fix(llama.cpp): embed metal file into result binary for darwin #4279

mudler · 2024-11-27T19:04:59Z

Description

This PR fixes #4274

Notes for Reviewers

Yes, I signed my commits.

netlify · 2024-11-27T19:05:41Z

✅ Deploy Preview for localai ready!

Name	Link
🔨 Latest commit	`725eecd`
🔍 Latest deploy log	https://app.netlify.com/sites/localai/deploys/6747be3d528683000898b449
😎 Deploy Preview	https://deploy-preview-4279--localai.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

To edit notification comments on pull requests, go to your Netlify site configuration.

mintyleaf · 2024-11-27T20:25:33Z

@mudler

...
12:24AM DBG [llama-cpp-fallback] llama-cpp variant available
12:24AM DBG Loading GRPC Process: /tmp/localai/backend_data/backend-assets/grpc/llama-cpp-fallback
12:24AM DBG GRPC Service for Phi2 will be running at: '127.0.0.1:64983'
12:24AM DBG GRPC Service state dir: /var/folders/d7/46zkm5yj39nbb6dp9dtrs_d00000gn/T/go-processmanager3344125854
12:24AM DBG GRPC Service Started
12:24AM DBG Wait for the service to start up
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stdout Server listening on 127.0.0.1:64983
12:24AM DBG GRPC Service Ready
12:24AM DBG GRPC: Loading model with options: {state:{NoUnkeyedLiterals:{} DoNotCompare:[] DoNotCopy:[] atomicMessageInfo:<nil>} sizeCache:0 unknownFields:[] Model:phi-2.Q2_K ContextSize:512 Seed:2133683766 NBatch:512 F16Memory:false MLock:false MMap:true VocabOnly:false LowVRAM:false Embeddings:false NUMA:false NGPULayers:99999999 MainGPU: TensorSplit: Threads:8 LibrarySearchPath: RopeFreqBase:0 RopeFreqScale:0 RMSNormEps:0 NGQA:0 ModelFile:/Users/mintyleaf/Projects/work/LocalAI/models/phi-2.Q2_K Device: UseTriton:false ModelBaseName: UseFastTokenizer:false PipelineType: SchedulerType: CUDA:false CFGScale:0 IMG2IMG:false CLIPModel: CLIPSubfolder: CLIPSkip:0 ControlNet: Tokenizer: LoraBase: LoraAdapter: LoraScale:0 NoMulMatQ:false DraftModel: AudioPath: Quantization: GPUMemoryUtilization:0 TrustRemoteCode:false EnforceEager:false SwapSpace:0 MaxModelLen:0 TensorParallelSize:0 LoadFormat: MMProj: RopeScaling: YarnExtFactor:0 YarnAttnFactor:0 YarnBetaFast:0 YarnBetaSlow:0 Type: FlashAttention:false NoKVOffload:false ModelPath:/Users/mintyleaf/Projects/work/LocalAI/models LoraAdapters:[] LoraScales:[]}
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_load_model_from_file: using device Metal (Apple M3) - 5461 MiB free
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_model_loader: loaded meta data with 20 key-value pairs and 325 tensors from /Users/mintyleaf/Projects/work/LocalAI/models/phi-2.Q2_K (version GGUF V3 (latest))
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_model_loader: - kv   0:                       general.architecture str              = phi2
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_model_loader: - kv   1:                               general.name str              = Phi2
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_model_loader: - kv   2:                        phi2.context_length u32              = 2048
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_model_loader: - kv   3:                      phi2.embedding_length u32              = 2560
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_model_loader: - kv   4:                   phi2.feed_forward_length u32              = 10240
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_model_loader: - kv   5:                           phi2.block_count u32              = 32
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_model_loader: - kv   6:                  phi2.attention.head_count u32              = 32
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_model_loader: - kv   7:               phi2.attention.head_count_kv u32              = 32
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_model_loader: - kv   8:          phi2.attention.layer_norm_epsilon f32              = 0.000010
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_model_loader: - kv   9:                  phi2.rope.dimension_count u32              = 32
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_model_loader: - kv  10:                          general.file_type u32              = 10
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_model_loader: - kv  11:               tokenizer.ggml.add_bos_token bool             = false
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_model_loader: - kv  12:                       tokenizer.ggml.model str              = gpt2
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_model_loader: - kv  13:                      tokenizer.ggml.tokens arr[str,51200]   = ["!", "\"", "#", "$", "%", "&", "'", ...
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_model_loader: - kv  14:                  tokenizer.ggml.token_type arr[i32,51200]   = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_model_loader: - kv  15:                      tokenizer.ggml.merges arr[str,50000]   = ["Ġ t", "Ġ a", "h e", "i n", "r e",...
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_model_loader: - kv  16:                tokenizer.ggml.bos_token_id u32              = 50256
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_model_loader: - kv  17:                tokenizer.ggml.eos_token_id u32              = 50256
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_model_loader: - kv  18:            tokenizer.ggml.unknown_token_id u32              = 50256
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_model_loader: - kv  19:               general.quantization_version u32              = 2
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_model_loader: - type  f32:  195 tensors
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_model_loader: - type q2_K:   33 tensors
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_model_loader: - type q3_K:   96 tensors
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_model_loader: - type q6_K:    1 tensors
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_vocab: missing pre-tokenizer type, using: 'default'
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_vocab:                                             
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_vocab: ************************************        
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_vocab: GENERATION QUALITY WILL BE DEGRADED!        
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_vocab: CONSIDER REGENERATING THE MODEL             
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_vocab: ************************************        
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_vocab:                                             
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_vocab: special tokens cache size = 944
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_vocab: token to piece cache size = 0.3151 MB
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: format           = GGUF V3 (latest)
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: arch             = phi2
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: vocab type       = BPE
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: n_vocab          = 51200
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: n_merges         = 50000
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: vocab_only       = 0
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: n_ctx_train      = 2048
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: n_embd           = 2560
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: n_layer          = 32
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: n_head           = 32
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: n_head_kv        = 32
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: n_rot            = 32
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: n_swa            = 0
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: n_embd_head_k    = 80
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: n_embd_head_v    = 80
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: n_gqa            = 1
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: n_embd_k_gqa     = 2560
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: n_embd_v_gqa     = 2560
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: f_norm_eps       = 1.0e-05
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: f_norm_rms_eps   = 0.0e+00
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: f_clamp_kqv      = 0.0e+00
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: f_max_alibi_bias = 0.0e+00
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: f_logit_scale    = 0.0e+00
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: n_ff             = 10240
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: n_expert         = 0
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: n_expert_used    = 0
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: causal attn      = 1
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: pooling type     = 0
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: rope type        = 2
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: rope scaling     = linear
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: freq_base_train  = 10000.0
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: freq_scale_train = 1
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: n_ctx_orig_yarn  = 2048
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: rope_finetuned   = unknown
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: ssm_d_conv       = 0
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: ssm_d_inner      = 0
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: ssm_d_state      = 0
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: ssm_dt_rank      = 0
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: ssm_dt_b_c_rms   = 0
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: model type       = 3B
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: model ftype      = Q2_K - Medium
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: model params     = 2.78 B
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: model size       = 1.09 GiB (3.37 BPW) 
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: general.name     = Phi2
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: BOS token        = 50256 '<|endoftext|>'
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: EOS token        = 50256 '<|endoftext|>'
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: EOT token        = 50256 '<|endoftext|>'
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: UNK token        = 50256 '<|endoftext|>'
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: LF token         = 128 'Ä'
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: EOG token        = 50256 '<|endoftext|>'
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_print_meta: max token length = 256
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_tensors: tensor 'token_embd.weight' (q2_K) (and 0 others) cannot be used with preferred buffer type CPU_AARCH64, using CPU instead
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr ggml_backend_metal_log_allocated_size: allocated buffer, size =  1076.52 MiB, ( 1076.59 /  5461.34)
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_tensors: offloading 32 repeating layers to GPU
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_tensors: offloading output layer to GPU
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_tensors: offloaded 33/33 layers to GPU
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_tensors: Metal_Mapped model buffer size =  1076.51 MiB
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llm_load_tensors:   CPU_Mapped model buffer size =    41.02 MiB
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr .........................................................................................
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_new_context_with_model: n_seq_max     = 1
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_new_context_with_model: n_ctx         = 512
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_new_context_with_model: n_ctx_per_seq = 512
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_new_context_with_model: n_batch       = 512
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_new_context_with_model: n_ubatch      = 512
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_new_context_with_model: flash_attn    = 0
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_new_context_with_model: freq_base     = 10000.0
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_new_context_with_model: freq_scale    = 1
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_new_context_with_model: n_ctx_per_seq (512) < n_ctx_train (2048) -- the full capacity of the model will not be utilized
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr ggml_metal_init: allocating
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr ggml_metal_init: found device: Apple M3
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr ggml_metal_init: picking default device: Apple M3
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr ggml_metal_init: default.metallib not found, loading from source
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr ggml_metal_init: GGML_METAL_PATH_RESOURCES = nil
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr ggml_metal_init: error: could not use bundle path to find ggml-metal.metal, falling back to trying cwd
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr ggml_metal_init: loading 'ggml-metal.metal'
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr ggml_metal_init: error: Error Domain=NSCocoaErrorDomain Code=260 "The file “ggml-metal.metal” couldn’t be opened because there is no such file." UserInfo={NSFilePath=ggml-metal.metal, NSURL=ggml-metal.metal -- file:///private/tmp/localai/backend_data/backend-assets/grpc/, NSUnderlyingError=0x600001becde0 {Error Domain=NSPOSIXErrorDomain Code=2 "No such file or directory"}}
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr ggml_backend_metal_device_init: error: failed to allocate context
12:24AM DBG GRPC(Phi2-127.0.0.1:64983): stderr llama_new_context_with_model: failed to initialize Metal backend
...

mintyleaf · 2024-11-27T20:29:12Z

...and if ggml-metal.metal is put back - we have the same error:

...
12:27AM DBG GRPC(Phi2-127.0.0.1:65077): stderr ggml_metal_init: allocating
12:27AM DBG GRPC(Phi2-127.0.0.1:65077): stderr ggml_metal_init: found device: Apple M3
12:27AM DBG GRPC(Phi2-127.0.0.1:65077): stderr ggml_metal_init: picking default device: Apple M3
12:27AM DBG GRPC(Phi2-127.0.0.1:65077): stderr ggml_metal_init: default.metallib not found, loading from source
12:27AM DBG GRPC(Phi2-127.0.0.1:65077): stderr ggml_metal_init: GGML_METAL_PATH_RESOURCES = nil
12:27AM DBG GRPC(Phi2-127.0.0.1:65077): stderr ggml_metal_init: loading '/private/tmp/localai/backend_data/backend-assets/grpc/ggml-metal.metal'
12:27AM DBG GRPC(Phi2-127.0.0.1:65077): stderr ggml_metal_init: error: Error Domain=MTLLibraryErrorDomain Code=3 "program_source:7:10: fatal error: '../ggml-common.h' file not found
12:27AM DBG GRPC(Phi2-127.0.0.1:65077): stderr #include "../ggml-common.h"
12:27AM DBG GRPC(Phi2-127.0.0.1:65077): stderr          ^~~~~~~~~~~~~~~~~~
12:27AM DBG GRPC(Phi2-127.0.0.1:65077): stderr " UserInfo={NSLocalizedDescription=program_source:7:10: fatal error: '../ggml-common.h' file not found
12:27AM DBG GRPC(Phi2-127.0.0.1:65077): stderr #include "../ggml-common.h"
12:27AM DBG GRPC(Phi2-127.0.0.1:65077): stderr          ^~~~~~~~~~~~~~~~~~
12:27AM DBG GRPC(Phi2-127.0.0.1:65077): stderr }
12:27AM DBG GRPC(Phi2-127.0.0.1:65077): stderr ggml_backend_metal_device_init: error: failed to allocate context
12:27AM DBG GRPC(Phi2-127.0.0.1:65077): stderr llama_new_context_with_model: failed to initialize Metal backend
12:27AM DBG GRPC(Phi2-127.0.0.1:65077): stderr common_init_from_params: failed to create context with model '/Users/mintyleaf/Projects/work/LocalAI/models/phi-2.Q2_K'
...

mintyleaf · 2024-11-27T21:00:36Z

@mudler

LocalAI/backend/cpp/llama/Makefile

Line 35 in 5f688d7

CMAKE_ARGS+=-DGGML_METAL_EMBED_LIBRARY=OFF

Removing this line fixes the whole thing

Signed-off-by: Ettore Di Giacinto <[email protected]>

mudler · 2024-11-27T21:33:23Z

@mudler

LocalAI/backend/cpp/llama/Makefile

Line 35 in 5f688d7

CMAKE_ARGS+=-DGGML_METAL_EMBED_LIBRARY=OFF

Removing this line fixes the whole thing

argh right, good catch! I forgot we explicitly disabled it in the backend Makefile. Thanks for testing

dave-gray101

Can confirm that this works and we're using the embedded metal lib

mudler mentioned this pull request Nov 27, 2024

darwin arm64 regression trying "Example: Build on mac", llama-cpp ''../ggml-common.h' file not found' #4274

Closed

fix(llama.cpp): embed metal file into result binary for darwin

1d391cd

Signed-off-by: Ettore Di Giacinto <[email protected]>

mudler force-pushed the fix/metal_embed branch from f5f73d1 to 1d391cd Compare November 27, 2024 21:32

mudler added the bug Something isn't working label Nov 27, 2024

dave-gray101 self-requested a review November 27, 2024 22:04

dave-gray101 approved these changes Nov 27, 2024

View reviewed changes

dave-gray101 enabled auto-merge (squash) November 27, 2024 22:15

Merge branch 'master' into fix/metal_embed

725eecd

dave-gray101 merged commit cbedf2f into master Nov 28, 2024
31 checks passed

dave-gray101 deleted the fix/metal_embed branch November 28, 2024 04:17

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fix(llama.cpp): embed metal file into result binary for darwin #4279

fix(llama.cpp): embed metal file into result binary for darwin #4279

mudler commented Nov 27, 2024

netlify bot commented Nov 27, 2024 •

edited

Loading

mintyleaf commented Nov 27, 2024

mintyleaf commented Nov 27, 2024

mintyleaf commented Nov 27, 2024

mudler commented Nov 27, 2024

dave-gray101 left a comment

fix(llama.cpp): embed metal file into result binary for darwin #4279

fix(llama.cpp): embed metal file into result binary for darwin #4279

Conversation

mudler commented Nov 27, 2024

netlify bot commented Nov 27, 2024 • edited Loading

✅ Deploy Preview for localai ready!

mintyleaf commented Nov 27, 2024

mintyleaf commented Nov 27, 2024

mintyleaf commented Nov 27, 2024

mudler commented Nov 27, 2024

dave-gray101 left a comment

Choose a reason for hiding this comment

netlify bot commented Nov 27, 2024 •

edited

Loading