Skip to content

Commit

Permalink
Fixed all models' control vectors
Browse files Browse the repository at this point in the history
  • Loading branch information
@jukofyork
jukofyork authored Jun 22, 2024
1 parent 2c42c51 commit f57ebf2
Showing 1 changed file with 111 additions and 32 deletions.
143 changes: 111 additions & 32 deletions llama.cpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -7818,8 +7818,7 @@ struct llm_build_context {
cur = ggml_add(ctx0, cur, ffn_inp);
cb(cur, "ffn_out", il);

ggml_tensor * layer_dir = lctx.cvec.tensor_for(il);
if (layer_dir != nullptr) {
if (ggml_tensor * layer_dir = lctx.cvec.tensor_for(il)) {
cur = ggml_add(ctx0, cur, layer_dir);
}
cb(cur, "l_out", il);
Expand Down Expand Up @@ -7936,6 +7935,9 @@ struct llm_build_context {
}

cur = ggml_add(ctx0, cur, ffn_inp);
if (ggml_tensor * layer_dir = lctx.cvec.tensor_for(il)) {
cur = ggml_add(ctx0, cur, layer_dir);
}
cb(cur, "l_out", il);

// input for next layer
Expand Down Expand Up @@ -8040,6 +8042,9 @@ struct llm_build_context {
}

cur = ggml_add(ctx0, cur, ffn_inp);
if (ggml_tensor * layer_dir = lctx.cvec.tensor_for(il)) {
cur = ggml_add(ctx0, cur, layer_dir);
}
cb(cur, "l_out", il);

// input for next layer
Expand Down Expand Up @@ -8158,6 +8163,9 @@ struct llm_build_context {
cb(cur, "l_out", il);

cur = ggml_add(ctx0, cur, inpL);
if (ggml_tensor * layer_dir = lctx.cvec.tensor_for(il)) {
cur = ggml_add(ctx0, cur, layer_dir);
}
cb(cur, "l_out", il);

// input for next layer
Expand Down Expand Up @@ -8309,8 +8317,7 @@ struct llm_build_context {
cur = ggml_add(ctx0, cur, ffn_inp);
cb(cur, "ffn_out", il);

ggml_tensor * layer_dir = lctx.cvec.tensor_for(il);
if (layer_dir != nullptr) {
if (ggml_tensor * layer_dir = lctx.cvec.tensor_for(il)) {
cur = ggml_add(ctx0, cur, layer_dir);
}
cb(cur, "l_out", il);
Expand Down Expand Up @@ -8443,8 +8450,7 @@ struct llm_build_context {
cur = ggml_add(ctx0, cur, ffn_inp);
cb(cur, "ffn_out", il);

ggml_tensor * layer_dir = lctx.cvec.tensor_for(il);
if (layer_dir != nullptr) {
if (ggml_tensor * layer_dir = lctx.cvec.tensor_for(il)) {
cur = ggml_add(ctx0, cur, layer_dir);
}
cb(cur, "l_out", il);
Expand Down Expand Up @@ -8552,8 +8558,14 @@ struct llm_build_context {
cb(cur, "ffn_out", il);
}

inpL = ggml_add(ctx0, cur, ffn_inp);
cb(inpL, "l_out", il);
cur = ggml_add(ctx0, cur, ffn_inp);
if (ggml_tensor * layer_dir = lctx.cvec.tensor_for(il)) {
cur = ggml_add(ctx0, cur, layer_dir);
}
cb(cur, "l_out", il);

// input for next layer
inpL = cur;
}

cur = llm_build_norm(ctx0, inpL, hparams,
Expand Down Expand Up @@ -8641,6 +8653,9 @@ struct llm_build_context {
}

cur = ggml_add(ctx0, cur, ffn_inp);
if (ggml_tensor * layer_dir = lctx.cvec.tensor_for(il)) {
cur = ggml_add(ctx0, cur, layer_dir);
}
cb(cur, "l_out", il);

// input for next layer
Expand Down Expand Up @@ -8936,8 +8951,14 @@ struct llm_build_context {
cb(cur, "ffn_out", il);
}

inpL = ggml_add(ctx0, cur, ffn_inp);
cb(inpL, "l_out", il);
cur = ggml_add(ctx0, cur, ffn_inp);
if (ggml_tensor * layer_dir = lctx.cvec.tensor_for(il)) {
cur = ggml_add(ctx0, cur, layer_dir);
}
cb(cur, "l_out", il);

// input for next layer
inpL = cur;
}

cur = llm_build_norm(ctx0, inpL, hparams,
Expand Down Expand Up @@ -9071,6 +9092,9 @@ struct llm_build_context {
}

cur = ggml_add(ctx0, cur, ffn_inp);
if (ggml_tensor * layer_dir = lctx.cvec.tensor_for(il)) {
cur = ggml_add(ctx0, cur, layer_dir);
}
cb(cur, "l_out", il);

// input for next layer
Expand Down Expand Up @@ -9219,6 +9243,9 @@ struct llm_build_context {
}

cur = ggml_add(ctx0, cur, ffn_inp);
if (ggml_tensor * layer_dir = lctx.cvec.tensor_for(il)) {
cur = ggml_add(ctx0, cur, layer_dir);
}
cb(cur, "l_out", il);

// input for next layer
Expand Down Expand Up @@ -9331,6 +9358,9 @@ struct llm_build_context {
}

cur = ggml_add(ctx0, cur, ffn_inp);
if (ggml_tensor * layer_dir = lctx.cvec.tensor_for(il)) {
cur = ggml_add(ctx0, cur, layer_dir);
}
cb(cur, "l_out", il);

// input for next layer
Expand Down Expand Up @@ -9442,13 +9472,9 @@ struct llm_build_context {
cb(cur, "ffn_out", il);

cur = ggml_add(ctx0, cur, ffn_inp);

// add control vector
ggml_tensor * layer_dir = lctx.cvec.tensor_for(il);
if (layer_dir != nullptr) {
if (ggml_tensor * layer_dir = lctx.cvec.tensor_for(il)) {
cur = ggml_add(ctx0, cur, layer_dir);
}

cb(cur, "l_out", il);

// input for next layer
Expand Down Expand Up @@ -9594,6 +9620,9 @@ struct llm_build_context {
}

cur = ggml_add(ctx0, cur, ffn_inp);
if (ggml_tensor * layer_dir = lctx.cvec.tensor_for(il)) {
cur = ggml_add(ctx0, cur, layer_dir);
}
cb(cur, "l_out", il);

// input for next layer
Expand Down Expand Up @@ -9717,8 +9746,12 @@ struct llm_build_context {
cb(cur, "l_out", il);

cur = ggml_add(ctx0, cur, inpL);
if (ggml_tensor * layer_dir = lctx.cvec.tensor_for(il)) {
cur = ggml_add(ctx0, cur, layer_dir);
}
cb(cur, "l_out", il);

// input for next layer
inpL = cur;
}

Expand Down Expand Up @@ -9850,8 +9883,12 @@ struct llm_build_context {
}

cur = ggml_add(ctx0, residual, cur);
if (ggml_tensor * layer_dir = lctx.cvec.tensor_for(il)) {
cur = ggml_add(ctx0, cur, layer_dir);
}
cb(cur, "l_out", il);

// input for next layer
inpL = cur;
}

Expand Down Expand Up @@ -9953,6 +9990,9 @@ struct llm_build_context {
cb(cur, "l_out", il);

cur = ggml_add(ctx0, cur, inpL);
if (ggml_tensor * layer_dir = lctx.cvec.tensor_for(il)) {
cur = ggml_add(ctx0, cur, layer_dir);
}
cb(cur, "l_out", il);

// input for next layer
Expand Down Expand Up @@ -10058,8 +10098,14 @@ struct llm_build_context {
cb(cur, "ffn_out", il);
}

inpL = ggml_add(ctx0, cur, ffn_inp);
cb(inpL, "l_out", il);
cur = ggml_add(ctx0, cur, ffn_inp);
if (ggml_tensor * layer_dir = lctx.cvec.tensor_for(il)) {
cur = ggml_add(ctx0, cur, layer_dir);
}
cb(cur, "l_out", il);

// input for next layer
inpL = cur;
}

cur = llm_build_norm(ctx0, inpL, hparams,
Expand Down Expand Up @@ -10165,8 +10211,14 @@ struct llm_build_context {
cb(cur, "ffn_out", il);
}

inpL = ggml_add(ctx0, cur, ffn_inp);
cb(inpL, "l_out", il);
cur = ggml_add(ctx0, cur, ffn_inp);
if (ggml_tensor * layer_dir = lctx.cvec.tensor_for(il)) {
cur = ggml_add(ctx0, cur, layer_dir);
}
cb(cur, "l_out", il);

// input for next layer
inpL = cur;
}

cur = llm_build_norm(ctx0, inpL, hparams,
Expand Down Expand Up @@ -10278,6 +10330,9 @@ struct llm_build_context {
cb(cur, "ffn_out", il);

cur = ggml_add(ctx0, cur, ffn_inp);
if (ggml_tensor * layer_dir = lctx.cvec.tensor_for(il)) {
cur = ggml_add(ctx0, cur, layer_dir);
}
cb(cur, "l_out", il);

// input for next layer
Expand Down Expand Up @@ -10395,6 +10450,9 @@ struct llm_build_context {
cb(cur, "ffn_out", il);

cur = ggml_add(ctx0, cur, ffn_inp);
if (ggml_tensor * layer_dir = lctx.cvec.tensor_for(il)) {
cur = ggml_add(ctx0, cur, layer_dir);
}
cb(cur, "l_out", il);

// input for next layer
Expand Down Expand Up @@ -10536,6 +10594,9 @@ struct llm_build_context {
cb(cur, "hidden_scaled_ffn", -1);

cur = ggml_add(ctx0, cur, ffn_inp);
if (ggml_tensor * layer_dir = lctx.cvec.tensor_for(il)) {
cur = ggml_add(ctx0, cur, layer_dir);
}
cb(cur, "l_out", il);

// input for next layer
Expand Down Expand Up @@ -10648,6 +10709,9 @@ struct llm_build_context {
}

cur = ggml_add(ctx0, cur, sa_out);
if (ggml_tensor * layer_dir = lctx.cvec.tensor_for(il)) {
cur = ggml_add(ctx0, cur, layer_dir);
}
cb(cur, "l_out", il);

// input for next layer
Expand Down Expand Up @@ -10765,6 +10829,9 @@ struct llm_build_context {
LLM_FFN_GELU, LLM_FFN_SEQ, cb, il);
cb(cur, "ffn_out", il);
cur = ggml_add(ctx0, cur, ffn_inp);
if (ggml_tensor * layer_dir = lctx.cvec.tensor_for(il)) {
cur = ggml_add(ctx0, cur, layer_dir);
}
cb(cur, "l_out", il);

// input for next layer
Expand Down Expand Up @@ -10913,6 +10980,9 @@ struct llm_build_context {

// residual
cur = ggml_add(ctx0, cur, inpL);
if (ggml_tensor * layer_dir = lctx.cvec.tensor_for(il)) {
cur = ggml_add(ctx0, cur, layer_dir);
}
cb(cur, "l_out", il);

// input for next layer
Expand Down Expand Up @@ -11054,13 +11124,9 @@ struct llm_build_context {
// add together residual + FFN + self-attention
cur = ggml_add(ctx0, cur, inpL);
cur = ggml_add(ctx0, cur, attn_out);

// add control vector
ggml_tensor * layer_dir = lctx.cvec.tensor_for(il);
if (layer_dir != nullptr) {
if (ggml_tensor * layer_dir = lctx.cvec.tensor_for(il)) {
cur = ggml_add(ctx0, cur, layer_dir);
}

cb(cur, "l_out", il);

// input for next layer
Expand Down Expand Up @@ -11196,8 +11262,7 @@ struct llm_build_context {
cur = ggml_add(ctx0, cur, ffn_inp);
cb(cur, "ffn_out", il);

ggml_tensor * layer_dir = lctx.cvec.tensor_for(il);
if (layer_dir != nullptr) {
if (ggml_tensor * layer_dir = lctx.cvec.tensor_for(il)) {
cur = ggml_add(ctx0, cur, layer_dir);
}
cb(cur, "l_out", il);
Expand Down Expand Up @@ -11313,8 +11378,14 @@ struct llm_build_context {
cur = ggml_add(ctx0, cur, inpL);
cb(cur, "ffn_out", il);

inpL = ggml_add(ctx0, cur, attn_out);
cb(inpL, "l_out", il);
cur = ggml_add(ctx0, cur, attn_out);
if (ggml_tensor * layer_dir = lctx.cvec.tensor_for(il)) {
cur = ggml_add(ctx0, cur, layer_dir);
}
cb(cur, "l_out", il);

// input for next layer
inpL = cur;
} else {
// attention and ffn are computed sequentially
// x = x + attn(ln1(x))
Expand All @@ -11337,8 +11408,14 @@ struct llm_build_context {
LLM_FFN_GELU, LLM_FFN_SEQ, cb, il);
cb(cur, "ffn_out", il);

inpL = ggml_add(ctx0, cur, ffn_inp);
cb(inpL, "l_out", il);
cur = ggml_add(ctx0, cur, ffn_inp);
if (ggml_tensor * layer_dir = lctx.cvec.tensor_for(il)) {
cur = ggml_add(ctx0, cur, layer_dir);
}
cb(cur, "l_out", il);

// input for next layer
inpL = cur;
}
}

Expand Down Expand Up @@ -11465,8 +11542,7 @@ struct llm_build_context {
cur = ggml_add(ctx0, cur, ffn_out);
cb(cur, "ffn_out", il);

ggml_tensor * layer_dir = lctx.cvec.tensor_for(il);
if (layer_dir != nullptr) {
if (ggml_tensor * layer_dir = lctx.cvec.tensor_for(il)) {
cur = ggml_add(ctx0, cur, layer_dir);
}
cb(cur, "l_out", il);
Expand Down Expand Up @@ -11701,6 +11777,9 @@ struct llm_build_context {
}

cur = ggml_add(ctx0, cur, ffn_inp);
if (ggml_tensor * layer_dir = lctx.cvec.tensor_for(il)) {
cur = ggml_add(ctx0, cur, layer_dir);
}
cb(cur, "l_out", il);

// input for next layer
Expand Down

0 comments on commit f57ebf2

Please sign in to comment.