Segmentation fault on finetune with -ngl > 0, Debian 12 stable

[Basiliotornado]

Specs: rtx 3060ti w/ 8gb vram, r7 5700x, 32gb ram

main says main: build = 2769 (8843a98c) main: built with cc (Debian 12.2.0-14) 12.2.0 for x86_64-linux-gnu

make says GNU Make 4.3 Built for x86_64-pc-linux-gnu Copyright (C) 1988-2020 Free Software Foundation, Inc.

Compiled using make LLAMA_CUDA=1, with CUDA 11.8.89~11.8.0-5~deb12u1. Ran using ./finetune --model-base ./models/tinyllama1.1b.gguf --train-data ../data.txt -ngl 100 on TinyLlama-1.1B-Chat

Nothing seems to be off when using it compiled with debug

Output of finetune in debug

~/Downloads/Llama/llama.cpp$ ./finetune --model-base ./models/tinyllama1.1b.gguf --train-data ../data.txt -ngl 100 main: seed: 1714445201 main: model base = './models/tinyllama1.1b.gguf' llama_model_loader: loaded meta data with 22 key-value pairs and 201 tensors from ./models/tinyllama1.1b.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.block_count u32 = 22 llama_model_loader: - kv 3: llama.context_length u32 = 2048 llama_model_loader: - kv 4: llama.embedding_length u32 = 2048 llama_model_loader: - kv 5: llama.feed_forward_length u32 = 5632 llama_model_loader: - kv 6: llama.attention.head_count u32 = 32 llama_model_loader: - kv 7: llama.attention.head_count_kv u32 = 4 llama_model_loader: - kv 8: llama.rope.freq_base f32 = 10000.000000 llama_model_loader: - kv 9: llama.attention.layer_norm_rms_epsilon f32 = 0.000010 llama_model_loader: - kv 10: general.file_type u32 = 1 llama_model_loader: - kv 11: llama.vocab_size u32 = 32000 llama_model_loader: - kv 12: llama.rope.dimension_count u32 = 64 llama_model_loader: - kv 13: tokenizer.ggml.model str = llama llama_model_loader: - kv 14: tokenizer.ggml.tokens arr[str,32000] = ["", "", "", "<0x00>", "<... llama_model_loader: - kv 15: tokenizer.ggml.scores arr[f32,32000] = [0.000000, 0.000000, 0.000000, 0.0000... llama_model_loader: - kv 16: tokenizer.ggml.token_type arr[i32,32000] = [2, 3, 3, 6, 6, 6, 6, 6, 6, 6, 6, 6, ... 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: - type f32: 45 tensors llama_model_loader: - type f16: 156 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 = 2048 llm_load_print_meta: n_embd = 2048 llm_load_print_meta: n_head = 32 llm_load_print_meta: n_head_kv = 4 llm_load_print_meta: n_layer = 22 llm_load_print_meta: n_rot = 64 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_yarn_orig_ctx = 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: model type = 1B llm_load_print_meta: model ftype = F16 llm_load_print_meta: model params = 1.10 B llm_load_print_meta: model size = 2.05 GiB (16.00 BPW) llm_load_print_meta: general.name = TinyLlama_TinyLlama-1.1B-Chat-v1.0 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 = 2 '' llm_load_print_meta: LF token = 13 '<0x0A>' ggml_cuda_init: GGML_CUDA_FORCE_MMQ: no ggml_cuda_init: CUDA_USE_TENSOR_CORES: yes ggml_cuda_init: found 1 CUDA devices: Device 0: NVIDIA GeForce RTX 3060 Ti, compute capability 8.6, VMM: yes llm_load_tensors: ggml ctx size = 0.20 MiB llm_load_tensors: offloading 22 repeating layers to GPU llm_load_tensors: offloading non-repeating layers to GPU llm_load_tensors: offloaded 23/23 layers to GPU llm_load_tensors: CPU buffer size = 125.00 MiB llm_load_tensors: CUDA0 buffer size = 1973.35 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: CUDA0 KV buffer size = 11.00 MiB llama_new_context_with_model: KV self size = 11.00 MiB, K (f16): 5.50 MiB, V (f16): 5.50 MiB llama_new_context_with_model: CUDA_Host output buffer size = 0.12 MiB ggml_gallocr_reserve_n: reallocating CUDA0 buffer from size 0.00 MiB to 66.50 MiB ggml_gallocr_reserve_n: reallocating CUDA_Host buffer from size 0.00 MiB to 5.01 MiB llama_new_context_with_model: CUDA0 compute buffer size = 66.50 MiB llama_new_context_with_model: CUDA_Host compute buffer size = 5.01 MiB llama_new_context_with_model: graph nodes = 710 llama_new_context_with_model: graph splits = 2 main: init model print_params: n_vocab : 32000 print_params: n_ctx : 128 print_params: n_embd : 2048 print_params: n_ff : 5632 print_params: n_head : 32 print_params: n_head_kv : 4 print_params: n_layer : 22 print_params: norm_rms_eps : 0.000010 print_params: rope_freq_base : 10000.000000 print_params: rope_freq_scale : 1.000000 print_lora_params: n_rank_attention_norm : 1 print_lora_params: n_rank_wq : 4 print_lora_params: n_rank_wk : 4 print_lora_params: n_rank_wv : 4 print_lora_params: n_rank_wo : 4 print_lora_params: n_rank_ffn_norm : 1 print_lora_params: n_rank_ffn_gate : 4 print_lora_params: n_rank_ffn_down : 4 print_lora_params: n_rank_ffn_up : 4 print_lora_params: n_rank_tok_embeddings : 4 print_lora_params: n_rank_norm : 1 print_lora_params: n_rank_output : 4 main: total train_iterations 0 main: seen train_samples 0 main: seen train_tokens 0 main: completed train_epochs 0 main: lora_size = 28472224 bytes (27.2 MB) main: opt_size = 42223360 bytes (40.3 MB) main: opt iter 0 main: input_size = 131076128 bytes (125.0 MB) ggml_gallocr_reserve_n: reallocating CPU buffer from size 0.00 MiB to 3825.01 MiB ggml_gallocr_reserve_n: reallocating CPU buffer from size 0.00 MiB to 4706.01 MiB main: compute_size = 4010812000 bytes (3825.0 MB) main: evaluation order = LEFT_TO_RIGHT ggml_gallocr_needs_realloc: graph has different number of nodes ggml_gallocr_alloc_graph: reallocating buffers automatically ggml_gallocr_reserve_n: reallocating CPU buffer from size 0.00 MiB to 3825.01 MiB main: tokenize training data from ../data.txt main: sample-start: main: include-sample-start: false tokenize_file: total number of samples: 31763 main: number of training tokens: 31891 main: number of unique tokens: 3308 main: train data seems to have changed. restarting shuffled epoch. main: begin training main: work_size = 768376 bytes (0.7 MB) train_opt_callback: iter= 0 sample=1/31763 sched=0.000000 loss=0.000000 |-> Segmentation fault (core dumped)

In KDE System Monitor, it seems to crash before anything can be done on the GPU

[Basiliotornado]

Occasionally I'll get a segfault in main as well. Albeit, using text-generation-webui, so likely on an old version of llamacpp. Doubt it's the same issue but thought i'd share.

GGML_ASSERT: /home/runner/work/llama-cpp-python-cuBLAS-wheels/llama-cpp-python-cuBLAS-wheels/vendor/llama.cpp/ggml.c:5513: a->ne[2] == b->ne[0]
Segmentation fault (core dumped)
Unable to attach: program terminated with signal SIGSEGV, Segmentation fault.
No stack.
The program is not being run.

[clort81]

I segfault on the finetune example as well. Perhaps it's our cuda version?

nvidia-cuda-toolkit (11.8.0-5~deb12u1