Bug: llama-cli templating does buf.resize(-1) if the model's template is not supported, causing crash

[justinsteven]

What happened?

common.cpp's llama_chat_apply_template says:

https://github.com/ggerganov/llama.cpp/blob/ac146628e47451c531a3c7e62e6a973a2bb467a0/common/common.cpp#L2630-L2637

res can be -1 (e.g. in the case that the model's Jinja template is not matched by any pattern in llama_chat_apply_template_internal?). When cast to size_t it becomes significantly bigger than buf.size() which leads to buf.resize(-1). This is followed by a crash of llama-cli on my machine.

llama-server seems to fall back to chatml if no pattern matches the model's Jinja template:

https://github.com/ggerganov/llama.cpp/blob/ac146628e47451c531a3c7e62e6a973a2bb467a0/examples/server/server.cpp#L2600-L2605

Perhaps the same needs to be done for llama-cli, and perhaps common.cpp's llama_chat_apply_template should be more defensive when llama_chat_apply_template_internal returns -1, rather than trying to resize buf to -1

Name and Version

$ ./llama-cli --version
version: 3246 (ac146628)
built with cc (Debian 12.2.0-14) 12.2.0 for x86_64-linux-gnu

What operating system are you seeing the problem on?

Linux

Relevant log output

$ ./llama-cli -m /models/bartowski/DeepSeek-Coder-V2-Lite-Instruct-GGUF/DeepSeek-Coder-V2-Lite-Instruct-Q6_K.gguf -c 8192
Log start
main: build = 3246 (ac146628)
main: built with cc (Debian 12.2.0-14) 12.2.0 for x86_64-linux-gnu
main: seed  = 1719466734
llama_model_loader: loaded meta data with 42 key-value pairs and 377 tensors from /models/bartowski/DeepSeek-Coder-V2-Lite-Instruct-GGUF/DeepSeek-Coder-V2-Lite-Instruct-Q6_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              = deepseek2
llama_model_loader: - kv   1:                               general.name str              = DeepSeek-Coder-V2-Lite-Instruct
llama_model_loader: - kv   2:                      deepseek2.block_count u32              = 27
llama_model_loader: - kv   3:                   deepseek2.context_length u32              = 163840
llama_model_loader: - kv   4:                 deepseek2.embedding_length u32              = 2048
llama_model_loader: - kv   5:              deepseek2.feed_forward_length u32              = 10944
llama_model_loader: - kv   6:             deepseek2.attention.head_count u32              = 16
llama_model_loader: - kv   7:          deepseek2.attention.head_count_kv u32              = 16
llama_model_loader: - kv   8:                   deepseek2.rope.freq_base f32              = 10000.000000
llama_model_loader: - kv   9: deepseek2.attention.layer_norm_rms_epsilon f32              = 0.000001
llama_model_loader: - kv  10:                deepseek2.expert_used_count u32              = 6
llama_model_loader: - kv  11:                          general.file_type u32              = 18
llama_model_loader: - kv  12:        deepseek2.leading_dense_block_count u32              = 1
llama_model_loader: - kv  13:                       deepseek2.vocab_size u32              = 102400
llama_model_loader: - kv  14:           deepseek2.attention.kv_lora_rank u32              = 512
llama_model_loader: - kv  15:             deepseek2.attention.key_length u32              = 192
llama_model_loader: - kv  16:           deepseek2.attention.value_length u32              = 128
llama_model_loader: - kv  17:       deepseek2.expert_feed_forward_length u32              = 1408
llama_model_loader: - kv  18:                     deepseek2.expert_count u32              = 64
llama_model_loader: - kv  19:              deepseek2.expert_shared_count u32              = 2
llama_model_loader: - kv  20:             deepseek2.expert_weights_scale f32              = 1.000000
llama_model_loader: - kv  21:             deepseek2.rope.dimension_count u32              = 64
llama_model_loader: - kv  22:                deepseek2.rope.scaling.type str              = yarn
llama_model_loader: - kv  23:              deepseek2.rope.scaling.factor f32              = 40.000000
llama_model_loader: - kv  24: deepseek2.rope.scaling.original_context_length u32              = 4096
llama_model_loader: - kv  25: deepseek2.rope.scaling.yarn_log_multiplier f32              = 0.070700
llama_model_loader: - kv  26:                       tokenizer.ggml.model str              = gpt2
llama_model_loader: - kv  27:                         tokenizer.ggml.pre str              = deepseek-llm
llama_model_loader: - kv  28:                      tokenizer.ggml.tokens arr[str,102400]  = ["!", "\"", "#", "$", "%", "&", "'", ...
llama_model_loader: - kv  29:                  tokenizer.ggml.token_type arr[i32,102400]  = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
llama_model_loader: - kv  30:                      tokenizer.ggml.merges arr[str,99757]   = ["Ġ Ġ", "Ġ t", "Ġ a", "i n", "h e...
llama_model_loader: - kv  31:                tokenizer.ggml.bos_token_id u32              = 100000
llama_model_loader: - kv  32:                tokenizer.ggml.eos_token_id u32              = 100001
llama_model_loader: - kv  33:            tokenizer.ggml.padding_token_id u32              = 100001
llama_model_loader: - kv  34:               tokenizer.ggml.add_bos_token bool             = true
llama_model_loader: - kv  35:               tokenizer.ggml.add_eos_token bool             = false
llama_model_loader: - kv  36:                    tokenizer.chat_template str              = {% if not add_generation_prompt is de...
llama_model_loader: - kv  37:               general.quantization_version u32              = 2
llama_model_loader: - kv  38:                      quantize.imatrix.file str              = /models/DeepSeek-Coder-V2-Lite-Instru...
llama_model_loader: - kv  39:                   quantize.imatrix.dataset str              = /training_data/calibration_datav3.txt
llama_model_loader: - kv  40:             quantize.imatrix.entries_count i32              = 293
llama_model_loader: - kv  41:              quantize.imatrix.chunks_count i32              = 139
llama_model_loader: - type  f32:  108 tensors
llama_model_loader: - type q8_0:   27 tensors
llama_model_loader: - type q6_K:  242 tensors
llm_load_vocab: special tokens cache size = 2400
llm_load_vocab: token to piece cache size = 0.6661 MB
llm_load_print_meta: format           = GGUF V3 (latest)
llm_load_print_meta: arch             = deepseek2
llm_load_print_meta: vocab type       = BPE
llm_load_print_meta: n_vocab          = 102400
llm_load_print_meta: n_merges         = 99757
llm_load_print_meta: n_ctx_train      = 163840
llm_load_print_meta: n_embd           = 2048
llm_load_print_meta: n_head           = 16
llm_load_print_meta: n_head_kv        = 16
llm_load_print_meta: n_layer          = 27
llm_load_print_meta: n_rot            = 64
llm_load_print_meta: n_embd_head_k    = 192
llm_load_print_meta: n_embd_head_v    = 128
llm_load_print_meta: n_gqa            = 1
llm_load_print_meta: n_embd_k_gqa     = 3072
llm_load_print_meta: n_embd_v_gqa     = 2048
llm_load_print_meta: f_norm_eps       = 0.0e+00
llm_load_print_meta: f_norm_rms_eps   = 1.0e-06
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             = 10944
llm_load_print_meta: n_expert         = 64
llm_load_print_meta: n_expert_used    = 6
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     = yarn
llm_load_print_meta: freq_base_train  = 10000.0
llm_load_print_meta: freq_scale_train = 0.025
llm_load_print_meta: n_ctx_orig_yarn  = 4096
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       = 16B
llm_load_print_meta: model ftype      = Q6_K
llm_load_print_meta: model params     = 15.71 B
llm_load_print_meta: model size       = 13.10 GiB (7.16 BPW)
llm_load_print_meta: general.name     = DeepSeek-Coder-V2-Lite-Instruct
llm_load_print_meta: BOS token        = 100000 '<｜begin▁of▁sentence｜>'
llm_load_print_meta: EOS token        = 100001 '<｜end▁of▁sentence｜>'
llm_load_print_meta: PAD token        = 100001 '<｜end▁of▁sentence｜>'
llm_load_print_meta: LF token         = 126 'Ä'
llm_load_print_meta: max token length = 256
llm_load_print_meta: n_layer_dense_lead   = 1
llm_load_print_meta: n_lora_q             = 0
llm_load_print_meta: n_lora_kv            = 512
llm_load_print_meta: n_ff_exp             = 1408
llm_load_print_meta: n_expert_shared      = 2
llm_load_print_meta: expert_weights_scale = 1.0
llm_load_print_meta: rope_yarn_log_mul    = 0.0707
llm_load_tensors: ggml ctx size =    0.16 MiB
llm_load_tensors:        CPU buffer size = 13411.50 MiB
......................................................................................
llama_new_context_with_model: n_ctx      = 8192
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 = 0.025
llama_kv_cache_init:        CPU KV buffer size =  2160.00 MiB
llama_new_context_with_model: KV self size  = 2160.00 MiB, K (f16): 1296.00 MiB, V (f16):  864.00 MiB
llama_new_context_with_model:        CPU  output buffer size =     0.39 MiB
llama_new_context_with_model:        CPU compute buffer size =   296.01 MiB
llama_new_context_with_model: graph nodes  = 1924
llama_new_context_with_model: graph splits = 1
terminate called after throwing an instance of 'std::length_error'
  what():  vector::_M_default_append
Aborted (core dumped)

[LordSyd]

I have the same error, in my case the command I tried to run was:

./llama-cli -m ../sauerkrautlm-una-solar-instruct.Q4_K_M.gguf --color -c 4096 --temp 0.7 --repeat_penalty 1.1 -n -1 -p "### User:\n{prompt}\n\n### Assistant:"

In my case I tried it on a Hugging Face space with basic hardware, just as a general test.

[ngxson]

This will be fixed after #8160 merge