Skip to content

fix: lora with split qkv compatibility check #1836

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

Open
stduhpf wants to merge 1 commit into
base: master
Choose a base branch
from
+60 −63
Open

fix: lora with split qkv compatibility check #1836

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
123 changes: 60 additions & 63 deletions src/model/adapter/lora.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -588,14 +588,17 @@ struct LoraModel : public GGMLRunner {
const std::string& model_tensor_name) {
ggml_tensor* out_diff = nullptr;
int index = 0;

std::vector<std::string> used_tensors;
bool is_conv2d = forward_params.op_type == WeightAdapter::ForwardParams::op_type_t::OP_CONV2D;

while (true) {
std::string key;
if (index == 0) {
key = model_tensor_name;
} else {
key = model_tensor_name + "." + std::to_string(index);
}
bool is_conv2d = forward_params.op_type == WeightAdapter::ForwardParams::op_type_t::OP_CONV2D;

std::string lokr_w1_name = "lora." + key + ".lokr_w1";
std::string lokr_w1_a_name = "lora." + key + ".lokr_w1_a";
Expand Down Expand Up @@ -663,7 +666,6 @@ struct LoraModel : public GGMLRunner {
if (iter != lora_tensors.end()) {
float alpha = ggml_ext_backend_tensor_get_f32(iter->second);
scale_value = alpha / rank;
applied_lora_tensors.insert(alpha_name);
}

if (rank == 1) {
Expand All @@ -678,19 +680,27 @@ struct LoraModel : public GGMLRunner {
out_diff = ggml_concat(ctx, out_diff, curr_out_diff, 0);
}

if (lokr_w1)
applied_lora_tensors.insert(lokr_w1_name);
if (lokr_w1_a)
applied_lora_tensors.insert(lokr_w1_a_name);
if (lokr_w1_b)
applied_lora_tensors.insert(lokr_w1_b_name);
if (lokr_w2)
applied_lora_tensors.insert(lokr_w2_name);
if (lokr_w2_a)
applied_lora_tensors.insert(lokr_w2_a_name);
if (lokr_w2_b)
applied_lora_tensors.insert(lokr_w2_b_name);
applied_lora_tensors.insert(alpha_name);
if (lokr_w1) {
used_tensors.push_back(lokr_w1_name);
}
if (lokr_w1_a) {
used_tensors.push_back(lokr_w1_a_name);
}
if (lokr_w1_b) {
used_tensors.push_back(lokr_w1_b_name);
}
if (lokr_w2) {
used_tensors.push_back(lokr_w2_name);
}
if (lokr_w2_a) {
used_tensors.push_back(lokr_w2_a_name);
}
if (lokr_w2_b) {
used_tensors.push_back(lokr_w2_b_name);
}
if (iter != lora_tensors.end()) {
used_tensors.push_back(alpha_name);
}

index++;
continue;
Expand Down Expand Up @@ -736,64 +746,18 @@ struct LoraModel : public GGMLRunner {
break;
}

if (!is_conv2d) {
const int64_t down_in = lora_down->ne[0];
const int64_t down_out = lora_down->ne[1];
const int64_t up_in = lora_up->ne[0];
const int64_t up_out = lora_up->ne[1];

bool compatible = down_in == model_weight->ne[0] &&
up_out == model_weight->ne[1];
if (lora_mid != nullptr) {
compatible = compatible &&
lora_mid->ne[0] == down_out &&
up_in == lora_mid->ne[1];
} else {
compatible = compatible && up_in == down_out;
}

if (!compatible) {
skipped_incompatible_lora_tensors.insert(lora_down_name);
skipped_incompatible_lora_tensors.insert(lora_up_name);
skipped_incompatible_lora_tensors.insert(lora_mid_name);
skipped_incompatible_lora_tensors.insert(scale_name);
skipped_incompatible_lora_tensors.insert(alpha_name);
if (warned_incompatible_model_tensors.insert(model_tensor_name).second) {
LOG_WARN("skip incompatible LoRA tensor |%s|: model shape = [%lld, %lld], down shape = [%lld, %lld], up shape = [%lld, %lld]",
model_tensor_name.c_str(),
static_cast<long long>(model_weight->ne[0]),
static_cast<long long>(model_weight->ne[1]),
static_cast<long long>(down_in),
static_cast<long long>(down_out),
static_cast<long long>(up_in),
static_cast<long long>(up_out));
}
index++;
continue;
}
}

applied_lora_tensors.insert(lora_up_name);
applied_lora_tensors.insert(lora_down_name);

if (lora_mid) {
applied_lora_tensors.insert(lora_mid_name);
}

float scale_value = 1.0f;

int64_t rank = lora_down->ne[ggml_n_dims(lora_down) - 1];
iter = lora_tensors.find(scale_name);
if (iter != lora_tensors.end()) {
scale_value = ggml_ext_backend_tensor_get_f32(iter->second);
applied_lora_tensors.insert(scale_name);
} else {
iter = lora_tensors.find(alpha_name);
if (iter != lora_tensors.end()) {
float alpha = ggml_ext_backend_tensor_get_f32(iter->second);
scale_value = alpha / rank;
// LOG_DEBUG("rank %s %ld %.2f %.2f", alpha_name.c_str(), rank, alpha, scale_value);
applied_lora_tensors.insert(alpha_name);
}
}
scale_value *= multiplier;
Expand Down Expand Up @@ -853,15 +817,48 @@ struct LoraModel : public GGMLRunner {
}

auto curr_out_diff = ggml_ext_scale(ctx, lx, scale_value, true);

if (out_diff == nullptr) {
out_diff = curr_out_diff;
} else {
out_diff = ggml_concat(ctx, out_diff, curr_out_diff, 0);
out_diff = ggml_concat(ctx, out_diff, curr_out_diff, is_conv2d ? 2 : 0);
}

used_tensors.push_back(lora_up_name);
used_tensors.push_back(lora_down_name);
if (lora_mid) {
used_tensors.push_back(lora_mid_name);
}
if (lora_tensors.find(scale_name) != lora_tensors.end()) {
used_tensors.push_back(scale_name);
}
if (lora_tensors.find(alpha_name) != lora_tensors.end()) {
used_tensors.push_back(alpha_name);
}

index++;
}

if (out_diff == nullptr)
return nullptr;

int64_t expected_out_dim = is_conv2d ? model_weight->ne[3] : model_weight->ne[1];
int64_t actual_out_dim = out_diff->ne[is_conv2d ? 2 : 0];

if (actual_out_dim != expected_out_dim) {
for (const auto& name : used_tensors) {
skipped_incompatible_lora_tensors.insert(name);
}
if (warned_incompatible_model_tensors.insert(model_tensor_name).second) {
LOG_WARN("skip incompatible LoRA tensors for |%s|: output dim %lld != model dim %lld",
model_tensor_name.c_str(), actual_out_dim, expected_out_dim);
}
return nullptr;
}

for (const auto& name : used_tensors) {
applied_lora_tensors.insert(name);
}

return out_diff;
}

Expand Down
Loading