Closed
Description
Name and Version
It works fine with the cpu backend. I'm using -ngl 0 because the decoding is faster on the cpu, but the igpu makes a massive improvement in the image processing. Both ggufs were downloaded from the ggml-org Huggingface repo.
Vulkan inference works fine with moondream2, glm, internvlm, smolvlm and qwen2.5vlm.
The error is:
llama.cpp/ggml/src/ggml-vulkan/ggml-vulkan.cpp:6524: GGML_ASSERT(ggml_vk_op_supports_incontiguous(op) || ggml_vk_dim01_contiguous(src0)) failed
llama.cpp version:
llama-mtmd-cli --version
ggml_vulkan: Found 1 Vulkan devices:
ggml_vulkan: 0 = Intel(R) Iris(R) Plus Graphics (ICL GT2) (Intel open-source Mesa driver) | uma: 1 | fp16: 1 | warp size: 32 | shared memory: 65536 | int dot: 0 | matrix cores: none
version: 5471 (ffd0eae6)
built with cc (GCC) 15.1.1 20250425 for x86_64-pc-linux-gnu
Full log is:
llama.cpp/build_vulkan/bin/llama-mtmd-cli -ngl 0 -m /vlms/mistral-small-31-24b-text-IQ2_M.gguf --mmproj vlms/mistral-small-31-24b-mmproj-f16.gguf --image /tmp/tmp5po9zi2y.jpg -p 'Describe this image in more than 10 words but less than 50 words.'
ggml_vulkan: Found 1 Vulkan devices:
ggml_vulkan: 0 = Intel(R) Iris(R) Plus Graphics (ICL GT2) (Intel open-source Mesa driver) | uma: 1 | fp16: 1 | warp size: 32 | shared memory: 65536 | int dot: 0 | matrix cores: none
build: 5471 (ffd0eae6) with cc (GCC) 15.1.1 20250425 for x86_64-pc-linux-gnu
llama_model_load_from_file_impl: using device Vulkan0 (Intel(R) Iris(R) Plus Graphics (ICL GT2)) - 7771 MiB free
llama_model_loader: loaded meta data with 39 key-value pairs and 363 tensors from vlms/mistral-small-31-24b-text-IQ2_M.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.type str = model
llama_model_loader: - kv 2: general.name str = Mistral-Small-3.1-24B-Instruct-2503
llama_model_loader: - kv 3: general.version str = 2503
llama_model_loader: - kv 4: general.finetune str = Instruct
llama_model_loader: - kv 5: general.basename str = Mistral-Small-3.1-24B-Instruct-2503
llama_model_loader: - kv 6: general.quantized_by str = Unsloth
llama_model_loader: - kv 7: general.size_label str = 24B
llama_model_loader: - kv 8: general.repo_url str = https://huggingface.co/unsloth
llama_model_loader: - kv 9: llama.block_count u32 = 40
llama_model_loader: - kv 10: llama.context_length u32 = 131072
llama_model_loader: - kv 11: llama.embedding_length u32 = 5120
llama_model_loader: - kv 12: llama.feed_forward_length u32 = 32768
llama_model_loader: - kv 13: llama.attention.head_count u32 = 32
llama_model_loader: - kv 14: llama.attention.head_count_kv u32 = 8
llama_model_loader: - kv 15: llama.rope.freq_base f32 = 1000000000.000000
llama_model_loader: - kv 16: llama.attention.layer_norm_rms_epsilon f32 = 0.000010
llama_model_loader: - kv 17: llama.attention.key_length u32 = 128
llama_model_loader: - kv 18: llama.attention.value_length u32 = 128
llama_model_loader: - kv 19: llama.vocab_size u32 = 131072
llama_model_loader: - kv 20: llama.rope.dimension_count u32 = 128
llama_model_loader: - kv 21: tokenizer.ggml.model str = gpt2
llama_model_loader: - kv 22: tokenizer.ggml.pre str = tekken
llama_model_loader: - kv 23: tokenizer.ggml.tokens arr[str,131072] = ["<unk>", "<s>", "</s>", "[INST]", "[...
llama_model_loader: - kv 24: tokenizer.ggml.token_type arr[i32,131072] = [3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, ...
llama_model_loader: - kv 25: tokenizer.ggml.merges arr[str,269443] = ["Ġ Ġ", "Ġ t", "e r", "i n", "Ġ �...
llama_model_loader: - kv 26: tokenizer.ggml.bos_token_id u32 = 1
llama_model_loader: - kv 27: tokenizer.ggml.eos_token_id u32 = 2
llama_model_loader: - kv 28: tokenizer.ggml.unknown_token_id u32 = 0
llama_model_loader: - kv 29: tokenizer.ggml.padding_token_id u32 = 11
llama_model_loader: - kv 30: tokenizer.ggml.add_bos_token bool = true
llama_model_loader: - kv 31: tokenizer.ggml.add_eos_token bool = false
llama_model_loader: - kv 32: tokenizer.ggml.add_space_prefix bool = false
llama_model_loader: - kv 33: general.quantization_version u32 = 2
llama_model_loader: - kv 34: general.file_type u32 = 29
llama_model_loader: - kv 35: quantize.imatrix.file str = Mistral-Small-3.1-24B-Instruct-2503-G...
llama_model_loader: - kv 36: quantize.imatrix.dataset str = unsloth_calibration_Mistral-Small-3.1...
llama_model_loader: - kv 37: quantize.imatrix.entries_count i32 = 280
llama_model_loader: - kv 38: quantize.imatrix.chunks_count i32 = 55
llama_model_loader: - type f32: 81 tensors
llama_model_loader: - type q3_K: 1 tensors
llama_model_loader: - type q4_K: 40 tensors
llama_model_loader: - type q5_K: 1 tensors
llama_model_loader: - type iq2_xs: 22 tensors
llama_model_loader: - type iq3_xxs: 72 tensors
llama_model_loader: - type iq3_s: 58 tensors
llama_model_loader: - type iq2_s: 88 tensors
print_info: file format = GGUF V3 (latest)
print_info: file type = IQ2_M - 2.7 bpw
print_info: file size = 8.15 GiB (2.97 BPW)
load: special_eos_id is not in special_eog_ids - the tokenizer config may be incorrect
load: special tokens cache size = 1000
load: token to piece cache size = 0.8498 MB
print_info: arch = llama
print_info: vocab_only = 0
print_info: n_ctx_train = 131072
print_info: n_embd = 5120
print_info: n_layer = 40
print_info: n_head = 32
print_info: n_head_kv = 8
print_info: n_rot = 128
print_info: n_swa = 0
print_info: is_swa_any = 0
print_info: n_embd_head_k = 128
print_info: n_embd_head_v = 128
print_info: n_gqa = 4
print_info: n_embd_k_gqa = 1024
print_info: n_embd_v_gqa = 1024
print_info: f_norm_eps = 0.0e+00
print_info: f_norm_rms_eps = 1.0e-05
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 = 0.0e+00
print_info: n_ff = 32768
print_info: n_expert = 0
print_info: n_expert_used = 0
print_info: causal attn = 1
print_info: pooling type = 0
print_info: rope type = 0
print_info: rope scaling = linear
print_info: freq_base_train = 1000000000.0
print_info: freq_scale_train = 1
print_info: n_ctx_orig_yarn = 131072
print_info: rope_finetuned = unknown
print_info: ssm_d_conv = 0
print_info: ssm_d_inner = 0
print_info: ssm_d_state = 0
print_info: ssm_dt_rank = 0
print_info: ssm_dt_b_c_rms = 0
print_info: model type = 13B
print_info: model params = 23.57 B
print_info: general.name = Mistral-Small-3.1-24B-Instruct-2503
print_info: vocab type = BPE
print_info: n_vocab = 131072
print_info: n_merges = 269443
print_info: BOS token = 1 '<s>'
print_info: EOS token = 2 '</s>'
print_info: UNK token = 0 '<unk>'
print_info: PAD token = 11 '<pad>'
print_info: LF token = 1010 'Ċ'
print_info: EOG token = 2 '</s>'
print_info: max token length = 150
load_tensors: loading model tensors, this can take a while... (mmap = true)
load_tensors: offloading 0 repeating layers to GPU
load_tensors: offloaded 0/41 layers to GPU
load_tensors: CPU_Mapped model buffer size = 8342.83 MiB
..............................................................................................
llama_context: constructing llama_context
llama_context: n_seq_max = 1
llama_context: n_ctx = 4096
llama_context: n_ctx_per_seq = 4096
llama_context: n_batch = 2048
llama_context: n_ubatch = 512
llama_context: causal_attn = 1
llama_context: flash_attn = 0
llama_context: freq_base = 1000000000.0
llama_context: freq_scale = 1
llama_context: n_ctx_per_seq (4096) < n_ctx_train (131072) -- the full capacity of the model will not be utilized
llama_context: CPU output buffer size = 0.50 MiB
llama_kv_cache_unified: CPU KV buffer size = 640.00 MiB
llama_kv_cache_unified: size = 640.00 MiB ( 4096 cells, 40 layers, 1 seqs), K (f16): 320.00 MiB, V (f16): 320.00 MiB
llama_context: Vulkan0 compute buffer size = 706.00 MiB
llama_context: Vulkan_Host compute buffer size = 18.01 MiB
llama_context: graph nodes = 1446
llama_context: graph splits = 444 (with bs=512), 1 (with bs=1)
common_init_from_params: setting dry_penalty_last_n to ctx_size = 4096
common_init_from_params: warming up the model with an empty run - please wait ... (--no-warmup to disable)
Failed to infer a tool call example (possible template bug)
Failed to infer a tool call example (possible template bug)
mtmd_cli_context: chat template example:
[SYSTEM_PROMPT]You are a helpful assistant[/SYSTEM_PROMPT][INST]Hello[/INST]Hi there</s>[INST]How are you?[/INST]
clip_ctx: CLIP using Vulkan0 backend
clip_model_loader: model name:
clip_model_loader: description:
clip_model_loader: GGUF version: 3
clip_model_loader: alignment: 32
clip_model_loader: n_tensors: 223
clip_model_loader: n_kv: 27
load_hparams: projector: pixtral
load_hparams: has_vision_encoder: 1
load_hparams: has_audio_encoder: 0
load_hparams: n_embd: 1024
load_hparams: n_head: 16
load_hparams: n_ff: 4096
load_hparams: n_layer: 24
load_hparams: ffn_op: gelu
load_hparams: projection_dim: 5120
load_hparams: image_size: 1540
load_hparams: patch_size: 14
load_hparams: has_llava_proj: 0
load_hparams: minicpmv_version: 0
load_hparams: proj_scale_factor: 0
load_hparams: n_wa_pattern: 0
load_hparams: model size: 837.36 MiB
load_hparams: metadata size: 0.08 MiB
alloc_compute_meta: Vulkan0 compute buffer size = 2.97 MiB
alloc_compute_meta: CPU compute buffer size = 0.14 MiB
llama.cpp/ggml/src/ggml-vulkan/ggml-vulkan.cpp:6524: GGML_ASSERT(ggml_vk_op_supports_incontiguous(op) || ggml_vk_dim01_contiguous(src0)) failed
Operating systems
Linux
GGML backends
Vulkan
Hardware
Intel Iris Plus Graphics G7 on i7-1065G7
i915 driver on Linux 6.14.6
Mesa version 25.0.5 supporting Vulkan 1.4.305
Models
https://huggingface.co/ggml-org/Mistral-Small-3.1-24B-Instruct-2503-GGUF
Problem description & steps to reproduce
llama.cpp/build_vulkan/bin/llama-mtmd-cli -ngl 0 -m /vlms/mistral-small-31-24b-text-IQ2_M.gguf --mmproj vlms/mistral-small-31-24b-mmproj-f16.gguf --image /tmp/tmp5po9zi2y.jpg -p 'Describe this image in more than 10 words but less than 50 words.'
fails with:
llama.cpp/ggml/src/ggml-vulkan/ggml-vulkan.cpp:6524: GGML_ASSERT(ggml_vk_op_supports_incontiguous(op) || ggml_vk_dim01_contiguous(src0)) failed
First Bad Commit
No response
Relevant log output
See above