cb);
alloc->add(cur, true);
} }
if (cur->ne[1] >= 32) { if (cur->ne[1] >= 32) {
cur = ggml_cast(ctx0, cur, GGML_TYPE_F16); cur = ggml_cast(ctx0, cur, GGML_TYPE_F16);
alloc->add(cur);
} }
ggml_build_forward_expand(gf, cur); ggml_build_forward_expand(gf, cur);
attn.push_back(cur); attn.push_back(cur);
if (id != model.main_gpu) {
data->alloc[id].cur_offset = offset;
}
} }
GGML_ASSERT(!attn.empty()); GGML_ASSERT(!attn.empty());
if (attn.size() == 1) return attn.front(); if (attn.size() == 1) return attn.front();
auto alloc = &data->alloc[model.main_gpu];
if (model.main_gpu != 0) alloc->add(attn[0]);
if (model.main_gpu != 1) alloc->add(attn[1]);
auto cur = ggml_add(ctx0, attn[0], attn[1]); auto cur = ggml_add(ctx0, attn[0], attn[1]);
alloc->add(cur, true);
cb(cur, "combine_attn", il); cb(cur, "combine_attn", il);
cur->op_params[0] = 0xff; cur->op_params[0] = 0xff;
for (int id = 2; id < (int)attn.size(); ++id) { for (int id = 2; id < (int)attn.size(); ++id) {
if (id != model.main_gpu) alloc->add(attn[id]);
cur = ggml_add(ctx0, cur, attn[id]); cur = ggml_add(ctx0, cur, attn[id]);
cb(cur, "combine_attn", il); cb(cur, "combine_attn", il);
alloc->add(cur, true);
} }
alloc->remove_from(cur, true);
return cur; return cur;
} }
} }

12
src/llama-build-context.h
View File

@@ -6,6 +6,7 @@
#include <cstdint> #include <cstdint>
#include <functional> #include <functional>
#include <tuple> #include <tuple>
#include <memory>
struct llama_model; struct llama_model;
struct llama_context; struct llama_context;
@@ -37,6 +38,9 @@ enum llm_norm_type {
LLM_NORM_RMS, LLM_NORM_RMS,
}; };
struct llm_build_context_data;
struct GraphAllocator;
struct llm_build_context { struct llm_build_context {
const llama_model & model; const llama_model & model;
llama_context & lctx; llama_context & lctx;
@@ -94,6 +98,8 @@ struct llm_build_context {
struct ggml_context * ctx0 = nullptr; struct ggml_context * ctx0 = nullptr;
std::unique_ptr<llm_build_context_data> data;
// TODO: consider making the entire interface noexcept // TODO: consider making the entire interface noexcept
llm_build_context( llm_build_context(
llama_context & lctx, llama_context & lctx,
@@ -102,6 +108,8 @@ struct llm_build_context {
bool worst_case, bool worst_case,
bool warmup); bool warmup);
~llm_build_context();
void init(); void init();
void free(); void free();
@@ -148,7 +156,7 @@ struct llm_build_context {
ggml_tensor * wq, ggml_tensor * bq, ggml_tensor * wq, ggml_tensor * bq,
ggml_tensor * wk, ggml_tensor * bk, ggml_tensor * wk, ggml_tensor * bk,
ggml_tensor * wv, ggml_tensor * bv, ggml_tensor * wv, ggml_tensor * bv,
float attention_scale, int il) const; float attention_scale, int il, GraphAllocator * alloc = nullptr) const;
std::tuple<ggml_tensor*, ggml_tensor*, ggml_tensor*> llm_build_mul_mat_qkv(ggml_cgraph * gf, ggml_tensor * cur, std::tuple<ggml_tensor*, ggml_tensor*, ggml_tensor*> llm_build_mul_mat_qkv(ggml_cgraph * gf, ggml_tensor * cur,
ggml_tensor * wqkv, ggml_tensor * bqkv, ggml_tensor * wqkv, ggml_tensor * bqkv,
@@ -156,7 +164,7 @@ struct llm_build_context {
ggml_tensor * wq, ggml_tensor * bq, ggml_tensor * wq, ggml_tensor * bq,
ggml_tensor * wk, ggml_tensor * bk, ggml_tensor * wk, ggml_tensor * bk,
ggml_tensor * wv, ggml_tensor * bv, ggml_tensor * wv, ggml_tensor * bv,
ggml_tensor * q_norm, ggml_tensor * k_norm, float attention_scale, int il) const; ggml_tensor * q_norm, ggml_tensor * k_norm, float attention_scale, int il, GraphAllocator * alloc = nullptr) const;
ggml_cgraph * build_llama(); ggml_cgraph * build_llama();