Misc. bug: Unable to load model

[VJHack]

Name and Version

./build/bin/llama-cli --version
register_backend: registered backend Metal (1 devices)
register_device: registered device Metal (Intel(R) Iris(TM) Plus Graphics 655)
register_backend: registered backend BLAS (1 devices)
register_device: registered device BLAS (Accelerate)
register_backend: registered backend CPU (1 devices)
register_device: registered device CPU (Intel(R) Core(TM) i5-8279U CPU @ 2.40GHz)
version: 4233 (f68367d2)
built with Apple clang version 14.0.0 (clang-1400.0.29.202) for x86_64-apple-darwin21.6.0

Operating systems

Mac

Which llama.cpp modules do you know to be affected?

Other (Please specify in the next section)

Problem description & steps to reproduce

I am unable to load and run inference on any model. Previously with the make build system, I didn't have any issues. But now that our build system has switched to cmake I've been having this issue shown below. I'm not able to use either llama-cli or llama-server.

I'm running on MacOS with 16 GB of memory and Intel Core i5 processor.

How I'm building:
cmake -B build
cmake --build build --config Release

command:
./build/bin/llama-server -m models/tinyllama-1.1b-chat-v1.0.Q2_K.gguf

Any help would be appreciated!

First Bad Commit

No response

Relevant log output

$ ./build/bin/llama-server -m models/tinyllama-1.1b-chat-v1.0.Q2_K.gguf

register_backend: registered backend Metal (1 devices)
register_device: registered device Metal (Intel(R) Iris(TM) Plus Graphics 655)
register_backend: registered backend BLAS (1 devices)
register_device: registered device BLAS (Accelerate)
register_backend: registered backend CPU (1 devices)
register_device: registered device CPU (Intel(R) Core(TM) i5-8279U CPU @ 2.40GHz)
build: 4180 (c6807b3f) with Apple clang version 14.0.0 (clang-1400.0.29.202) for x86_64-apple-darwin21.6.0 (debug)
system info: n_threads = 4, n_threads_batch = 4, total_threads = 8

system_info: n_threads = 4 (n_threads_batch = 4) / 8 | Metal : EMBED_LIBRARY = 1 | CPU : SSE3 = 1 | SSSE3 = 1 | AVX = 1 | AVX2 = 1 | F16C = 1 | FMA = 1 | LLAMAFILE = 1 | AARCH64_REPACK = 1 | 

main: HTTP server is listening, hostname: 127.0.0.1, port: 8080, http threads: 7
main: loading model
srv    load_model: loading model 'models/tinyllama-1.1b-chat-v1.0.Q2_K.gguf'
llama_load_model_from_file: using device Metal (Intel(R) Iris(TM) Plus Graphics 655) - 1531 MiB free
llama_model_loader: loaded meta data with 23 key-value pairs and 201 tensors from models/tinyllama-1.1b-chat-v1.0.Q2_K.gguf (version GGUF V3 (latest))
llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
llama_model_loader: - kv   0:                       general.architecture str              = llama
llama_model_loader: - kv   1:                               general.name str              = tinyllama_tinyllama-1.1b-chat-v1.0
llama_model_loader: - kv   2:                       llama.context_length u32              = 2048
llama_model_loader: - kv   3:                     llama.embedding_length u32              = 2048
llama_model_loader: - kv   4:                          llama.block_count u32              = 22
llama_model_loader: - kv   5:                  llama.feed_forward_length u32              = 5632
llama_model_loader: - kv   6:                 llama.rope.dimension_count u32              = 64
llama_model_loader: - kv   7:                 llama.attention.head_count u32              = 32
llama_model_loader: - kv   8:              llama.attention.head_count_kv u32              = 4
llama_model_loader: - kv   9:     llama.attention.layer_norm_rms_epsilon f32              = 0.000010
llama_model_loader: - kv  10:                       llama.rope.freq_base f32              = 10000.000000
llama_model_loader: - kv  11:                          general.file_type u32              = 10
llama_model_loader: - kv  12:                       tokenizer.ggml.model str              = llama
llama_model_loader: - kv  13:                      tokenizer.ggml.tokens arr[str,32000]   = ["<unk>", "<s>", "</s>", "<0x00>", "<...
llama_model_loader: - kv  14:                      tokenizer.ggml.scores arr[f32,32000]   = [0.000000, 0.000000, 0.000000, 0.0000...
llama_model_loader: - kv  15:                  tokenizer.ggml.token_type arr[i32,32000]   = [2, 3, 3, 6, 6, 6, 6, 6, 6, 6, 6, 6, ...
llama_model_loader: - kv  16:                      tokenizer.ggml.merges arr[str,61249]   = ["▁ t", "e r", "i n", "▁ a", "e n...
llama_model_loader: - kv  17:                tokenizer.ggml.bos_token_id u32              = 1
llama_model_loader: - kv  18:                tokenizer.ggml.eos_token_id u32              = 2
llama_model_loader: - kv  19:            tokenizer.ggml.unknown_token_id u32              = 0
llama_model_loader: - kv  20:            tokenizer.ggml.padding_token_id u32              = 2
llama_model_loader: - kv  21:                    tokenizer.chat_template str              = {% for message in messages %}\n{% if m...
llama_model_loader: - kv  22:               general.quantization_version u32              = 2
llama_model_loader: - type  f32:   45 tensors
llama_model_loader: - type q2_K:   45 tensors
llama_model_loader: - type q3_K:  110 tensors
llama_model_loader: - type q6_K:    1 tensors
llm_load_vocab: special_eos_id is not in special_eog_ids - the tokenizer config may be incorrect
llm_load_vocab: special tokens cache size = 3
llm_load_vocab: token to piece cache size = 0.1684 MB
llm_load_print_meta: format           = GGUF V3 (latest)
llm_load_print_meta: arch             = llama
llm_load_print_meta: vocab type       = SPM
llm_load_print_meta: n_vocab          = 32000
llm_load_print_meta: n_merges         = 0
llm_load_print_meta: vocab_only       = 0
llm_load_print_meta: n_ctx_train      = 2048
llm_load_print_meta: n_embd           = 2048
llm_load_print_meta: n_layer          = 22
llm_load_print_meta: n_head           = 32
llm_load_print_meta: n_head_kv        = 4
llm_load_print_meta: n_rot            = 64
llm_load_print_meta: n_swa            = 0
llm_load_print_meta: n_embd_head_k    = 64
llm_load_print_meta: n_embd_head_v    = 64
llm_load_print_meta: n_gqa            = 8
llm_load_print_meta: n_embd_k_gqa     = 256
llm_load_print_meta: n_embd_v_gqa     = 256
llm_load_print_meta: f_norm_eps       = 0.0e+00
llm_load_print_meta: f_norm_rms_eps   = 1.0e-05
llm_load_print_meta: f_clamp_kqv      = 0.0e+00
llm_load_print_meta: f_max_alibi_bias = 0.0e+00
llm_load_print_meta: f_logit_scale    = 0.0e+00
llm_load_print_meta: n_ff             = 5632
llm_load_print_meta: n_expert         = 0
llm_load_print_meta: n_expert_used    = 0
llm_load_print_meta: causal attn      = 1
llm_load_print_meta: pooling type     = 0
llm_load_print_meta: rope type        = 0
llm_load_print_meta: rope scaling     = linear
llm_load_print_meta: freq_base_train  = 10000.0
llm_load_print_meta: freq_scale_train = 1
llm_load_print_meta: n_ctx_orig_yarn  = 2048
llm_load_print_meta: rope_finetuned   = unknown
llm_load_print_meta: ssm_d_conv       = 0
llm_load_print_meta: ssm_d_inner      = 0
llm_load_print_meta: ssm_d_state      = 0
llm_load_print_meta: ssm_dt_rank      = 0
llm_load_print_meta: ssm_dt_b_c_rms   = 0
llm_load_print_meta: model type       = 1B
llm_load_print_meta: model ftype      = Q2_K - Medium
llm_load_print_meta: model params     = 1.10 B
llm_load_print_meta: model size       = 459.11 MiB (3.50 BPW) 
llm_load_print_meta: general.name     = tinyllama_tinyllama-1.1b-chat-v1.0
llm_load_print_meta: BOS token        = 1 '<s>'
llm_load_print_meta: EOS token        = 2 '</s>'
llm_load_print_meta: UNK token        = 0 '<unk>'
llm_load_print_meta: PAD token        = 2 '</s>'
llm_load_print_meta: LF token         = 13 '<0x0A>'
llm_load_print_meta: EOG token        = 2 '</s>'
llm_load_print_meta: max token length = 48
llm_load_tensors: offloading 22 repeating layers to GPU
llm_load_tensors: offloading output layer to GPU
llm_load_tensors: offloaded 23/23 layers to GPU
llm_load_tensors: Metal_Mapped model buffer size =   387.33 MiB
llm_load_tensors:   CPU_Mapped model buffer size =   459.10 MiB
...................................................................................
llama_new_context_with_model: n_seq_max     = 1
llama_new_context_with_model: n_ctx         = 4096
llama_new_context_with_model: n_ctx_per_seq = 4096
llama_new_context_with_model: n_batch       = 2048
llama_new_context_with_model: n_ubatch      = 512
llama_new_context_with_model: flash_attn    = 0
llama_new_context_with_model: freq_base     = 10000.0
llama_new_context_with_model: freq_scale    = 1
llama_new_context_with_model: n_ctx_pre_seq (4096) > n_ctx_train (2048) -- possible training context overflow
ggml_metal_init: allocating
ggml_metal_init: found device: Intel(R) Iris(TM) Plus Graphics 655
ggml_metal_init: picking default device: Intel(R) Iris(TM) Plus Graphics 655
ggml_metal_init: using embedded metal library
ggml_metal_init: error: Error Domain=MTLLibraryErrorDomain Code=3 "program_source:56:35: error: unknown type name 'block_q4_0'
void dequantize_q4_0(device const block_q4_0 *xb, short il, thread type4x4 & reg) {
                                  ^
program_source:75:35: error: unknown type name 'block_q4_1'
void dequantize_q4_1(device const block_q4_1 *xb, short il, thread type4x4 & reg) {
                                  ^
program_source:94:35: error: unknown type name 'block_q5_0'
void dequantize_q5_0(device const block_q5_0 *xb, short il, thread type4x4 & reg) {
                                  ^
program_source:126:35: error: unknown type name 'block_q5_1'
void dequantize_q5_1(device const block_q5_1 *xb, short il, thread type4x4 & reg) {
                                  ^


...


ggml_backend_metal_device_init: error: failed to allocate context
llama_new_context_with_model: failed to initialize Metal backend
common_init_from_params: failed to create context with model 'models/tinyllama-1.1b-chat-v1.0.Q2_K.gguf'
main: error: unable to load model

[eugeniosegala]

Could it be related to #10208?

[ggerganov]

Could it be related to #10208?

No. The problem here is trying to run the Metal backend on Intel device which is not supported.

[eugeniosegala]

Thank you @ggerganov.

My setup is:

version: 1 (43ed389)
built with Apple clang version 16.0.0 (clang-1600.0.26.4) for arm64-apple-darwin24.1.0

Is this supported?

As I'm experiencing the same problem.

[ggerganov]

It is supported. Show the commands and the full log that you get.

[eugeniosegala]

Command:

./build/bin/llama-cli -m ./FastLlama-3.2-1B-Instruct.i1-IQ3_XS.gguf -p "I believe the meaning of life is"

Error:

ggml_backend_metal_device_init: error: failed to allocate context
llama_new_context_with_model: failed to initialize Metal backend
common_init_from_params: failed to create context with model './FastLlama-3.2-1B-Instruct.i1-IQ3_XS.gguf'
main: error: unable to load model

However, it works fine when I re-build everything without Metal using -DGGML_METAL=OFF.

[VJHack]

@ggerganov and @eugeniosegala Thanks for the support! Re-building everything without the metal backend worked for me. I'm closing this issue as it is not a bug.

[eugeniosegala]

@VJHack the fact that the Metal build does not work is okay though?

Is it a working in progress feature?

[ggerganov]

Provide more logs: ./build/bin/llama-cli -m ./FastLlama-3.2-1B-Instruct.i1-IQ3_XS.gguf -p "I believe the meaning of life is" -lv 1

[eugeniosegala]

@ggerganov this is what I got back:

debug.log

I got the same with other models too, like llama 3 8b.

[ggerganov]

This should fix it:

rm -rf build

cmake -B build
cmake --build build --config Release

./build/bin/llama-cli -m ./FastLlama-3.2-1B-Instruct.i1-IQ3_XS.gguf -p "I believe the meaning of life is"

[eugeniosegala]

@ggerganov I get the same error.

I have tried to re-build it several times.

P.S: I'm building from latest master

I will try to build it from some previous commits as well.

[ggerganov]

Clone the repo from scratch. You likely have some local modifications.

[eugeniosegala]

@ggerganov you are correct, I had to re-clone everything and re-build again, then it worked!

I did not have local changes or anything but maybe something was cached or similar (I did delete the build folder though).

Thank you!