Skip to content

models : fix graph splits #19866

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

Merged
ggerganov merged 1 commit into from
Feb 24, 2026
+8 −3
Merged

models : fix graph splits #19866

Show file tree
Hide file tree
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
2 changes: 2 additions & 0 deletions src/models/kimi-linear.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -116,6 +116,8 @@ llm_build_kimi_linear::llm_build_kimi_linear(const llama_model & model, const ll
cur = build_norm(inpL, layer.attn_norm, NULL, LLM_NORM_RMS, il);
cb(cur, "attn_norm", il);

ggml_build_forward_expand(gf, cur);

// Check layer type by checking which tensors exist
// KDA layers have ssm_a_log tensor, MLA layers have wkv_a_mqa tensor
bool is_kda = (layer.ssm_a != nullptr);
Expand Down
3 changes: 2 additions & 1 deletion src/models/qwen35.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -29,6 +29,8 @@ llm_build_qwen35::llm_build_qwen35(const llama_model & model, const llm_graph_pa
cur = build_norm(inpL, model.layers[il].attn_norm, nullptr, LLM_NORM_RMS, il);
cb(cur, "attn_norm", il);

ggml_build_forward_expand(gf, cur);

// Determine layer type and build appropriate attention mechanism
if (hparams.is_recurrent(il)) {
// Linear attention layer (gated delta net)
Expand Down Expand Up @@ -269,7 +271,6 @@ ggml_tensor * llm_build_qwen35::build_layer_attn_linear(
cb(state_update_target, "state_update_target", il);

ggml_build_forward_expand(gf, ggml_cpy(ctx0, last_conv_states, state_update_target));
cb(conv_states_all, "conv_states_updated", il);

ggml_tensor * state = build_rs(inp, ssm_states_all, hparams.n_embd_s(), n_seqs);
state = ggml_reshape_4d(ctx0, state, head_v_dim, head_v_dim, num_v_heads, n_seqs);
Expand Down
3 changes: 2 additions & 1 deletion src/models/qwen35moe.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -29,6 +29,8 @@ llm_build_qwen35moe::llm_build_qwen35moe(const llama_model & model, const llm_gr
cur = build_norm(inpL, model.layers[il].attn_norm, nullptr, LLM_NORM_RMS, il);
cb(cur, "attn_norm", il);

ggml_build_forward_expand(gf, cur);

// Determine layer type and build appropriate attention mechanism
if (hparams.is_recurrent(il)) {
// Linear attention layer (gated delta net)
Expand Down Expand Up @@ -269,7 +271,6 @@ ggml_tensor * llm_build_qwen35moe ::build_layer_attn_linear(
cb(state_update_target, "state_update_target", il);

ggml_build_forward_expand(gf, ggml_cpy(ctx0, last_conv_states, state_update_target));
cb(conv_states_all, "conv_states_updated", il);

ggml_tensor * state = build_rs(inp, ssm_states_all, hparams.n_embd_s(), n_seqs);
state = ggml_reshape_4d(ctx0, state, head_v_dim, head_v_dim, num_v_heads, n_seqs);
Expand Down
3 changes: 2 additions & 1 deletion src/models/qwen3next.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,8 @@ llm_build_qwen3next::llm_build_qwen3next(const llama_model & model, const llm_gr
cur = build_norm(inpL, model.layers[il].attn_norm, nullptr, LLM_NORM_RMS, il);
cb(cur, "attn_norm", il);

ggml_build_forward_expand(gf, cur);

// Determine layer type and build appropriate attention mechanism
if (hparams.is_recurrent(il)) {
// Linear attention layer (gated delta net)
Expand Down Expand Up @@ -354,7 +356,6 @@ ggml_tensor * llm_build_qwen3next::build_layer_attn_linear(
cb(state_update_target, "state_update_target", il);

ggml_build_forward_expand(gf, ggml_cpy(ctx0, last_conv_states, state_update_target));
cb(conv_states_all, "conv_states_updated", il);

ggml_tensor * state = build_rs(inp, ssm_states_all, hparams.n_embd_s(), n_seqs);
state = ggml_reshape_4d(ctx0, state, head_v_dim, head_v_dim, num_v_heads, n_seqs);
Expand Down
Loading