Bug: passing `tfs_z` crashes the server

[z80maniac]

What happened?

If you pass tfs_z param to the server, it crashes sometimes.

Starting the server:

~/test/llama.cpp/llama-server -m /opt/models/text/gemma-2-27b-it-Q8_0.gguf --verbose

startup logs

build: 3802 (a5b57b08) with cc (Ubuntu 13.2.0-23ubuntu4) 13.2.0 for x86_64-linux-gnu
system info: n_threads = 12, n_threads_batch = 12, total_threads = 24

system_info: n_threads = 12 (n_threads_batch = 12) / 24 | AVX = 1 | AVX_VNNI = 0 | AVX2 = 1 | AVX512 = 1 | AVX512_VBMI = 1 | AVX512_VNNI = 1 | AVX512_BF16 = 1 | FMA = 1 | NEON = 0 | SVE = 0 | ARM_FMA = 0 | F16C = 1 | FP16_VA = 0 | RISCV_VECT = 0 | WASM_SIMD = 0 | BLAS = 1 | SSE3 = 1 | SSSE3 = 1 | VSX = 0 | MATMUL_INT8 = 0 | LLAMAFILE = 1 |

main: HTTP server is listening, hostname: 127.0.0.1, port: 8080, http threads: 23
main: loading model
llama_model_loader: loaded meta data with 33 key-value pairs and 508 tensors from /opt/models/text/gemma-2-27b-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              = gemma2
llama_model_loader: - kv   1:                               general.name str              = gemma-2-27b-it
llama_model_loader: - kv   2:                      gemma2.context_length u32              = 8192
llama_model_loader: - kv   3:                    gemma2.embedding_length u32              = 4608
llama_model_loader: - kv   4:                         gemma2.block_count u32              = 46
llama_model_loader: - kv   5:                 gemma2.feed_forward_length u32              = 36864
llama_model_loader: - kv   6:                gemma2.attention.head_count u32              = 32
llama_model_loader: - kv   7:             gemma2.attention.head_count_kv u32              = 16
llama_model_loader: - kv   8:    gemma2.attention.layer_norm_rms_epsilon f32              = 0.000001
llama_model_loader: - kv   9:                gemma2.attention.key_length u32              = 128
llama_model_loader: - kv  10:              gemma2.attention.value_length u32              = 128
llama_model_loader: - kv  11:                          general.file_type u32              = 7
llama_model_loader: - kv  12:              gemma2.attn_logit_softcapping f32              = 50.000000
llama_model_loader: - kv  13:             gemma2.final_logit_softcapping f32              = 30.000000
llama_model_loader: - kv  14:            gemma2.attention.sliding_window u32              = 4096
llama_model_loader: - kv  15:                       tokenizer.ggml.model str              = llama
llama_model_loader: - kv  16:                         tokenizer.ggml.pre str              = default
llama_model_loader: - kv  17:                      tokenizer.ggml.tokens arr[str,256000]  = ["<pad>", "<eos>", "<bos>", "<unk>", ...
llama_model_loader: - kv  18:                      tokenizer.ggml.scores arr[f32,256000]  = [-1000.000000, -1000.000000, -1000.00...
llama_model_loader: - kv  19:                  tokenizer.ggml.token_type arr[i32,256000]  = [3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, ...
llama_model_loader: - kv  20:                tokenizer.ggml.bos_token_id u32              = 2
llama_model_loader: - kv  21:                tokenizer.ggml.eos_token_id u32              = 1
llama_model_loader: - kv  22:            tokenizer.ggml.unknown_token_id u32              = 3
llama_model_loader: - kv  23:            tokenizer.ggml.padding_token_id u32              = 0
llama_model_loader: - kv  24:               tokenizer.ggml.add_bos_token bool             = true
llama_model_loader: - kv  25:               tokenizer.ggml.add_eos_token bool             = false
llama_model_loader: - kv  26:                    tokenizer.chat_template str              = {{ bos_token }}{% if messages[0]['rol...
llama_model_loader: - kv  27:            tokenizer.ggml.add_space_prefix bool             = false
llama_model_loader: - kv  28:               general.quantization_version u32              = 2
llama_model_loader: - kv  29:                      quantize.imatrix.file str              = /models_out/gemma-2-27b-it-GGUF/gemma...
llama_model_loader: - kv  30:                   quantize.imatrix.dataset str              = /training_dir/calibration_datav3.txt
llama_model_loader: - kv  31:             quantize.imatrix.entries_count i32              = 322
llama_model_loader: - kv  32:              quantize.imatrix.chunks_count i32              = 128
llama_model_loader: - type  f32:  185 tensors
llama_model_loader: - type q8_0:  323 tensors
llm_load_vocab: special tokens cache size = 217
llm_load_vocab: token to piece cache size = 1.6014 MB
llm_load_print_meta: format           = GGUF V3 (latest)
llm_load_print_meta: arch             = gemma2
llm_load_print_meta: vocab type       = SPM
llm_load_print_meta: n_vocab          = 256000
llm_load_print_meta: n_merges         = 0
llm_load_print_meta: vocab_only       = 0
llm_load_print_meta: n_ctx_train      = 8192
llm_load_print_meta: n_embd           = 4608
llm_load_print_meta: n_layer          = 46
llm_load_print_meta: n_head           = 32
llm_load_print_meta: n_head_kv        = 16
llm_load_print_meta: n_rot            = 128
llm_load_print_meta: n_swa            = 4096
llm_load_print_meta: n_embd_head_k    = 128
llm_load_print_meta: n_embd_head_v    = 128
llm_load_print_meta: n_gqa            = 2
llm_load_print_meta: n_embd_k_gqa     = 2048
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             = 36864
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        = 2
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_ctx_orig_yarn  = 8192
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: ssm_dt_b_c_rms   = 0
llm_load_print_meta: model type       = 27B
llm_load_print_meta: model ftype      = Q8_0
llm_load_print_meta: model params     = 27.23 B
llm_load_print_meta: model size       = 26.94 GiB (8.50 BPW)
llm_load_print_meta: general.name     = gemma-2-27b-it
llm_load_print_meta: BOS token        = 2 '<bos>'
llm_load_print_meta: EOS token        = 1 '<eos>'
llm_load_print_meta: UNK token        = 3 '<unk>'
llm_load_print_meta: PAD token        = 0 '<pad>'
llm_load_print_meta: LF token         = 227 '<0x0A>'
llm_load_print_meta: EOT token        = 107 '<end_of_turn>'
llm_load_print_meta: max token length = 48
ggml_cuda_init: GGML_CUDA_FORCE_MMQ:    no
ggml_cuda_init: GGML_CUDA_FORCE_CUBLAS: no
ggml_cuda_init: found 1 CUDA devices:
  Device 0: NVIDIA GeForce RTX 3060, compute capability 8.6, VMM: yes
llm_load_tensors: ggml ctx size =    0.23 MiB
llm_load_tensors: offloading 0 repeating layers to GPU
llm_load_tensors: offloaded 0/47 layers to GPU
llm_load_tensors:        CPU buffer size = 27591.06 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 = 1
llama_kv_cache_init:  CUDA_Host KV buffer size =  2944.00 MiB
llama_new_context_with_model: KV self size  = 2944.00 MiB, K (f16): 1472.00 MiB, V (f16): 1472.00 MiB
llama_new_context_with_model:  CUDA_Host  output buffer size =     1.95 MiB
llama_new_context_with_model:      CUDA0 compute buffer size =  1704.31 MiB
llama_new_context_with_model:  CUDA_Host compute buffer size =    41.01 MiB
llama_new_context_with_model: graph nodes  = 1850
llama_new_context_with_model: graph splits = 602
llama_init_from_gpt_params: warming up the model with an empty run - please wait ... (--no-warmup to disable)
srv          init: initializing slots, n_slots = 1
slot         init: id  0 | task -1 | new slot n_ctx_slot = 8192
slot        reset: id  0 | task -1 |
main: model loaded
main: chat template, built_in: 1, chat_example: '<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 127.0.0.1:8080 - starting the main loop
que    start_loop: processing new tasks
que    start_loop: update slots
srv  update_slots: all slots are idle
srv  kv_cache_cle: clearing KV cache
que    start_loop: waiting for new tasks

Request with tfs_z:

curl --data '{"prompt": "I see", "n_predict": 2, "tfs_z": 0.9}' http://127.0.0.1:8080/completion

Failure logs

srv  add_waiting_: add task 0 to waiting list. current waiting = 0 (before add)
que          post: new task, id = 0/1, front = 0
que    start_loop: processing new tasks
que    start_loop: processing task, id = 0
slot get_availabl: id  0 | task -1 | selected slot by lru, t_last = -1
slot        reset: id  0 | task -1 |
slot launch_slot_: id  0 | task 0 | processing task
que    start_loop: update slots
srv  update_slots: posting NEXT_RESPONSE
que          post: new task, id = 1, front = 0
slot update_slots: id  0 | task 0 | tokenizing prompt, len = 1
slot update_slots: id  0 | task 0 | prompt tokenized, n_ctx_slot = 8192, n_keep = 0, n_prompt_tokens = 3
slot update_slots: id  0 | task 0 | kv cache rm [0, end)
slot update_slots: id  0 | task 0 | prompt processing progress, n_past = 3, n_tokens = 3, progress = 1.000000
slot update_slots: id  0 | task 0 | prompt done, n_past = 3, n_tokens = 3
srv  update_slots: decoding batch, n_tokens = 3
slot process_toke: id  0 | task 0 | n_decoded = 1, n_remaining = 1, next token: ' a'
srv  update_slots: run slots completed
que    start_loop: waiting for new tasks
que    start_loop: processing new tasks
que    start_loop: processing task, id = 1
que    start_loop: update slots
srv  update_slots: posting NEXT_RESPONSE
que          post: new task, id = 2, front = 0
slot update_slots: id  0 | task 0 | slot decode token, n_ctx = 8192, n_past = 4, n_system_tokens = 0, n_cache_tokens = 0, truncated = 0
srv  update_slots: decoding batch, n_tokens = 1
src/llama-sampling.cpp:66: GGML_ASSERT(cur_p->size > 0) failed
Could not attach to process.  If your uid matches the uid of the target
process, check the setting of /proc/sys/kernel/yama/ptrace_scope, or try
again as the root user.  For more details, see /etc/sysctl.d/10-ptrace.conf
ptrace: Operation not permitted.
No stack.
The program is not being run.
fish: Job 1, '~/test/llama.cpp/llama-server -…' terminated by signal SIGABRT (Abort)

It may not crash on the first request. It may take up to 10 requests sometimes. I tested it with different models, and with CUDA/no-CUDA builds.

Name and Version

❯ ./llama-server --version
version: 3802 (a5b57b0)
built with cc (Ubuntu 13.2.0-23ubuntu4) 13.2.0 for x86_64-linux-gnu

What operating system are you seeing the problem on?

Linux

Relevant log output

No response

[Senemu]

Also happens with llama-cli, since df270ef:

$ llama-cli -m Mistral-Large-Instruct-2407-Q4_K_S.gguf --tfs 0.1
build: 3799 (d09770ca) with cc (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0 for x86_64-linux-gnu
main: llama backend init
main: load the model and apply lora adapter, if any
llama_model_loader: loaded meta data with 34 key-value pairs and 795 tensors from Mistral-Large-Instruct-2407-Q4_K_S.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 Large Instruct 2407
llama_model_loader: - kv   3:                            general.version str              = 2407
llama_model_loader: - kv   4:                           general.finetune str              = Instruct
llama_model_loader: - kv   5:                           general.basename str              = Mistral
llama_model_loader: - kv   6:                         general.size_label str              = Large
llama_model_loader: - kv   7:                            general.license str              = other
llama_model_loader: - kv   8:                       general.license.name str              = mrl
llama_model_loader: - kv   9:                       general.license.link str              = https://mistral.ai/licenses/MRL-0.1.md
llama_model_loader: - kv  10:                          general.languages arr[str,10]      = ["en", "fr", "de", "es", "it", "pt", ...
llama_model_loader: - kv  11:                          llama.block_count u32              = 88
llama_model_loader: - kv  12:                       llama.context_length u32              = 32768
llama_model_loader: - kv  13:                     llama.embedding_length u32              = 12288
llama_model_loader: - kv  14:                  llama.feed_forward_length u32              = 28672
llama_model_loader: - kv  15:                 llama.attention.head_count u32              = 96
llama_model_loader: - kv  16:              llama.attention.head_count_kv u32              = 8
llama_model_loader: - kv  17:                       llama.rope.freq_base f32              = 1000000,000000
llama_model_loader: - kv  18:     llama.attention.layer_norm_rms_epsilon f32              = 0,000010
llama_model_loader: - kv  19:                          general.file_type u32              = 14
llama_model_loader: - kv  20:                           llama.vocab_size u32              = 32768
llama_model_loader: - kv  21:                 llama.rope.dimension_count u32              = 128
llama_model_loader: - kv  22:            tokenizer.ggml.add_space_prefix bool             = false
llama_model_loader: - kv  23:                       tokenizer.ggml.model str              = llama
llama_model_loader: - kv  24:                         tokenizer.ggml.pre str              = default
llama_model_loader: - kv  25:                      tokenizer.ggml.tokens arr[str,32768]   = ["<unk>", "<s>", "</s>", "[INST]", "[...
llama_model_loader: - kv  26:                      tokenizer.ggml.scores arr[f32,32768]   = [-1000,000000, -1000,000000, -1000,00...
llama_model_loader: - kv  27:                  tokenizer.ggml.token_type arr[i32,32768]   = [3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4, 4, ...
llama_model_loader: - kv  28:                tokenizer.ggml.bos_token_id u32              = 1
llama_model_loader: - kv  29:                tokenizer.ggml.eos_token_id u32              = 2
llama_model_loader: - kv  30:            tokenizer.ggml.unknown_token_id u32              = 0
llama_model_loader: - kv  31:               tokenizer.ggml.add_bos_token bool             = true
llama_model_loader: - kv  32:               tokenizer.ggml.add_eos_token bool             = false
llama_model_loader: - kv  33:               general.quantization_version u32              = 2
llama_model_loader: - type  f32:  177 tensors
llama_model_loader: - type q4_K:  602 tensors
llama_model_loader: - type q5_K:   15 tensors
llama_model_loader: - type q6_K:    1 tensors
llm_load_vocab: special tokens cache size = 771
llm_load_vocab: token to piece cache size = 0,1732 MB
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          = 32768
llm_load_print_meta: n_merges         = 0
llm_load_print_meta: vocab_only       = 0
llm_load_print_meta: n_ctx_train      = 32768
llm_load_print_meta: n_embd           = 12288
llm_load_print_meta: n_layer          = 88
llm_load_print_meta: n_head           = 96
llm_load_print_meta: n_head_kv        = 8
llm_load_print_meta: n_rot            = 128
llm_load_print_meta: n_swa            = 0
llm_load_print_meta: n_embd_head_k    = 128
llm_load_print_meta: n_embd_head_v    = 128
llm_load_print_meta: n_gqa            = 12
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             = 28672
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  = 1000000,0
llm_load_print_meta: freq_scale_train = 1
llm_load_print_meta: n_ctx_orig_yarn  = 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: ssm_dt_b_c_rms   = 0
llm_load_print_meta: model type       = ?B
llm_load_print_meta: model ftype      = Q4_K - Small
llm_load_print_meta: model params     = 122,61 B
llm_load_print_meta: model size       = 64,79 GiB (4,54 BPW) 
llm_load_print_meta: general.name     = Mistral Large Instruct 2407
llm_load_print_meta: BOS token        = 1 '<s>'
llm_load_print_meta: EOS token        = 2 '</s>'
llm_load_print_meta: UNK token        = 0 '<unk>'
llm_load_print_meta: LF token         = 781 '<0x0A>'
llm_load_print_meta: max token length = 48
llm_load_tensors: ggml ctx size =    0,37 MiB
llm_load_tensors:        CPU buffer size = 66347,30 MiB
....................................................................................................
llama_new_context_with_model: n_ctx      = 32768
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  = 1000000,0
llama_new_context_with_model: freq_scale = 1
llama_kv_cache_init:        CPU KV buffer size = 11264,00 MiB
llama_new_context_with_model: KV self size  = 11264,00 MiB, K (f16): 5632,00 MiB, V (f16): 5632,00 MiB
llama_new_context_with_model:        CPU  output buffer size =     0,13 MiB
llama_new_context_with_model:        CPU compute buffer size =  6304,01 MiB
llama_new_context_with_model: graph nodes  = 2822
llama_new_context_with_model: graph splits = 1
llama_init_from_gpt_params: warming up the model with an empty run - please wait ... (--no-warmup to disable)
main: llama threadpool init, n_threads = 16

system_info: n_threads = 16 (n_threads_batch = 16) / 32 | AVX = 1 | AVX_VNNI = 0 | AVX2 = 1 | AVX512 = 0 | AVX512_VBMI = 0 | AVX512_VNNI = 0 | AVX512_BF16 = 0 | FMA = 1 | NEON = 0 | SVE = 0 | ARM_FMA = 0 | F16C = 1 | FP16_VA = 0 | RISCV_VECT = 0 | WASM_SIMD = 0 | BLAS = 0 | SSE3 = 1 | SSSE3 = 1 | VSX = 0 | MATMUL_INT8 = 0 | LLAMAFILE = 1 | 

sampler seed: 3428396780
sampler params: 
	repeat_last_n = 64, repeat_penalty = 1,000, frequency_penalty = 0,000, presence_penalty = 0,000
	top_k = 40, tfs_z = 0,100, top_p = 0,950, min_p = 0,050, typical_p = 1,000, temp = 0,800
	mirostat = 0, mirostat_lr = 0,100, mirostat_ent = 5,000
sampler chain: logits -> logit-bias -> penalties -> top-k -> tail-free -> typical -> top-p -> min-p -> temp-ext -> softmax -> dist 
generate: n_ctx = 32768, n_batch = 2048, n_predict = -1, n_keep = 1

src/llama-sampling.cpp:66: GGML_ASSERT(cur_p->size > 0) failed

[Septa2112]

Enabling TFS adds the llama_sampler_tail_free_apply to the sampler chain. A low tfs_z value can result in the number of output tokens being reduced to 0, which can trigger the error.

However, in below code

https://github.com/ggerganov/llama.cpp/blob/e62e9789cda3bf5573a747e55ec2a7ee32908f56/src/llama-sampling.cpp#L760-L772

there should theoretically be a min_keep setting to ensure a minimum number of tokens are generated. For some reason, though, it seems that the min_keep option has been removed, and the current default is set to 0.

As a result, cur_p->size is now updated to 0, leading to an error when the subsequent GGML_ASSERT(cur_p->size > 0); check is triggered.

And I tried to modify default value of min_keep in gpt_sampler_params from 0 to 1 and it appears to be working normally.

[github-actions]

This issue was closed because it has been inactive for 14 days since being marked as stale.