GGML_ASSERT: llama.cpp/ggml-cuda/argsort.cu:48: (ncols & (ncols - 1)) == 0

[schmorp]

I get this error when trying to calculate an imatrix for

https://huggingface.co/nbeerbower/HeroBophades-3x7B

The gguf file is created by running

convert.py --skip-unknown --vocab-type spm,hfft,bpe --pad-vocab

The switches are my default switches, so are likely irrelevant. The resulting gguf file seems to work fine when used with main, but crashes when used with imatrix:

$ imatrix -ofreq 10 -t 1 -ngl 0 -mg 0 -m HeroBophades-3x7B.gguf -o "HeroBophades-3x7B.imatrix~ -f imatrix-training.txt
main: build = 2569 (5106ef4)
main: built with gcc-12 (Debian 12.2.0-14) 12.2.0 for x86_64-linux-gnu
main: seed = 1712520807
llama_model_loader: loaded meta data with 24 key-value pairs and 515 tensors from HeroBophades-3x7B.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 = .
llama_model_loader: - kv 2: llama.context_length u32 = 32768
llama_model_loader: - kv 3: llama.embedding_length u32 = 4096
llama_model_loader: - kv 4: llama.block_count u32 = 32
llama_model_loader: - kv 5: llama.feed_forward_length u32 = 14336
llama_model_loader: - kv 6: llama.rope.dimension_count u32 = 128
llama_model_loader: - kv 7: llama.attention.head_count u32 = 32
llama_model_loader: - kv 8: llama.attention.head_count_kv u32 = 8
llama_model_loader: - kv 9: llama.expert_count u32 = 3
llama_model_loader: - kv 10: llama.expert_used_count u32 = 2
llama_model_loader: - kv 11: llama.attention.layer_norm_rms_epsilon f32 = 0.000010
llama_model_loader: - kv 12: llama.rope.freq_base f32 = 10000.000000
llama_model_loader: - kv 13: general.file_type u32 = 1
llama_model_loader: - kv 14: tokenizer.ggml.model str = llama
llama_model_loader: - kv 15: tokenizer.ggml.tokens arr[str,32000] = ["", "", "", "<0x00>", "<...
llama_model_loader: - kv 16: tokenizer.ggml.scores arr[f32,32000] = [0.000000, 0.000000, 0.000000, 0.0000...
llama_model_loader: - kv 17: tokenizer.ggml.token_type arr[i32,32000] = [2, 3, 3, 6, 6, 6, 6, 6, 6, 6, 6, 6, ...
llama_model_loader: - kv 18: tokenizer.ggml.bos_token_id u32 = 1
llama_model_loader: - kv 19: tokenizer.ggml.eos_token_id u32 = 2
llama_model_loader: - kv 20: tokenizer.ggml.unknown_token_id u32 = 0
llama_model_loader: - kv 21: tokenizer.ggml.padding_token_id u32 = 1
llama_model_loader: - kv 22: tokenizer.ggml.add_bos_token bool = true
llama_model_loader: - kv 23: tokenizer.ggml.add_eos_token bool = false
llama_model_loader: - type f32: 65 tensors
llama_model_loader: - type f16: 450 tensors
llm_load_vocab: special tokens definition check successful ( 259/32000 ).
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: n_ctx_train = 32768
llm_load_print_meta: n_embd = 4096
llm_load_print_meta: n_head = 32
llm_load_print_meta: n_head_kv = 8
llm_load_print_meta: n_layer = 32
llm_load_print_meta: n_rot = 128
llm_load_print_meta: n_embd_head_k = 128
llm_load_print_meta: n_embd_head_v = 128
llm_load_print_meta: n_gqa = 4
llm_load_print_meta: n_embd_k_gqa = 1024
llm_load_print_meta: n_embd_v_gqa = 1024
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 = 14336
llm_load_print_meta: n_expert = 3
llm_load_print_meta: n_expert_used = 2
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_yarn_orig_ctx = 32768
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: model type = 7B
llm_load_print_meta: model ftype = F16
llm_load_print_meta: model params = 18.52 B
llm_load_print_meta: model size = 34.49 GiB (16.00 BPW)
llm_load_print_meta: general.name = .
llm_load_print_meta: BOS token = 1 ''
llm_load_print_meta: EOS token = 2 ''
llm_load_print_meta: UNK token = 0 ''
llm_load_print_meta: PAD token = 1 ''
llm_load_print_meta: LF token = 13 '<0x0A>'
ggml_cuda_init: GGML_CUDA_FORCE_MMQ: yes
ggml_cuda_init: CUDA_USE_TENSOR_CORES: no
ggml_cuda_init: found 1 CUDA devices:
Device 0: NVIDIA GeForce RTX 4070 Ti, compute capability 8.9, VMM: yes
llm_load_tensors: ggml ctx size = 0.20 MiB
llm_load_tensors: offloading 0 repeating layers to GPU
llm_load_tensors: offloaded 0/33 layers to GPU
llm_load_tensors: CPU buffer size = 35317.77 MiB
....................................................................................................
llama_new_context_with_model: n_ctx = 512
llama_new_context_with_model: n_batch = 512
llama_new_context_with_model: n_ubatch = 512
llama_new_context_with_model: freq_base = 10000.0
llama_new_context_with_model: freq_scale = 1
llama_kv_cache_init: CUDA_Host KV buffer size = 64.00 MiB
llama_new_context_with_model: KV self size = 64.00 MiB, K (f16): 32.00 MiB, V (f16): 32.00 MiB
llama_new_context_with_model: CUDA_Host output buffer size = 0.12 MiB
llama_new_context_with_model: CUDA0 compute buffer size = 1064.00 MiB
llama_new_context_with_model: CUDA_Host compute buffer size = 9.01 MiB
llama_new_context_with_model: graph nodes = 1670
llama_new_context_with_model: graph splits = 388

system_info: n_threads = 4 / 28 | AVX = 1 | AVX_VNNI = 1 | AVX2 = 1 | AVX512 = 0 | AVX512_VBMI = 0 | AVX512_VNNI = 0 | FMA = 1 | NEON = 0 | ARM_FMA = 0 | F16C = 1 | FP16_VA = 0 | WASM_SIMD = 0 | BLAS = 1 | SSE3 = 1 | SSSE3 = 1 | VSX = 0 | MATMUL_INT8 = 0 |
compute_imatrix: tokenizing the input ..
compute_imatrix: tokenization took 171.649 ms
compute_imatrix: computing over 307 chunks with batch_size 512
GGML_ASSERT: llama.cpp/ggml-cuda/argsort.cu:48: (ncols & (ncols - 1)) == 0
[New LWP 46883]
[New LWP 46887]
[New LWP 46888]
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
0x00007fffeeef2b57 in __GI___wait4 (pid=46892, stat_loc=0x0, options=0, usage=0x0) at ../sysdeps/unix/sysv/linux/wait4.c:30
Download failed: Invalid argument. Continuing without source file ./posix/../sysdeps/unix/sysv/linux/wait4.c.
30 ../sysdeps/unix/sysv/linux/wait4.c: Inappropriate ioctl for device.
#0 0x00007fffeeef2b57 in __GI___wait4 (pid=46892, stat_loc=0x0, options=0, usage=0x0) at ../sysdeps/unix/sysv/linux/wait4.c:30
30 in ../sysdeps/unix/sysv/linux/wait4.c
#1 0x00005555555e68eb in ggml_print_backtrace () at /llama.cpp/ggml.c:145
warning: Source file is more recent than executable.
145 waitpid(pid, NULL, 0);
#2 0x00005555556588a3 in argsort_f32_i32_cuda(float const*, int*, int, int, ggml_sort_order, CUstream_st*) ()
#3 0x0000555555658b60 in ggml_cuda_op_argsort(ggml_backend_cuda_context&, ggml_tensor*) ()
#4 0x000055555564f778 in ggml_cuda_compute_forward(ggml_backend_cuda_context&, ggml_tensor*) ()
#5 0x0000555555650279 in ggml_backend_cuda_graph_compute(ggml_backend*, ggml_cgraph*) ()
#6 0x00005555555dce8a in ggml_backend_graph_compute_async (cgraph=0x7fffffffa510, backend=0x555557092fd0) at llama.cpp/ggml-backend.c:282
282 return backend->iface.graph_compute(backend, cgraph);
#7 ggml_backend_sched_compute_splits (sched=0x7fffc017e010) at llama.cpp/ggml-backend.c:1685
1685 enum ggml_status ec = ggml_backend_graph_compute_async(split_backend, &gv);
#8 ggml_backend_sched_graph_compute_async (graph=, sched=0x7fffc017e010) at llama.cpp/ggml-backend.c:1839
1839 return ggml_backend_sched_compute_splits(sched);
#9 llama_graph_compute (lctx=..., gf=, n_threads=) at llama.cpp/llama.cpp:9756
warning: Source file is more recent than executable.
9756 //
#10 0x0000555555642711 in llama_decode_internal(llama_context&, llama_batch) [clone .isra.0] (lctx=..., batch_all=...) at llama.cpp/llama.cpp:10001
10001
#11 0x000055555557e07b in llama_decode (batch=..., ctx=0x5555563bec20) at llama.cpp/llama.cpp:15135
15135 const auto & cell = kv_self.cells[i];
#12 compute_imatrix (ctx=ctx@entry=0x5555563bec20, params=..., compute_ppl=compute_ppl@entry=true, from_chunk=from_chunk@entry=0) at llama.cpp/examples/imatrix/imatrix.cpp:428
warning: Source file is more recent than executable.
428 }
#13 0x0000555555574734 in main (argc=, argv=) at llama.cpp/examples/imatrix/imatrix.cpp:632
632 fprintf(stderr, "%s\n", get_system_info(params).c_str());
[Inferior 1 (process 46881) detached]
/tmp/ai/imatrix-training: line 33: 46881 Aborted

[slaren]

This assert no longer exists in master.

[schmorp]

It still crashes all the same, though. Should I open a new issue for this?

GGML_ASSERT: llama.cpp/examples/imatrix/imatrix.cpp:112: n_as*ggml_nrows(ids)sizeof(int) == GGML_PAD(ggml_nbytes(ids), n_assizeof(int))

[schmorp]

To make it easier, I'll open a new issue.