tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-06-28 04:15:21 +00:00
Code Issues Packages Projects Releases Wiki Activity

mtmd : Support Pixtral 12B (#13065)

Browse Source 
* add pixtral text model (vision is wip)

* cgraph ok, just missing 2D RoPE

* fix bad rebase

* first working version

* fix problem with img_break token

* support dynamic image size

* update docs

* update test script
This commit is contained in:
Xuan-Son Nguyen
2025-04-23 20:21:59 +02:00
committed by GitHub
parent eb1776b15a
commit ecda2ec4b3
14 changed files with 643 additions and 31 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
gguf-py/gguf/tensor_mapping.py
View File

@ -914,6 +914,7 @@ class TensorNameMap:
"vision_tower.vision_model.embeddings.patch_embedding",
"vpm.embeddings.patch_embedding",
"model.vision_model.embeddings.patch_embedding", # SmolVLM
"vision_tower.patch_conv", # pixtral
),
MODEL_TENSOR.V_ENC_EMBD_POS: (
@ -926,52 +927,65 @@ class TensorNameMap:
"vision_tower.vision_model.encoder.layers.{bid}.self_attn.q_proj",
"vpm.encoder.layers.{bid}.self_attn.q_proj",
"model.vision_model.encoder.layers.{bid}.self_attn.q_proj", # SmolVLM
"vision_tower.transformer.layers.{bid}.attention.q_proj", # pixtral
),
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",
"model.vision_model.encoder.layers.{bid}.self_attn.k_proj", # SmolVLM
"vision_tower.transformer.layers.{bid}.attention.k_proj", # pixtral
),
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",
"model.vision_model.encoder.layers.{bid}.self_attn.v_proj", # SmolVLM
"vision_tower.transformer.layers.{bid}.attention.v_proj", # pixtral
),
MODEL_TENSOR.V_ENC_INPUT_NORM: (
"vision_tower.vision_model.encoder.layers.{bid}.layer_norm1",
"vpm.encoder.layers.{bid}.layer_norm1",
"model.vision_model.encoder.layers.{bid}.layer_norm1", # SmolVLM
"vision_tower.transformer.layers.{bid}.attention_norm", # pixtral
),
MODEL_TENSOR.V_ENC_OUTPUT: (
"vision_tower.vision_model.encoder.layers.{bid}.self_attn.out_proj",
"vpm.encoder.layers.{bid}.self_attn.out_proj",
"model.vision_model.encoder.layers.{bid}.self_attn.out_proj", # SmolVLM
"vision_tower.transformer.layers.{bid}.attention.o_proj", # pixtral
),
MODEL_TENSOR.V_ENC_OUTPUT_NORM: (
"vision_tower.vision_model.encoder.layers.{bid}.layer_norm2",
"vpm.encoder.layers.{bid}.layer_norm2",
"model.vision_model.encoder.layers.{bid}.layer_norm2", # SmolVLM
"vision_tower.transformer.layers.{bid}.ffn_norm", # pixtral
),
MODEL_TENSOR.V_ENC_FFN_UP: (
"vision_tower.vision_model.encoder.layers.{bid}.mlp.fc1",
"vpm.encoder.layers.{bid}.mlp.fc1",
"model.vision_model.encoder.layers.{bid}.mlp.fc2", # SmolVLM, gemma3 (note: name is swapped)
"vision_tower.transformer.layers.{bid}.feed_forward.up_proj", # pixtral
),
MODEL_TENSOR.V_ENC_FFN_GATE: (
"vision_tower.transformer.layers.{bid}.feed_forward.gate_proj", # pixtral
),
MODEL_TENSOR.V_ENC_FFN_DOWN: (
"vision_tower.vision_model.encoder.layers.{bid}.mlp.fc2",
"vpm.encoder.layers.{bid}.mlp.fc2",
"model.vision_model.encoder.layers.{bid}.mlp.fc1", # SmolVLM, gemma3 (note: name is swapped)
"vision_tower.transformer.layers.{bid}.feed_forward.down_proj", # pixtral
),
MODEL_TENSOR.V_PRE_NORM: (
"vision_tower.vision_model.pre_layrnorm",
"vision_tower.ln_pre", # pixtral
),
MODEL_TENSOR.V_POST_NORM: (
@ -1030,6 +1044,10 @@ class TensorNameMap:
MODEL_TENSOR.V_RESMPL_QUERY: (
"resampler.query",
),
MODEL_TENSOR.V_TOK_EMBD_IMG_BREAK: (
"v.token_embd.img_break", # for pixtral, this is a generated vector
),
}
# architecture-specific block mappings