Misc. bug: Regression with Vulkan backend **terminate called after throwing an instance of 'vk::DeviceLostError'** #17334
Description
Name and Version
Debian 13@6.16.3 llama.cpp version: 7083 (2376b7758)
$ llama-cli --version
ggml_vulkan: Found 1 Vulkan devices:
ggml_vulkan: 0 = Intel(R) Iris(R) Xe Graphics (ADL GT2) (Intel open-source Mesa driver) | uma: 1 | fp16: 1 | bf16: 0 | warp size: 32 | shared memory: 65536 | int dot: 1 | matrix cores: none
version: 7083 (2376b7758)
built with cc (Debian 14.2.0-19) 14.2.0 for x86_64-linux-gnu
seems bit similar to : #13248
Operating systems
Debian GNU/Linux 13@6.16.3
Which llama.cpp modules do you know to be affected?
llama-server
Command line
`$ llama-server -m .cache/llama.cpp/ggml-org_gemma-3n-E4B-it-GGUF_gemma-3n-E4B-it-Q8_0.gguf --no-mmap -ngl 99 -fa 1 --jinja --host 0.0.0.0 --port 8080 -c 8192`Problem description & steps to reproduce
Hi since few build i can spot some bug with Vulkan backend:
terminate called after throwing an instance of 'vk::DeviceLostError'
what(): vk::Queue::submit: ErrorDeviceLostggml_vulkan: Found 1 Vulkan devices:
ggml_vulkan: 0 = Intel(R) Iris(R) Xe Graphics (ADL GT2) (Intel open-source Mesa driver) | uma: 1 | fp16: 1 | bf16: 0 | warp size: 32 | shared memory: 65536 | int dot: 1 | matrix cores: none
main: setting n_parallel = 4 and kv_unified = true (add -kvu to disable this)
build: 7083 (2376b7758) with cc (Debian 14.2.0-19) 14.2.0 for x86_64-linux-gnu
system info: n_threads = 4, n_threads_batch = 4, total_threads = 16
system_info: n_threads = 4 (n_threads_batch = 4) / 16 | CPU : SSE3 = 1 | SSSE3 = 1 | AVX = 1 | AVX_VNNI = 1 | AVX2 = 1 | F16C = 1 | FMA = 1 | BMI2 = 1 | LLAMAFILE = 1 | OPENMP = 1 | REPACK = 1 |
main: binding port with default address family
main: HTTP server is listening, hostname: 0.0.0.0, port: 8080, http threads: 15
main: loading model
srv load_model: loading model '.cache/llama.cpp/ggml-org_gemma-3n-E4B-it-GGUF_gemma-3n-E4B-it-Q8_0.gguf'
llama_model_load_from_file_impl: using device Vulkan0 (Intel(R) Iris(R) Xe Graphics (ADL GT2)) (0000:00:02.0) - 21530 MiB free
llama_model_loader: loaded meta data with 42 key-value pairs and 847 tensors from .cache/llama.cpp/ggml-org_gemma-3n-E4B-it-GGUF_gemma-3n-E4B-it-Q8_0.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 = gemma3n
llama_model_loader: - kv 1: general.type str = model
llama_model_loader: - kv 2: general.size_label str = 6.9B
llama_model_loader: - kv 3: general.license str = gemma
llama_model_loader: - kv 4: general.base_model.count u32 = 1
llama_model_loader: - kv 5: general.base_model.0.name str = Gemma 3n E4B
llama_model_loader: - kv 6: general.base_model.0.organization str = Google
llama_model_loader: - kv 7: general.base_model.0.repo_url str = https://huggingface.co/google/gemma-3...
llama_model_loader: - kv 8: general.tags arr[str,5] = ["automatic-speech-recognition", "aut...
llama_model_loader: - kv 9: gemma3n.context_length u32 = 32768
llama_model_loader: - kv 10: gemma3n.embedding_length u32 = 2048
llama_model_loader: - kv 11: gemma3n.block_count u32 = 35
llama_model_loader: - kv 12: gemma3n.feed_forward_length u32 = 16384
llama_model_loader: - kv 13: gemma3n.attention.head_count u32 = 8
llama_model_loader: - kv 14: gemma3n.attention.layer_norm_rms_epsilon f32 = 0.000001
llama_model_loader: - kv 15: gemma3n.attention.key_length u32 = 256
llama_model_loader: - kv 16: gemma3n.attention.value_length u32 = 256
llama_model_loader: - kv 17: gemma3n.rope.freq_base f32 = 1000000.000000
llama_model_loader: - kv 18: gemma3n.attention.sliding_window u32 = 512
llama_model_loader: - kv 19: gemma3n.attention.head_count_kv u32 = 2
llama_model_loader: - kv 20: gemma3n.altup.active_idx u32 = 0
llama_model_loader: - kv 21: gemma3n.altup.num_inputs u32 = 4
llama_model_loader: - kv 22: gemma3n.embedding_length_per_layer_input u32 = 256
llama_model_loader: - kv 23: gemma3n.attention.shared_kv_layers f32 = 15.000000
llama_model_loader: - kv 24: gemma3n.activation_sparsity_scale arr[f32,35] = [1.644853, 1.644853, 1.644853, 1.6448...
llama_model_loader: - kv 25: gemma3n.attention.sliding_window_pattern arr[bool,35] = [true, true, true, true, false, true,...
llama_model_loader: - kv 26: tokenizer.chat_template str = {{ bos_token }}\n{%- if messages[0]['r...
llama_model_loader: - kv 27: tokenizer.ggml.model str = llama
llama_model_loader: - kv 28: tokenizer.ggml.pre str = default
llama_model_loader: - kv 29: tokenizer.ggml.tokens arr[str,262144] = ["<pad>", "<eos>", "<bos>", "<unk>", ...
llama_model_loader: - kv 30: tokenizer.ggml.scores arr[f32,262144] = [-1000.000000, -1000.000000, -1000.00...
llama_model_loader: - kv 31: tokenizer.ggml.token_type arr[i32,262144] = [3, 3, 3, 3, 3, 4, 3, 3, 3, 3, 3, 3, ...
llama_model_loader: - kv 32: tokenizer.ggml.bos_token_id u32 = 2
llama_model_loader: - kv 33: tokenizer.ggml.eos_token_id u32 = 1
llama_model_loader: - kv 34: tokenizer.ggml.unknown_token_id u32 = 3
llama_model_loader: - kv 35: tokenizer.ggml.padding_token_id u32 = 0
llama_model_loader: - kv 36: tokenizer.ggml.add_bos_token bool = true
llama_model_loader: - kv 37: tokenizer.ggml.add_sep_token bool = false
llama_model_loader: - kv 38: tokenizer.ggml.add_eos_token bool = false
llama_model_loader: - kv 39: tokenizer.ggml.add_space_prefix bool = false
llama_model_loader: - kv 40: general.quantization_version u32 = 2
llama_model_loader: - kv 41: general.file_type u32 = 7
llama_model_loader: - type f32: 422 tensors
llama_model_loader: - type f16: 108 tensors
llama_model_loader: - type q8_0: 317 tensors
print_info: file format = GGUF V3 (latest)
print_info: file type = Q8_0
print_info: file size = 6.84 GiB (8.56 BPW)
load: special_eos_id is not in special_eog_ids - the tokenizer config may be incorrect
load: printing all EOG tokens:
load: - 1 ('<eos>')
load: - 106 ('<end_of_turn>')
load: special tokens cache size = 6414
load: token to piece cache size = 1.9446 MB
print_info: arch = gemma3n
print_info: vocab_only = 0
print_info: n_ctx_train = 32768
print_info: n_embd = 2048
print_info: n_embd_inp = 2048
print_info: n_layer = 35
print_info: n_head = 8
print_info: n_head_kv = 2
print_info: n_rot = 256
print_info: n_swa = 512
print_info: is_swa_any = 1
print_info: n_embd_head_k = 256
print_info: n_embd_head_v = 256
print_info: n_gqa = 4
print_info: n_embd_k_gqa = 512
print_info: n_embd_v_gqa = 512
print_info: f_norm_eps = 0.0e+00
print_info: f_norm_rms_eps = 1.0e-06
print_info: f_clamp_kqv = 0.0e+00
print_info: f_max_alibi_bias = 0.0e+00
print_info: f_logit_scale = 0.0e+00
print_info: f_attn_scale = 1.0e+00
print_info: n_ff = 16384
print_info: n_expert = 0
print_info: n_expert_used = 0
print_info: n_expert_groups = 0
print_info: n_group_used = 0
print_info: causal attn = 1
print_info: pooling type = 0
print_info: rope type = 2
print_info: rope scaling = linear
print_info: freq_base_train = 1000000.0
print_info: freq_scale_train = 1
print_info: n_ctx_orig_yarn = 32768
print_info: rope_finetuned = unknown
print_info: model type = E4B
print_info: model params = 6.87 B
print_info: general.name = n/a
print_info: vocab type = SPM
print_info: n_vocab = 262144
print_info: n_merges = 0
print_info: BOS token = 2 '<bos>'
print_info: EOS token = 1 '<eos>'
print_info: EOT token = 106 '<end_of_turn>'
print_info: UNK token = 3 '<unk>'
print_info: PAD token = 0 '<pad>'
print_info: LF token = 248 '<0x0A>'
print_info: EOG token = 1 '<eos>'
print_info: EOG token = 106 '<end_of_turn>'
print_info: max token length = 48
load_tensors: loading model tensors, this can take a while... (mmap = false)
load_tensors: offloading 35 repeating layers to GPU
load_tensors: offloading output layer to GPU
load_tensors: offloaded 36/36 layers to GPU
load_tensors: Vulkan0 model buffer size = 7006.38 MiB
load_tensors: Vulkan_Host model buffer size = 544.00 MiB
.........................................................
llama_context: constructing llama_context
llama_context: n_seq_max = 4
llama_context: n_ctx = 8192
llama_context: n_ctx_seq = 8192
llama_context: n_batch = 2048
llama_context: n_ubatch = 512
llama_context: causal_attn = 1
llama_context: flash_attn = enabled
llama_context: kv_unified = true
llama_context: freq_base = 1000000.0
llama_context: freq_scale = 1
llama_context: n_ctx_seq (8192) < n_ctx_train (32768) -- the full capacity of the model will not be utilized
llama_context: Vulkan_Host output buffer size = 4.00 MiB
llama_kv_cache_iswa: creating non-SWA KV cache, size = 8192 cells
llama_kv_cache: Vulkan0 KV buffer size = 64.00 MiB
llama_kv_cache: size = 64.00 MiB ( 8192 cells, 4 layers, 4/1 seqs), K (f16): 32.00 MiB, V (f16): 32.00 MiB
llama_kv_cache_iswa: creating SWA KV cache, size = 2560 cells
llama_kv_cache: Vulkan0 KV buffer size = 80.00 MiB
llama_kv_cache: size = 80.00 MiB ( 2560 cells, 16 layers, 4/1 seqs), K (f16): 40.00 MiB, V (f16): 40.00 MiB
llama_context: Vulkan0 compute buffer size = 520.00 MiB
llama_context: Vulkan_Host compute buffer size = 25.03 MiB
llama_context: graph nodes = 3123
llama_context: graph splits = 2
common_init_from_params: added <eos> logit bias = -inf
common_init_from_params: added <end_of_turn> logit bias = -inf
common_init_from_params: setting dry_penalty_last_n to ctx_size = 8192
common_init_from_params: warming up the model with an empty run - please wait ... (--no-warmup to disable)
srv init: initializing slots, n_slots = 4
slot init: id 0 | task -1 | new slot, n_ctx = 8192
slot init: id 1 | task -1 | new slot, n_ctx = 8192
slot init: id 2 | task -1 | new slot, n_ctx = 8192
slot init: id 3 | task -1 | new slot, n_ctx = 8192
srv init: prompt cache is enabled, size limit: 8192 MiB
srv init: use `--cache-ram 0` to disable the prompt cache
srv init: for more info see https://github.com/ggml-org/llama.cpp/pull/16391
srv init: thinking = 0
main: model loaded
main: chat template, chat_template: {{ bos_token }}
{%- if messages[0]['role'] == 'system' -%}
{%- if messages[0]['content'] is string -%}
{%- set first_user_prefix = messages[0]['content'] + '
' -%}
{%- else -%}
{%- set first_user_prefix = messages[0]['content'][0]['text'] + '
' -%}
{%- endif -%}
{%- set loop_messages = messages[1:] -%}
{%- else -%}
{%- set first_user_prefix = "" -%}
{%- set loop_messages = messages -%}
{%- endif -%}
{%- for message in loop_messages -%}
{%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}
{{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }}
{%- endif -%}
{%- if (message['role'] == 'assistant') -%}
{%- set role = "model" -%}
{%- else -%}
{%- set role = message['role'] -%}
{%- endif -%}
{{ '<start_of_turn>' + role + '
' + (first_user_prefix if loop.first else "") }}
{%- if message['content'] is string -%}
{{ message['content'] | trim }}
{%- elif message['content'] is iterable -%}
{%- for item in message['content'] -%}
{%- if item['type'] == 'audio' -%}
{{ '<audio_soft_token>' }}
{%- elif item['type'] == 'image' -%}
{{ '<image_soft_token>' }}
{%- elif item['type'] == 'text' -%}
{{ item['text'] | trim }}
{%- endif -%}
{%- endfor -%}
{%- else -%}
{{ raise_exception("Invalid content type") }}
{%- endif -%}
{{ '<end_of_turn>
' }}
{%- endfor -%}
{%- if add_generation_prompt -%}
{{'<start_of_turn>model
'}}
{%- endif -%}
, example_format: '<start_of_turn>user
You are a helpful assistant
Hello<end_of_turn>
<start_of_turn>model
Hi there<end_of_turn>
<start_of_turn>user
How are you?<end_of_turn>
<start_of_turn>model
'
main: server is listening on http://0.0.0.0:8080 - starting the main loop
srv update_slots: all slots are idle
srv log_server_r: request: GET / 127.0.0.1 200
srv log_server_r: request: GET /props 127.0.0.1 200
srv log_server_r: request: GET /props 127.0.0.1 200
srv log_server_r: request: GET /props 127.0.0.1 200
srv update_slots: all slots are idle
srv log_server_r: request: GET /slots 127.0.0.1 200
srv params_from_: Chat format: Content-only
slot get_availabl: id 3 | task -1 | selected slot by LRU, t_last = -1
slot launch_slot_: id 3 | task -1 | sampler chain: logits -> logit-bias -> penalties -> dry -> top-n-sigma -> top-k -> typical -> top-p -> min-p -> xtc -> temp-ext -> dist
slot launch_slot_: id 3 | task 1 | processing task
slot update_slots: id 3 | task 1 | new prompt, n_ctx_slot = 8192, n_keep = 0, task.n_tokens = 271
slot update_slots: id 3 | task 1 | n_tokens = 0, memory_seq_rm [0, end)
slot update_slots: id 3 | task 1 | prompt processing progress, n_tokens = 207, batch.n_tokens = 207, progress = 0.763838
First Bad Commit
I can observe that error since few last release builds, don't remember exact crash point but something has changed over the weekend.
Relevant log output
Batch offset=0x0 len=0x0 on queue 0 (aperture: 10646.8Mb, 5.2Mb VRAM only)
BO: addr=0xffffeffeffffe000-0xffffeffeffffffff size= 8KB handle=00010 capture=1 vram_only=0 name=workaround
BO: addr=0x00000002bfff0000-0x00000002bfffffff size= 64KB handle=00038 capture=1 vram_only=0 name=indirect descriptors
BO: addr=0x00000007d8800000-0x00000007d8ba3fff size= 3728KB handle=00042 capture=0 vram_only=0 name=user
BO: addr=0x0000000300000000-0x000000030003ffff size= 256KB handle=00003 capture=1 vram_only=0 name=instruction pool
BO: addr=0x00000000c0000000-0x00000000c00fffff size= 1024KB handle=00006 capture=1 vram_only=0 name=binding table pool
BO: addr=0x000000007ff40000-0x000000007fffffff size= 768KB handle=00039 capture=0 vram_only=1 name=scratch
BO: addr=0xffffeffefffe1000-0xffffeffefffe2fff size= 8KB handle=00050 capture=1 vram_only=0 name=batch
BO: addr=0x000000007fac0000-0x000000007fc3ffff size= 1536KB handle=00123 capture=0 vram_only=1 name=scratch
BO: addr=0x000000007fc40000-0x000000007ff3ffff size= 3072KB handle=00046 capture=0 vram_only=1 name=scratch
BO: addr=0xffffeffefffd9000-0xffffeffefffdcfff size= 16KB handle=00053 capture=1 vram_only=0 name=batch
BO: addr=0x0000000140000000-0x000000014000ffff size= 64KB handle=00005 capture=1 vram_only=0 name=bindless surface state pool
BO: addr=0x00000002c0000000-0x00000002c000ffff size= 64KB handle=00007 capture=1 vram_only=0 name=indirect push descriptor pool
BO: addr=0x0000000100000000-0x000000010000ffff size= 64KB handle=00004 capture=1 vram_only=0 name=internal surface state pool
BO: addr=0x0000000100010000-0x000000010001ffff size= 64KB handle=00044 capture=1 vram_only=0 name=internal surface state pool
BO: addr=0x0000000100020000-0x000000010003ffff size= 128KB handle=00049 capture=1 vram_only=0 name=internal surface state pool
BO: addr=0x0000000100040000-0x000000010007ffff size= 256KB handle=00055 capture=1 vram_only=0 name=internal surface state pool
BO: addr=0x0000000100080000-0x00000001000fffff size= 512KB handle=00064 capture=1 vram_only=0 name=internal surface state pool
BO: addr=0x0000000100100000-0x00000001001fffff size= 1024KB handle=00084 capture=1 vram_only=0 name=internal surface state pool
BO: addr=0x00000003c0000000-0x00000003c003ffff size= 256KB handle=00002 capture=1 vram_only=0 name=dynamic pool
BO: addr=0x00000003c0040000-0x00000003c007ffff size= 256KB handle=00012 capture=1 vram_only=0 name=dynamic pool
BO: addr=0x00000003c0080000-0x00000003c00fffff size= 512KB handle=00047 capture=1 vram_only=0 name=dynamic pool
BO: addr=0x00000003c0100000-0x00000003c01fffff size= 1024KB handle=00066 capture=1 vram_only=0 name=dynamic pool
BO: addr=0x00000003c0200000-0x00000003c03fffff size= 2048KB handle=00097 capture=1 vram_only=0 name=dynamic pool
BO: addr=0x0000000000200000-0x000000000023ffff size= 256KB handle=00001 capture=1 vram_only=0 name=general pool
BO: addr=0x0000000300040000-0x000000030007ffff size= 256KB handle=00040 capture=1 vram_only=0 name=instruction pool
BO: addr=0x0000000300080000-0x00000003000fffff size= 512KB handle=00043 capture=1 vram_only=0 name=instruction pool
BO: addr=0x0000000300100000-0x00000003001fffff size= 1024KB handle=00045 capture=1 vram_only=0 name=instruction pool
BO: addr=0x00000000c0100000-0x00000000c01fffff size= 1024KB handle=00069 capture=1 vram_only=0 name=binding table pool
BO: addr=0x00000000c0200000-0x00000000c03fffff size= 2048KB handle=00122 capture=1 vram_only=0 name=binding table pool
BO: addr=0x00000004bffff000-0x00000004c003efff size= 256KB handle=00008 capture=1 vram_only=0 name=aux-tt pool
BO: addr=0x00000004c003f000-0x00000004c01fefff size= 1792KB handle=00009 capture=1 vram_only=0 name=aux-tt pool
BO: addr=0x0000000540000000-0x0000000561ffffff size= 557056KB handle=00014 capture=0 vram_only=0 name=user
BO: addr=0x0000000562000000-0x00000005f6bfffff size=2437120KB handle=00015 capture=0 vram_only=0 name=user
BO: addr=0x00000005f6c00000-0x0000000635bcdfff size=1031992KB handle=00016 capture=0 vram_only=0 name=user
BO: addr=0x0000000635c00000-0x0000000675b14fff size=1047636KB handle=00017 capture=0 vram_only=0 name=user
BO: addr=0x0000000675c00000-0x00000006b5b19fff size=1047656KB handle=00018 capture=0 vram_only=0 name=user
BO: addr=0x00000006b5c00000-0x00000006f3cc9fff size=1016616KB handle=00019 capture=0 vram_only=0 name=user
BO: addr=0x00000006f3e00000-0x00000006f619efff size= 36476KB handle=00020 capture=0 vram_only=0 name=user
BO: addr=0x00000006f6200000-0x0000000718200fff size= 557060KB handle=00021 capture=0 vram_only=0 name=user
BO: addr=0x0000000718400000-0x00000007acffffff size=2437120KB handle=00023 capture=0 vram_only=0 name=user
BO: addr=0x00000007ad000000-0x00000007ad400fff size= 4100KB handle=00032 capture=0 vram_only=0 name=user
BO: addr=0x00000007ad600000-0x00000007b15fffff size= 65536KB handle=00033 capture=0 vram_only=0 name=user
BO: addr=0x00000007b1600000-0x00000007b65fffff size= 81920KB handle=00034 capture=0 vram_only=0 name=user
BO: addr=0x00000007b6600000-0x00000007d6dfffff size= 532480KB handle=00035 capture=0 vram_only=0 name=user
BO: addr=0x00000007d6e00000-0x00000007d8708fff size= 25636KB handle=00036 capture=0 vram_only=0 name=user
BO: addr=0xffffeffefffdf000-0xffffeffefffe0fff size= 8KB handle=00051 capture=1 vram_only=0 name=batch
llama.cpp/build/bin/libggml-base.so.0(+0x149a5) [0x7f21443769a5]
llama.cpp/build/bin/libggml-base.so.0(ggml_print_backtrace+0x1df) [0x7f2144376d6f]
llama.cpp/build/bin/libggml-base.so.0(+0x26579) [0x7f2144388579]
/lib/x86_64-linux-gnu/libstdc++.so.6(+0xb344a) [0x7f21440b344a]
/lib/x86_64-linux-gnu/libstdc++.so.6(_ZSt10unexpectedv+0x0) [0x7f21440a15e9]
/lib/x86_64-linux-gnu/libstdc++.so.6(+0xb36c8) [0x7f21440b36c8]
llama.cpp/build/bin/libggml-vulkan.so.0(+0x7e20b) [0x7f214027e20b]
llama.cpp/build/bin/libggml-vulkan.so.0(+0x175ba2) [0x7f2140375ba2]
llama.cpp/build/bin/libggml-vulkan.so.0(+0x176769) [0x7f2140376769]
llama.cpp/build/bin/libggml-base.so.0(ggml_backend_sched_graph_compute_async+0x827) [0x7f2144391927]
llama.cpp/build/bin/libllama.so.0(_ZN13llama_context13graph_computeEP11ggml_cgraphb+0xa1) [0x7f21444960e1]
llama.cpp/build/bin/libllama.so.0(_ZN13llama_context14process_ubatchERK12llama_ubatch14llm_graph_typeP22llama_memory_context_iR11ggml_status+0xe1) [0x7f2144497d71]
llama.cpp/build/bin/libllama.so.0(_ZN13llama_context6decodeERK11llama_batch+0x3bf) [0x7f214449cd7f]
llama.cpp/build/bin/libllama.so.0(llama_decode+0xb) [0x7f214449dc0b]
llama-server(+0xc2c7a) [0x5621c42b2c7a]
llama-server(+0x9a02b) [0x5621c428a02b]
llama-server(+0x6317c) [0x5621c425317c]
/lib/x86_64-linux-gnu/libc.so.6(+0x29ca8) [0x7f2143e06ca8]
/lib/x86_64-linux-gnu/libc.so.6(__libc_start_main+0x85) [0x7f2143e06d65]
llama-server(+0x64b91) [0x5621c4254b91]
terminate called after throwing an instance of 'vk::DeviceLostError'
what(): vk::Queue::submit: ErrorDeviceLost