Skip to content

mtmd : support InternVL 3 38B and 78B mmproj #13443

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
merged 3 commits into from
May 11, 2025
Merged

mtmd : support InternVL 3 38B and 78B mmproj #13443

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
dcc53b1
Support InternVL 3 38B and 78B mmproj
city96 May 10, 2025
a505b92
Swap norms in clip.cpp
city96 May 10, 2025
814734e
Group variables together
city96 May 10, 2025
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
6 changes: 6 additions & 0 deletions gguf-py/gguf/constants.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -483,7 +483,9 @@ class MODEL_TENSOR(IntEnum):
V_ENC_EMBD_PATCH = auto()
V_ENC_EMBD_POS = auto()
V_ENC_ATTN_Q = auto()
V_ENC_ATTN_Q_NORM = auto()
V_ENC_ATTN_K = auto()
V_ENC_ATTN_K_NORM = auto()
V_ENC_ATTN_V = auto()
V_ENC_INPUT_NORM = auto()
V_ENC_OUTPUT = auto()
Expand Down Expand Up @@ -742,7 +744,9 @@ class MODEL_TENSOR(IntEnum):
MODEL_TENSOR.V_ENC_EMBD_PATCH: "v.patch_embd",
MODEL_TENSOR.V_ENC_EMBD_POS: "v.position_embd",
MODEL_TENSOR.V_ENC_ATTN_Q: "v.blk.{bid}.attn_q",
MODEL_TENSOR.V_ENC_ATTN_Q_NORM: "v.blk.{bid}.attn_q_norm",
MODEL_TENSOR.V_ENC_ATTN_K: "v.blk.{bid}.attn_k",
MODEL_TENSOR.V_ENC_ATTN_K_NORM: "v.blk.{bid}.attn_k_norm",
MODEL_TENSOR.V_ENC_ATTN_V: "v.blk.{bid}.attn_v",
MODEL_TENSOR.V_ENC_INPUT_NORM: "v.blk.{bid}.ln1",
MODEL_TENSOR.V_ENC_OUTPUT: "v.blk.{bid}.attn_out",
Expand Down Expand Up @@ -782,7 +786,9 @@ class MODEL_TENSOR(IntEnum):
MODEL_TENSOR.V_ENC_EMBD_PATCH,
MODEL_TENSOR.V_ENC_EMBD_POS,
MODEL_TENSOR.V_ENC_ATTN_Q,
MODEL_TENSOR.V_ENC_ATTN_Q_NORM,
MODEL_TENSOR.V_ENC_ATTN_K,
MODEL_TENSOR.V_ENC_ATTN_K_NORM,
MODEL_TENSOR.V_ENC_ATTN_V,
MODEL_TENSOR.V_ENC_INPUT_NORM,
MODEL_TENSOR.V_ENC_OUTPUT,
Expand Down
8 changes: 8 additions & 0 deletions gguf-py/gguf/tensor_mapping.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -938,6 +938,10 @@ class TensorNameMap:
"visual.blocks.{bid}.attn.q", # qwen2vl, generated
),

MODEL_TENSOR.V_ENC_ATTN_Q_NORM: (
"vision_tower.vision_model.encoder.layers.{bid}.attn.q_norm", # InternVL
),

MODEL_TENSOR.V_ENC_ATTN_K: (
"vision_tower.vision_model.encoder.layers.{bid}.self_attn.k_proj",
"vpm.encoder.layers.{bid}.self_attn.k_proj",
Expand All @@ -946,6 +950,10 @@ class TensorNameMap:
"visual.blocks.{bid}.attn.k", # qwen2vl, generated
),

MODEL_TENSOR.V_ENC_ATTN_K_NORM: (
"vision_tower.vision_model.encoder.layers.{bid}.attn.k_norm", # InternVL
),

MODEL_TENSOR.V_ENC_ATTN_V: (
"vision_tower.vision_model.encoder.layers.{bid}.self_attn.v_proj",
"vpm.encoder.layers.{bid}.self_attn.v_proj",
Expand Down
2 changes: 2 additions & 0 deletions tools/mtmd/clip-impl.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -53,6 +53,8 @@
#define TN_ATTN_Q "%s.blk.%d.attn_q.%s"
#define TN_ATTN_V "%s.blk.%d.attn_v.%s"
#define TN_ATTN_OUTPUT "%s.blk.%d.attn_out.%s"
#define TN_ATTN_K_NORM "%s.blk.%d.attn_k_norm.%s"
#define TN_ATTN_Q_NORM "%s.blk.%d.attn_q_norm.%s"
#define TN_FFN_DOWN "%s.blk.%d.ffn_down.%s"
#define TN_FFN_GATE "%s.blk.%d.ffn_gate.%s"
#define TN_FFN_UP "%s.blk.%d.ffn_up.%s"
Expand Down
15 changes: 15 additions & 0 deletions tools/mtmd/clip.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -205,6 +205,9 @@ struct clip_layer {
ggml_tensor * o_w = nullptr;
ggml_tensor * o_b = nullptr;

ggml_tensor * k_norm = nullptr;
ggml_tensor * q_norm = nullptr;

// layernorm 1
ggml_tensor * ln_1_w = nullptr;
ggml_tensor * ln_1_b = nullptr;
Expand Down Expand Up @@ -1363,6 +1366,16 @@ struct clip_graph {
Vcur = ggml_add(ctx0, Vcur, layer.v_b);
}

if (layer.q_norm) {
Qcur = build_norm(Qcur, layer.q_norm, NULL, norm_t, eps, il);
cb(Qcur, "Qcur_norm", il);
}

if (layer.k_norm) {
Kcur = build_norm(Kcur, layer.k_norm, NULL, norm_t, eps, il);
cb(Kcur, "Kcur_norm", il);
}

Qcur = ggml_reshape_3d(ctx0, Qcur, d_head, n_head, n_pos);
Kcur = ggml_reshape_3d(ctx0, Kcur, d_head, n_head, n_pos);
Vcur = ggml_reshape_3d(ctx0, Vcur, d_head, n_head, n_pos);
Expand Down Expand Up @@ -1988,6 +2001,8 @@ struct clip_model_loader {
layer.q_w = get_tensor(string_format(TN_ATTN_Q, "v", il, "weight"));
layer.v_w = get_tensor(string_format(TN_ATTN_V, "v", il, "weight"));
layer.o_w = get_tensor(string_format(TN_ATTN_OUTPUT, "v", il, "weight"));
layer.k_norm = get_tensor(string_format(TN_ATTN_K_NORM, "v", il, "weight"), false);
layer.q_norm = get_tensor(string_format(TN_ATTN_Q_NORM, "v", il, "weight"), false);
layer.ln_1_w = get_tensor(string_format(TN_LN_1, "v", il, "weight"), false);
layer.ln_2_w = get_tensor(string_format(TN_LN_2, "v", il, "weight"), false);
layer.ls_1_w = get_tensor(string_format(TN_LS_1, "v", il, "weight"), false); // no bias
Expand Down
Loading