Add support for CogVLM model by Tianyue-Zhao · Pull Request #15002 · ggml-org/llama.cpp

Tianyue-Zhao · 2025-08-01T01:50:42Z

This addresses the requests for CogVLM in #4387 and #4350.
CogVLM is a pretty popular model that now adds in cleanly after the recent additions to libmtmd.
I've converted a GGUF here: Link to GGUF files

Sample command and output:

build/bin/llama-mtmd-cli -m ../cogvlm-chat-hf/cogvlm-13B-chat-v1.1-F16.gguf --mmproj ../cogvlm-chat-hf/mmproj-cogvlm-chat-hf --image ./community.png --chat-template vicuna -p "Describe the picture"

load_hparams: model size:         8448.53 MiB
load_hparams: metadata size:      0.36 MiB
alloc_compute_meta:        CPU compute buffer size =   142.02 MiB
main: loading model: ../cogvlm-chat-hf/cogvlm-13B-chat-v1.1-F16.gguf
encoding image slice...
image slice encoded in 16135 ms
decoding image batch 1/1, n_tokens_batch = 1227
image decoded (batch 1/1) in 54065 ms

1. The image showcases a futuristic urban landscape with a mix of architectural styles. The buildings are multi-storied and have a combination of traditional and modern elements. There's a prominent tree in the foreground, suggesting a blend of nature and urban development. The scene appears to be bustling with activity, with various signs and billboards, indicating commercial or residential zones.


llama_perf_context_print:        load time =  108969.65 ms
llama_perf_context_print: prompt eval time =   85229.27 ms /  1241 tokens (   68.68 ms per token,    14.56 tokens per second)
llama_perf_context_print:        eval time =   19843.15 ms /    83 runs   (  239.07 ms per token,     4.18 tokens per second)
llama_perf_context_print:       total time =  126951.23 ms /  1324 tokens
llama_perf_context_print:    graphs reused =          0

I have read the contributing guidelines
Self-reported review complexity:
- Low
- Medium
- High

Tianyue-Zhao · 2025-08-02T16:47:30Z

I think I've fixed the typecheck and format check workflows that were failing before, can someone approve the workflows to run again?
Also, is there a way to run these Github workflows locally or without needing approval from a reviewer?
It would be good to run these CI/CD checks myself before posting the PR.

CISC · 2025-08-02T18:36:51Z

Also, is there a way to run these Github workflows locally or without needing approval from a reviewer? It would be good to run these CI/CD checks myself before posting the PR.

You can run flake8, pyright and editorconfig locally (or via IDE plugins), the build tests can be run manually with ctest.

CISC

This is not a complete review as I don't know enough about mtmd, just commenting...

convert_hf_to_gguf.py

src/llama-model.cpp

Tianyue-Zhao · 2025-08-06T00:12:09Z

Also, is there a way to run these Github workflows locally or without needing approval from a reviewer? It would be good to run these CI/CD checks myself before posting the PR.

You can run flake8, pyright and editorconfig locally (or via IDE plugins), the build tests can be run manually with ctest.

Thanks for the info! That's something I've been wondering about for a while.

src/llama-model.cpp

tools/mtmd/clip.cpp

CISC

Further refinement (merge cont+reshape).

src/llama-model.cpp

tools/mtmd/clip.cpp

CISC · 2025-08-29T11:19:10Z

Further refinement (merge cont+reshape).

After #15662 we can avoid these altogether and just create 3D views.

CISC · 2025-09-08T09:43:31Z

Further refinement (merge cont+reshape).

After #15662 we can avoid these altogether and just create 3D views.

Merged, rebase and apply updated suggestions.

Tianyue-Zhao · 2025-09-09T00:14:23Z

Further refinement (merge cont+reshape).

After #15662 we can avoid these altogether and just create 3D views.

Merged, rebase and apply updated suggestions.

Thanks for the reminder, I've rebased it and removed the extra ggml_cont calls.

ngxson · 2025-10-29T20:53:33Z

sorry I missed the notification to review this. will have a look & push commits to resolve the conflicts

ngxson

I don't have enough VRAM to test the model right now, but I think the code should be good to merge (after CI passed)

Feel free to give it a try even after the PR is merged. In case there are bugs, we can make follow-up PRs to fix it.

ngxson · 2025-10-30T09:55:56Z

No idea why the ASAN test failed, probably just a random runtime issue. I'm re-running the CI

CISC · 2025-10-30T10:00:31Z

No idea why the ASAN test failed, probably just a random runtime issue. I'm re-running the CI

It's ccache getting poisoned somehow, I've yet to track down the reason, when it happens you have to find and delete all the caches on the branch + master (I've just deleted the master ones) and rerun.

ngxson · 2025-10-30T23:25:31Z

btw @Tianyue-Zhao , it seems like this implementation still use the legacy llava preprocessing and does not support dynamic resolution. is this expected?

* model : Granite docling + Idefics3 preprocessing (SmolVLM) (ggml-org#16206) * feat: Add granite-docling conversion using trillion pretokenizer Branch: gabe-l-hart/GraniteDocling Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * feat: Add granite-docling vocab pre enum Branch: gabe-l-hart/GraniteDocling Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * fix: Use granite-docling pre Branch: gabe-l-hart/GraniteDocling Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * feat: Add clip_is_idefics3 Branch: gabe-l-hart/GraniteDocling Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * feat: Allow multi-token boundary sequences for image templating Branch: gabe-l-hart/GraniteDocling Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * feat: Add tiling support for idefices3 in clip.cpp This should likely be moved into llava_uhd::get_slice_instructions, but for now this avoids disrupting the logic there. Branch: gabe-l-hart/GraniteDocling Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * feat: Partial support for full templating for idefics3 in mtmd There are still errors encoding some of the image chunks, but the token sequence now matches transformers _almost_ perfectly, except for the double newline before the global image which shows up as two consecutive newline tokens instead of a single double-newline token. I think this is happening because the blocks are tokenized separately then concatenated. Branch: gabe-l-hart/GraniteDocling Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * feat: Fully working image preprocessing for idefics3 w/ resize and slicing Branch: gabe-l-hart/GraniteDocling Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * feat: Parse the preprocessor config's longest side and add it to the mmproj hparams Branch: GraniteDocling Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * fix: Use the longest side instead of size * scale_factor For Granite Docling, these come out to the same value, but that was just a conicidence. Branch: GraniteDocling Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * fix: Allow batch encoding and remove clip_is_idefics3 Branch: GraniteDocling Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * refactor: Remove unnecessary conditionals for empty token vectors Branch: GraniteDocling Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * refactor: Use image_manipulation util Branch: GraniteDocling Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * add test model --------- Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> Co-authored-by: Xuan Son Nguyen <son@huggingface.co> # Conflicts: # convert_hf_to_gguf.py # convert_hf_to_gguf_update.py # gguf-py/gguf/constants.py # gguf-py/gguf/gguf_writer.py # src/llama-vocab.cpp # src/llama-vocab.h * mtmd : support home-cooked Mistral Small Omni (ggml-org#14928) * model : add LightOnOCR-1B model (ggml-org#16764) * model : add LightOnOCR-1B model * add test # Conflicts: # convert_hf_to_gguf.py # gguf-py/gguf/constants.py * mtmd : fix idefics3 preprocessing (ggml-org#16806) * mtmd : fix idefics3 preprocessing * disable granite test * fix test for granite * model: Add support for CogVLM model (ggml-org#15002) * Added GGUF mappings for CogVLM model * Add tensor mapping for CogVLM visual encoder * Add CogVLM to conversion script, no vision part yet * Added CogVLM vision model to conversion script * Add graph for CogVLM CLIP model * Add graph for CogVLM * Fixes for CogVLM. Now compiles. * Model now runs * Fixes for cogvlm graph * Account for graph context change after rebase * Changes for whitespace * Changes in convert script according to comments * Switch CogVLM LLM graph to merged QKV tensor * Use rope_type variable instead of direct definition * Change CogVLM CLIP encoder to use SWIGLU * Switch CogVLM CLIP to use merged QKV * Apply rebase edits and remove ggml_cont call that is now unnecessary * clean up --------- Co-authored-by: Xuan Son Nguyen <son@huggingface.co> # Conflicts: # convert_hf_to_gguf.py # examples/mtmd/clip.cpp # gguf-py/gguf/constants.py # gguf-py/gguf/tensor_mapping.py # src/llama-arch.cpp # src/llama-arch.h # src/llama-model.cpp # src/llama-model.h * mtmd: refactor preprocessing + support max/min pixels (ggml-org#16878) * mtmd: refactor preprocessing + support max/min pixels * fix mlp type * implement mix/max pixels * improve hparams * better image preproc for qwen * fix * fix out of bound composite * fix (2) * fix token calculation * get_merge_kernel_size() * fix llama4 and lfm2 * gonna fix them all * use simple resize for qwen * qwen: increase min tokens * no resize if dst size == src size * restore to initial min/max tokens value for qwen # Conflicts: # examples/mtmd/clip.cpp * clip : use FA (ggml-org#16837) * clip : use FA * cont : add warning about unsupported ops * implement "auto" mode for clip flash attn * clip : print more detailed op support info during warmup * cont : remove obsolete comment [no ci] * improve debugging message * trailing space * metal : remove stray return --------- Co-authored-by: Xuan Son Nguyen <son@huggingface.co> * model: add Janus Pro for image understanding (ggml-org#16906) * Add support for Janus Pro * Update gguf-py/gguf/tensor_mapping.py Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com> * Update gguf-py/gguf/tensor_mapping.py Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com> * Address reviewer suggestions Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com> * Add JANUS_PRO constant * Update clip model handling Co-authored-by: Xuan-Son Nguyen <son@huggingface.co> * Update tools/mtmd/clip.cpp Co-authored-by: Xuan-Son Nguyen <thichthat@gmail.com> * Refactor JANUS_PRO handling in clip.cpp Co-authored-by: Xuan-Son Nguyen <son@huggingface.co> * Update tools/mtmd/clip.cpp Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com> * em whitespace --------- Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com> Co-authored-by: Xuan-Son Nguyen <son@huggingface.co> Co-authored-by: Xuan-Son Nguyen <thichthat@gmail.com> # Conflicts: # convert_hf_to_gguf.py # gguf-py/gguf/constants.py # gguf-py/gguf/tensor_mapping.py * mtmd: pad mask for qwen2.5vl (ggml-org#16954) * mtmd: pad mask for qwen2.5vl * improve * mtmd: add --image-min/max-tokens (ggml-org#16921) * mtmd: improve struct initialization (ggml-org#16981) * mtmd: allow QwenVL to process larger image by default (ggml-org#17020) * Disable flash attention * mtmd : fix embedding size for image input (ggml-org#17123) * mtmd: fix patch_size initialized to random value in audio models (ggml-org#17128) * mtmd: fix patch_size initialized to random value in audio models * add default hparams * add llama_model_n_embd_inp * Fix load qwen3 vl Change batch size * Add description * Fix cli build error --------- Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> Co-authored-by: Gabe Goodhart <ghart@us.ibm.com> Co-authored-by: Xuan Son Nguyen <son@huggingface.co> Co-authored-by: Tianyue-Zhao <zhaotianyue@outlook.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Zhiyong Wang <85110830+ravenouse@users.noreply.github.com> Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com> Co-authored-by: Xuan-Son Nguyen <thichthat@gmail.com> Co-authored-by: firecoperana <firecoperana>

* Added GGUF mappings for CogVLM model * Add tensor mapping for CogVLM visual encoder * Add CogVLM to conversion script, no vision part yet * Added CogVLM vision model to conversion script * Add graph for CogVLM CLIP model * Add graph for CogVLM * Fixes for CogVLM. Now compiles. * Model now runs * Fixes for cogvlm graph * Account for graph context change after rebase * Changes for whitespace * Changes in convert script according to comments * Switch CogVLM LLM graph to merged QKV tensor * Use rope_type variable instead of direct definition * Change CogVLM CLIP encoder to use SWIGLU * Switch CogVLM CLIP to use merged QKV * Apply rebase edits and remove ggml_cont call that is now unnecessary * clean up --------- Co-authored-by: Xuan Son Nguyen <son@huggingface.co>

github-actions bot added examples python python script changes labels Aug 1, 2025

Tianyue-Zhao marked this pull request as ready for review August 1, 2025 02:15

Tianyue-Zhao force-pushed the cogvlm_support branch from 42113d1 to de22157 Compare August 2, 2025 16:45

CISC reviewed Aug 2, 2025

View reviewed changes

convert_hf_to_gguf.py Show resolved Hide resolved

convert_hf_to_gguf.py Outdated Show resolved Hide resolved

convert_hf_to_gguf.py Outdated Show resolved Hide resolved

convert_hf_to_gguf.py Outdated Show resolved Hide resolved

src/llama-model.cpp Outdated Show resolved Hide resolved

Tianyue-Zhao force-pushed the cogvlm_support branch from de22157 to a571d9a Compare August 8, 2025 00:35

CISC reviewed Aug 8, 2025

View reviewed changes

src/llama-model.cpp Outdated Show resolved Hide resolved

tools/mtmd/clip.cpp Outdated Show resolved Hide resolved

Tianyue-Zhao force-pushed the cogvlm_support branch from a571d9a to ac3992d Compare August 10, 2025 23:40

CISC reviewed Aug 18, 2025

View reviewed changes

src/llama-model.cpp Outdated Show resolved Hide resolved

tools/mtmd/clip.cpp Outdated Show resolved Hide resolved

Tianyue-Zhao added 16 commits September 8, 2025 23:48

Added GGUF mappings for CogVLM model

da81af4

Add tensor mapping for CogVLM visual encoder

6de9d16

Add CogVLM to conversion script, no vision part yet

11fac0c

Added CogVLM vision model to conversion script

a3adac1

Add graph for CogVLM CLIP model

3026781

Add graph for CogVLM

183ca2e

Fixes for CogVLM. Now compiles.

ac3e348

Model now runs

0634dc9

Fixes for cogvlm graph

4c080bf

Account for graph context change after rebase

bc3f084

Changes for whitespace

bcbd6ef

Changes in convert script according to comments

c65f5aa

Switch CogVLM LLM graph to merged QKV tensor

00af7ee

Use rope_type variable instead of direct definition

79e5640

Change CogVLM CLIP encoder to use SWIGLU

86a10bc

Switch CogVLM CLIP to use merged QKV

a959a1f

Apply rebase edits and remove ggml_cont call that is now unnecessary

06a0719

Tianyue-Zhao force-pushed the cogvlm_support branch from ac3992d to 06a0719 Compare September 9, 2025 00:13

CISC requested a review from ngxson September 9, 2025 07:59

CISC mentioned this pull request Oct 29, 2025

[model] add support for qwen3vl series #16780

Merged

ngxson added 2 commits October 29, 2025 23:59

Merge branch 'master' into cogvlm_support

f5cb848

clean up

85764bd

ngxson approved these changes Oct 29, 2025

View reviewed changes

ngxson merged commit bacddc0 into ggml-org:master Oct 30, 2025
189 of 192 checks passed

Conversation

Tianyue-Zhao commented Aug 1, 2025

Uh oh!

Tianyue-Zhao commented Aug 2, 2025

Uh oh!

CISC commented Aug 2, 2025

Uh oh!

CISC left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Tianyue-Zhao commented Aug 6, 2025

Uh oh!

Uh oh!

Uh oh!

CISC left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

CISC commented Aug 29, 2025

Uh oh!

CISC commented Sep 8, 2025

Uh oh!

Tianyue-Zhao commented Sep 9, 2025

Uh oh!

ngxson commented Oct 29, 2025

Uh oh!

ngxson left a comment

Choose a reason for hiding this comment

Uh oh!

ngxson commented Oct 30, 2025

Uh oh!

CISC commented Oct 30, 2025

Uh oh!

Uh oh!

ngxson commented Oct 30, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants