Add Trigger at PR creation #15386

alitariq4589 · 2025-08-18T07:50:11Z

@ggerganov This is a reference PR to #14439 for adding CI with RVV1.0 hardware

In the previously merged PR, the workflow did not have a pull_request flag, so the builds are not getting triggered. I have just added the pull_request flag.

I am working on creating a file that mimics the functionality of the build.yml file, which contains all the tests that can run on RISC-V. Until then, please consider this change so that the project can just build on RISC-V vector hardware for now.

…ctions workflow Removed the legacy .devops/cloud-v-pipeline Jenkins CI configuration and introduced .github/workflows/build-riscv-native.yml for native RISC-V builds using GitHub Actions.

Replace Jenkins Cloud-V pipeline with GitHub Actions RISC-V native build

CISC · 2025-08-18T07:59:00Z

Since it's building on native it would be useful if it ran tests like the other native builds:

llama.cpp/.github/workflows/build.yml

Lines 286 to 290 in b143fbc

    
                 - name: Test 
        
                   id: cmake_test 
        
                   run: | 
        
                     cd build 
        
                     ctest -L main --verbose --timeout 900

CISC · 2025-08-18T08:03:56Z

Adding ccache is probably a good idea too.

alitariq4589 · 2025-08-18T08:06:59Z

Since it's building on native it would be useful if it ran tests like the other native builds:

llama.cpp/.github/workflows/build.yml

Lines 286 to 290 in b143fbc

- name: Test

id: cmake_test

run: |

cd build

ctest -L main --verbose --timeout 900

@CISC

Yes. You are right.

But since the setup is currently running CIs inside the podman container, which is running on a non-root user on RISC-V, the docker execution is not yet successful. I am testing this and will take some time.

I have created this PR since the last PR, with just the building project functionality merged, but without the pull_request flag so it was not triggering at all. So till I have a native build tests ready, merging this will prove useful (check this link for infrastructure details) and will provide insights to what is failing in the current CI so I can add changes in the next PR or improve the infrastructure.

Also, the RISC-V doesn't seem to have full Vulkan support right now, so all tests as written in build.yml may not run successfully but I am working on this.

CISC · 2025-08-18T08:19:15Z

Also, the RISC-V doesn't seem to have full Vulkan support right now, so all tests as written in build.yml may not run successfully but I am working on this.

CPU-only is fine, but looking at how long the build takes I'd say ccache is a must.

CISC · 2025-08-18T10:01:27Z

Just installing ccache is probably not enough as the .ccache folder needs to be persisted across runs, see: https://github.com/hendrikmuhs/ccache-action?tab=readme-ov-file#how-it-works

alitariq4589 · 2025-08-18T13:31:37Z

Just installing ccache is probably not enough as the .ccache folder needs to be persisted across runs, see: https://github.com/hendrikmuhs/ccache-action?tab=readme-ov-file#how-it-works

In case of GitHub-hosted runners, the ccache has to be made to stay consistent, but as this self-hosted runner is the same for every runner, I suppose the ccache is naturally the same for every run. I have added a step to set up a directory of the ccache with a 5 GB cache capacity.

alitariq4589 · 2025-08-18T13:35:09Z

The podman container is not ephemeral and is always up. So the folder should remain consistent in every run. But I have added a step for ccache in the workflow file.

CISC · 2025-08-21T12:16:09Z

Cleanup and fix https://github.com/ggml-org/llama.cpp/actions/runs/17126302903/job/48578719271?pr=15386 and I will merge.

* Changed the CI file to hw * Changed the CI file to hw * Added to sudoers for apt * Removed the clone command and used checkout * Added libcurl * Added gcc-14 * Checking gcc --version * added gcc-14 symlink * added CC and C++ variables * Added the gguf weight * Changed the weights path * Added system specification * Removed white spaces * ci: Replace Jenkins riscv native build Cloud-V pipeline with GitHub Actions workflow Removed the legacy .devops/cloud-v-pipeline Jenkins CI configuration and introduced .github/workflows/build-riscv-native.yml for native RISC-V builds using GitHub Actions. * removed trailing whitespaces * Added the trigger at PR creation * Corrected OS name * Added ccache as setup package * Added ccache for self-hosted runner * Added directory for ccache size storage Co-authored-by: Sigbjørn Skjæret <[email protected]> * Changed the build command and added ccache debug log * Added the base dir for the ccache * Re-trigger CI * Cleanup and refactored ccache steps * Cleanup and refactored ccache steps --------- Co-authored-by: Akif Ejaz <[email protected]> Co-authored-by: Sigbjørn Skjæret <[email protected]>

ixgbe · 2025-08-29T09:43:20Z

.github/workflows/build-riscv-native.yml

          sudo apt-get update || true
          sudo apt-get install -y --no-install-recommends \
                  build-essential \
                  gcc-14-riscv64-linux-gnu \


Hi, Since we are building directly on a RISC-V Linux machine, are we still required to use a cross-compilation toolchain?
It seems that native compilation should be sufficient in this case. Could you help clarify the motivation for using a cross toolchain here? Thnaks you.

I think the cross-compilation was only introduced because we didnt have the hardware for checking the RVV1.0 support on RISC-V. But I did not add this CI, so author may better tell us about the motivation.

I think @ixgbe is referring to the gcc-14-riscv64-linux-gnu package (which comes from your #14439 PR), which usually installs a cross-compiler, but build_essential should already give you a native riscv64 compiler.

Installing gcc-14-riscv64-linux-gnu doesn't install a cross compiler. In fact, this is the native toolchain when the command is executed inside the riscv compute.

Also, the build essential installs 13.2.0 GCC by default, which has a vector intrinsics issue with RISC-V.

Check this issue and comment: #12693 (comment)

CISC · 2025-09-20T11:16:33Z

@alitariq4589 The CI has been failing for a few days now with this error:

sudo: a terminal is required to read the password; either use the -S option to read from standard input or configure an askpass helper
sudo: a password is required

alitariq4589 · 2025-09-20T12:11:03Z

@alitariq4589 The CI has been failing for a few days now with this error:
sudo: a terminal is required to read the password; either use the -S option to read from standard input or configure an askpass helper
sudo: a password is required

Thank you for letting me know. I will resolve this by today.

alitariq4589 · 2025-09-20T13:33:08Z

@CISC I have checked the latest failing CI and this is not the runner that I integrated with my CI. If you have access to the actions settings, there should be a runner named 87a313462bbe in there. But this build is running on ggml-4-x86-cuda-v100 as shown in the following logs.

Current runner version: '2.328.0'
Runner name: 'ggml-4-x86-cuda-v100'
Runner group name: 'ggml-ci'
Machine name: 'ggml-4-x86-cuda-v100'

I don't know if you can assign a label to the runner from inside the repo settings. If you can, then I can change the CI to use that label.

The cause of the issue is that CI workflow has runs-on: self-hosted and i am assuming that ggml-4-x86-cuda-v100 is also self-hosted

If you cannot change the label, then can you check if there is a unique label which is not common in both 87a313462bbe and ggml-4-x86-cuda-v100? e.g. RISC-V or something?

alitariq4589 · 2025-09-20T13:35:19Z

Also, my version of github runner is 2.321.0. I have patched the GitHub Actions version 2.328.0 for RISC-V from upstream. I am working to update all the running instances to use tha,t but it will take some time.

ggerganov · 2025-09-20T13:37:53Z

I don't know if you can assign a label to the runner from inside the repo settings. If you can, then I can change the CI to use that label.

Yes, please use the label RISCV64:

alitariq4589 · 2025-09-21T15:56:24Z

Changed the runner label in #16149 . I think the workflow needs approval to run.

ggerganov · 2025-09-21T15:57:34Z

Thanks, I just opened a PR too: #16150. Will merge that one.

alitariq4589 · 2025-09-21T15:59:19Z

Cool. I will close mine then.

alitariq4589 · 2025-09-26T11:14:03Z

Since it's building on native it would be useful if it ran tests like the other native builds:

llama.cpp/.github/workflows/build.yml

Lines 286 to 290 in b143fbc

- name: Test

id: cmake_test

run: |

cd build

ctest -L main --verbose --timeout 900

@CISC, I am adding complete CI for RISC-V, and I would like to know if it is okay if I add the tests to the build.yml file instead of creating a separate file for RISC-V? (I will delete the old CI file once done)

CISC · 2025-09-26T11:41:59Z

@CISC, I am adding complete CI for RISC-V, and I would like to know if it is okay if I add the tests to the build.yml file instead of creating a separate file for RISC-V? (I will delete the old CI file once done)

Yes, please do.

alitariq4589 · 2025-10-03T13:26:31Z

@CISC There seems to be some issue with the ggml-vocabs magic characters. Two of the CI tests are failing because of this.

Start 14: test-tokenizers-ggml-vocabs
14/35 Test #14: test-tokenizers-ggml-vocabs .......***Failed    1.71 sec
Cloning into '/home/runner/_work/llama.cpp/llama.cpp/models/ggml-vocabs'...
main : reading vocab from: '/home/runner/_work/llama.cpp/llama.cpp/models/ggml-vocabs/PLaMo2/ggml-vocab-plamo2.gguf'
register_backend: registered backend CPU (1 devices)
register_device: registered device CPU (Spacemit(R) X60)
gguf_init_from_file_impl: invalid magic characters: 'vers', expected 'GGUF'
llama_model_load: error loading model: llama_model_loader: failed to load model from /home/runner/_work/llama.cpp/llama.cpp/models/ggml-vocabs/PLaMo2/ggml-vocab-plamo2.gguf
llama_model_load_from_file_impl: failed to load model
main: error: failed to load vocab '/home/runner/_work/llama.cpp/llama.cpp/models/ggml-vocabs/PLaMo2/ggml-vocab-plamo2.gguf'
main : reading vocab from: '/home/runner/_work/llama.cpp/llama.cpp/models/ggml-vocabs/RWKV/ggml-vocab-rwkv-7-world.gguf'
register_backend: registered backend CPU (1 devices)
register_device: registered device CPU (Spacemit(R) X60)
gguf_init_from_file_impl: invalid magic characters: 'vers', expected 'GGUF'
llama_model_load: error loading model: llama_model_loader: failed to load model from /home/runner/_work/llama.cpp/llama.cpp/models/ggml-vocabs/RWKV/ggml-vocab-rwkv-7-world.gguf
llama_model_load_from_file_impl: failed to load model
main: error: failed to load vocab '/home/runner/_work/llama.cpp/llama.cpp/models/ggml-vocabs/RWKV/ggml-vocab-rwkv-7-world.gguf'
main : reading vocab from: '/home/runner/_work/llama.cpp/llama.cpp/models/ggml-vocabs/SPM/ggml-vocab-gemma-3.gguf'
register_backend: registered backend CPU (1 devices)
register_device: registered device CPU (Spacemit(R) X60)
gguf_init_from_file_impl: invalid magic characters: 'vers', expected 'GGUF'
llama_model_load: error loading model: llama_model_loader: failed to load model from /home/runner/_work/llama.cpp/llama.cpp/models/ggml-vocabs/SPM/ggml-vocab-gemma-3.gguf
llama_model_load_from_file_impl: failed to load model
main: error: failed to load vocab '/home/runner/_work/llama.cpp/llama.cpp/models/ggml-vocabs/SPM/ggml-vocab-gemma-3.gguf'
main : reading vocab from: '/home/runner/_work/llama.cpp/llama.cpp/models/ggml-vocabs/UGM/ggml-vocab-nomic-bert-moe.gguf'
register_backend: registered backend CPU (1 devices)
register_device: registered device CPU (Spacemit(R) X60)
gguf_init_from_file_impl: invalid magic characters: 'vers', expected 'GGUF'
llama_model_load: error loading model: llama_model_loader: failed to load model from /home/runner/_work/llama.cpp/llama.cpp/models/ggml-vocabs/UGM/ggml-vocab-nomic-bert-moe.gguf
llama_model_load_from_file_impl: failed to load model
main: error: failed to load vocab '/home/runner/_work/llama.cpp/llama.cpp/models/ggml-vocabs/UGM/ggml-vocab-nomic-bert-moe.gguf'
main : reading vocab from: '/home/runner/_work/llama.cpp/llama.cpp/models/ggml-vocabs/WPM/ggml-vocab-jina-v2-en.gguf'
register_backend: registered backend CPU (1 devices)
register_device: registered device CPU (Spacemit(R) X60)
gguf_init_from_file_impl: invalid magic characters: 'vers', expected 'GGUF'
llama_model_load: error loading model: llama_model_loader: failed to load model from /home/runner/_work/llama.cpp/llama.cpp/models/ggml-vocabs/WPM/ggml-vocab-jina-v2-en.gguf
llama_model_load_from_file_impl: failed to load model
main: error: failed to load vocab '/home/runner/_work/llama.cpp/llama.cpp/models/ggml-vocabs/WPM/ggml-vocab-jina-v2-en.gguf'

Check this test.

Is this something expected? This seems to be the hugging face repo. Should I create an issue for this?

CISC · 2025-10-03T13:38:04Z

@CISC There seems to be some issue with the ggml-vocabs magic characters. Two of the CI tests are failing because of this.

Is this something expected? This seems to be the hugging face repo. Should I create an issue for this?

Not sure why it's vers for you? The original files are little endian, but automatically byteswapped for s390x, see #15925, should probably be made to work for all big endian platforms.

alitariq4589 · 2025-10-03T13:43:43Z

@CISC There seems to be some issue with the ggml-vocabs magic characters. Two of the CI tests are failing because of this.
Is this something expected? This seems to be the hugging face repo. Should I create an issue for this?

Not sure why it's vers for you? The original files are little endian, but automatically byteswapped for s390x, see #15925, should probably be made to work for all big endian platforms.

Most of the RISC-V chips are little endian and I am assuming k1 (chip on the RISC-V board which I am using for builds) should also be little endian. Is this something directly related to llama.cpp or is this parsed in some other repository?

CISC · 2025-10-03T13:48:45Z

Not sure why it's vers for you? The original files are little endian, but automatically byteswapped for s390x, see #15925, should probably be made to work for all big endian platforms.

Most of the RISC-V chips are little endian and I am assuming k1 (chip on the RISC-V board which I am using for builds) should also be little endian. Is this something directly related to llama.cpp or is this parsed in some other repository?

I wonder, vers suggests maybe you are not getting the GGUFs at all, but some other content?

alitariq4589 · 2025-10-03T14:02:09Z

Not sure why it's vers for you? The original files are little endian, but automatically byteswapped for s390x, see #15925, should probably be made to work for all big endian platforms.

Most of the RISC-V chips are little endian and I am assuming k1 (chip on the RISC-V board which I am using for builds) should also be little endian. Is this something directly related to llama.cpp or is this parsed in some other repository?

I wonder, vers suggests maybe you are not getting the GGUFs at all, but some other content?

I don't yet understand how this works. I will have to look closely at this CI script to determine which model/file is causing the incorrect magic header.

CISC · 2025-10-03T14:06:19Z

Not sure why it's vers for you? The original files are little endian, but automatically byteswapped for s390x, see #15925, should probably be made to work for all big endian platforms.

Most of the RISC-V chips are little endian and I am assuming k1 (chip on the RISC-V board which I am using for builds) should also be little endian. Is this something directly related to llama.cpp or is this parsed in some other repository?

I wonder, vers suggests maybe you are not getting the GGUFs at all, but some other content?

I don't yet understand how this works. I will have to look closely at this CI script to determine which model/file is causing the incorrect magic header.

It looks like all of them, downloaded from here: https://huggingface.co/ggml-org/vocabs/tree/main

alitariq4589 · 2025-10-09T13:38:35Z

Not sure why it's vers for you? The original files are little endian, but automatically byteswapped for s390x, see #15925, should probably be made to work for all big endian platforms.

Most of the RISC-V chips are little endian and I am assuming k1 (chip on the RISC-V board which I am using for builds) should also be little endian. Is this something directly related to llama.cpp or is this parsed in some other repository?

I wonder, vers suggests maybe you are not getting the GGUFs at all, but some other content?

I don't yet understand how this works. I will have to look closely at this CI script to determine which model/file is causing the incorrect magic header.

It looks like all of them, downloaded from here: https://huggingface.co/ggml-org/vocabs/tree/main

An update on this bug...

There seems to be an issue with how the test runs on the fork. I ran the CI test on the master of my fork (with no added changes, so it stays synced with fork) on my x86 work laptop and got the following error:

      Start 14: test-tokenizers-ggml-vocabs
14/35 Test #14: test-tokenizers-ggml-vocabs .......***Failed    0.56 sec
Already up to date.
main : reading vocab from: '/home/user0/.WORKDIR/llama.cpp/models/ggml-vocabs/PLaMo2/ggml-vocab-plamo2.gguf'
register_backend: registered backend CPU (1 devices)
register_device: registered device CPU (11th Gen Intel(R) Core(TM) i5-1135G7 @ 2.40GHz)
gguf_init_from_file_impl: invalid magic characters: 'vers', expected 'GGUF'
llama_model_load: error loading model: llama_model_loader: failed to load model from /home/user0/.WORKDIR/llama.cpp/models/ggml-vocabs/PLaMo2/ggml-vocab-plamo2.gguf
llama_model_load_from_file_impl: failed to load model
main: error: failed to load vocab '/home/user0/.WORKDIR/llama.cpp/models/ggml-vocabs/PLaMo2/ggml-vocab-plamo2.gguf'
main : reading vocab from: '/home/user0/.WORKDIR/llama.cpp/models/ggml-vocabs/RWKV/ggml-vocab-rwkv-7-world.gguf'
register_backend: registered backend CPU (1 devices)
register_device: registered device CPU (11th Gen Intel(R) Core(TM) i5-1135G7 @ 2.40GHz)
gguf_init_from_file_impl: invalid magic characters: 'vers', expected 'GGUF'
llama_model_load: error loading model: llama_model_loader: failed to load model from /home/user0/.WORKDIR/llama.cpp/models/ggml-vocabs/RWKV/ggml-vocab-rwkv-7-world.gguf
llama_model_load_from_file_impl: failed to load model
main: error: failed to load vocab '/home/user0/.WORKDIR/llama.cpp/models/ggml-vocabs/RWKV/ggml-vocab-rwkv-7-world.gguf'
main : reading vocab from: '/home/user0/.WORKDIR/llama.cpp/models/ggml-vocabs/SPM/ggml-vocab-gemma-3.gguf'
register_backend: registered backend CPU (1 devices)
register_device: registered device CPU (11th Gen Intel(R) Core(TM) i5-1135G7 @ 2.40GHz)
gguf_init_from_file_impl: invalid magic characters: 'vers', expected 'GGUF'
llama_model_load: error loading model: llama_model_loader: failed to load model from /home/user0/.WORKDIR/llama.cpp/models/ggml-vocabs/SPM/ggml-vocab-gemma-3.gguf
llama_model_load_from_file_impl: failed to load model
main: error: failed to load vocab '/home/user0/.WORKDIR/llama.cpp/models/ggml-vocabs/SPM/ggml-vocab-gemma-3.gguf'
main : reading vocab from: '/home/user0/.WORKDIR/llama.cpp/models/ggml-vocabs/UGM/ggml-vocab-nomic-bert-moe.gguf'
register_backend: registered backend CPU (1 devices)
register_device: registered device CPU (11th Gen Intel(R) Core(TM) i5-1135G7 @ 2.40GHz)
gguf_init_from_file_impl: invalid magic characters: 'vers', expected 'GGUF'
llama_model_load: error loading model: llama_model_loader: failed to load model from /home/user0/.WORKDIR/llama.cpp/models/ggml-vocabs/UGM/ggml-vocab-nomic-bert-moe.gguf
llama_model_load_from_file_impl: failed to load model
main: error: failed to load vocab '/home/user0/.WORKDIR/llama.cpp/models/ggml-vocabs/UGM/ggml-vocab-nomic-bert-moe.gguf'
main : reading vocab from: '/home/user0/.WORKDIR/llama.cpp/models/ggml-vocabs/WPM/ggml-vocab-jina-v2-en.gguf'
register_backend: registered backend CPU (1 devices)
register_device: registered device CPU (11th Gen Intel(R) Core(TM) i5-1135G7 @ 2.40GHz)
gguf_init_from_file_impl: invalid magic characters: 'vers', expected 'GGUF'
llama_model_load: error loading model: llama_model_loader: failed to load model from /home/user0/.WORKDIR/llama.cpp/models/ggml-vocabs/WPM/ggml-vocab-jina-v2-en.gguf
llama_model_load_from_file_impl: failed to load model
main: error: failed to load vocab '/home/user0/.WORKDIR/llama.cpp/models/ggml-vocabs/WPM/ggml-vocab-jina-v2-en.gguf'

I dont see this while running on the upstream repo

CISC · 2025-10-09T13:41:47Z

It looks like all of them, downloaded from here: https://huggingface.co/ggml-org/vocabs/tree/main

An update on this bug...

There seems to be an issue with how the test runs on the fork. I ran the CI test on the master of my fork (with no added changes, so it stays synced with fork) on my x86 work laptop and got the following error:

Ahhh, I know, you don't have git-lfs installed. :)

alitariq4589 · 2025-10-09T15:41:27Z

I have checked it. It is installed.

alitariq4589 · 2025-10-09T15:42:11Z

$ sudo apt install git-lfs -y
Reading package lists... Done
Building dependency tree... Done
Reading state information... Done
git-lfs is already the newest version (3.0.2-1ubuntu0.3+esm1).
0 upgraded, 0 newly installed, 0 to remove and 27 not upgraded.

CISC · 2025-10-09T18:56:20Z

I have checked it. It is installed.

Ok, but it is definitely a git-lfs issue, because what you're seeing is the first 4 bytes of
version https://git-lfs.github.com/spec/v1

So, for some reason LFS files are not resolved.

ggerganov · 2025-10-10T08:13:29Z

Not sure if this is relevant, but git-lfs requires you to run the following command once after installing the package:

git lfs install

alitariq4589 · 2025-10-13T10:06:40Z

Yes, that was right. It was git-lfs issue. All the tests are now passing except one ggml-ci-riscv64-native-cpu-high-perf.

I think it is because of the missing Python transformers module. I am installing it. If this is not installable, then I think I will have to add the condition in the test scripts to convert the weights to the appropriate format and then transfer to RISC-V for execution.

alitariq4589 and others added 19 commits June 29, 2025 13:41

Changed the CI file to hw

c24dc3a

Changed the CI file to hw

c465ebe

Added to sudoers for apt

b599746

Removed the clone command and used checkout

9342a37

Added libcurl

bb19818

Added gcc-14

b402426

Checking gcc --version

e608c00

added gcc-14 symlink

5857204

added CC and C++ variables

48fc7c5

Added the gguf weight

d8c9241

Changed the weights path

fd500f7

Added system specification

95f4d8e

Removed white spaces

fc453a0

Merge branch 'master' into alitariq4589/rvv_ci

96d1aa6

ci: Replace Jenkins riscv native build Cloud-V pipeline with GitHub A…

c1d61aa

…ctions workflow Removed the legacy .devops/cloud-v-pipeline Jenkins CI configuration and introduced .github/workflows/build-riscv-native.yml for native RISC-V builds using GitHub Actions.

Merge pull request #2 from alitariq4589/ae/rvv-native-build

d70b2cd

Replace Jenkins Cloud-V pipeline with GitHub Actions RISC-V native build

removed trailing whitespaces

4d21e10

Added the trigger at PR creation

1cc9920

Corrected OS name

c9bb340

github-actions bot added the devops improvements to build systems and github actions label Aug 18, 2025

Merge branch 'master' into alitariq4589/rvv_ci

767c0ad

Added ccache as setup package

304cee2

Added ccache for self-hosted runner

e58e012

alitariq4589 added 2 commits August 21, 2025 17:42

Cleanup and refactored ccache steps

71372a6

Cleanup and refactored ccache steps

7344565

CISC merged commit 029bb39 into ggml-org:master Aug 21, 2025
3 checks passed

ixgbe reviewed Aug 29, 2025

View reviewed changes

ggerganov mentioned this pull request Sep 21, 2025

ci : add label for the RISC-V runner #16150

Merged

Add Trigger at PR creation #15386

Add Trigger at PR creation #15386

Uh oh!

Conversation

alitariq4589 commented Aug 18, 2025

Uh oh!

CISC commented Aug 18, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

CISC commented Aug 18, 2025

Uh oh!

alitariq4589 commented Aug 18, 2025

Uh oh!

CISC commented Aug 18, 2025

Uh oh!

CISC commented Aug 18, 2025

Uh oh!

alitariq4589 commented Aug 18, 2025

Uh oh!

alitariq4589 commented Aug 18, 2025

Uh oh!

CISC commented Aug 21, 2025

Uh oh!

Uh oh!

ixgbe Aug 29, 2025

Choose a reason for hiding this comment

Uh oh!

alitariq4589 Aug 29, 2025

Choose a reason for hiding this comment

Uh oh!

CISC Aug 29, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

alitariq4589 Aug 29, 2025

Choose a reason for hiding this comment

Uh oh!

CISC commented Sep 20, 2025

Uh oh!

alitariq4589 commented Sep 20, 2025

Uh oh!

alitariq4589 commented Sep 20, 2025

Uh oh!

alitariq4589 commented Sep 20, 2025

Uh oh!

ggerganov commented Sep 20, 2025

Uh oh!

alitariq4589 commented Sep 21, 2025

Uh oh!

ggerganov commented Sep 21, 2025

Uh oh!

alitariq4589 commented Sep 21, 2025

Uh oh!

alitariq4589 commented Sep 26, 2025

Uh oh!

CISC commented Sep 26, 2025

Uh oh!

alitariq4589 commented Oct 3, 2025

Uh oh!

CISC commented Oct 3, 2025

Uh oh!

alitariq4589 commented Oct 3, 2025

Uh oh!

CISC commented Oct 3, 2025

Uh oh!

alitariq4589 commented Oct 3, 2025

Uh oh!

CISC commented Oct 3, 2025

Uh oh!

alitariq4589 commented Oct 9, 2025

Uh oh!

CISC commented Oct 9, 2025

Uh oh!

alitariq4589 commented Oct 9, 2025

Uh oh!

alitariq4589 commented Oct 9, 2025

Uh oh!

CISC commented Oct 9, 2025

Uh oh!

ggerganov commented Oct 10, 2025

CISC commented Aug 18, 2025 •

edited

Loading

CISC Aug 29, 2025 •

edited

Loading