Bug: BigLlama-3.1-681B-Instruct requires llama_model_max_nodes to return a higher value

[nicoboss]

What happened?

This issue is caused by a reappearance of issue #8615 (PR #8622). I recommend reading them for more information about this problem.

For Meta-Llama-3.1-405B-Instruct it turned out that the default llama_model_max_nodes of 8192 is still enough. For its self-merge available under https://huggingface.co/mlabonne/BigLlama-3.1-681B-Instruct (GGUFs available under https://huggingface.co/mradermacher/BigLlama-3.1-681B-Instruct-GGUF/tree/main) it is unfortunately not.

For this issue to be fixed the commented out logic under

llama.cpp/src/llama.cpp

Lines 3571 to 3573 in 3071c0a

	//if (model.arch == LLM_ARCH_LLAMA && model.hparams.n_layer > ??) { // llama-3 405B
	// return 32768;
	//}

needs to be reenabled and set to something like model.hparams.n_layer > 200 for this model to work. Maybe a good approach would be having 0-200 return 8192, >200 return 16384 and >400 return 32768. To play around with this model I made the llama_model_max_nodes function always return 16384 which fixed the issue.

Name and Version

version: 3557 (3071c0a)
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 /bpool/BigLlama-3.1-681B-Instruct.Q4_K_S.gguf -p "I believe the meaning of life is" -c 128 -n 64 -ngl 0

main: build = 3557 (3071c0a5)                                                                                                                                
main: built with cc (Debian 12.2.0-14) 12.2.0 for x86_64-linux-gnu                                                                                           
main: seed  = 1723208630                                                                                                                                     
llama_model_loader: loaded meta data with 32 key-value pairs and 1894 tensors from /bpool/BigLlama-3.1-681B-Instruct.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              = Meta Llama 3.1 405B Instruct                                     
llama_model_loader: - kv   3:                       general.organization str              = Meta Llama                                                       
llama_model_loader: - kv   4:                           general.finetune str              = Instruct
llama_model_loader: - kv   5:                           general.basename str              = Meta-Llama-3.1                                                   
llama_model_loader: - kv   6:                         general.size_label str              = 405B                                                             
llama_model_loader: - kv   7:                   general.base_model.count u32              = 1                                                                
llama_model_loader: - kv   8:                  general.base_model.0.name str              = Meta Llama 3.1 405B Instruct                                     
llama_model_loader: - kv   9:          general.base_model.0.organization str              = Meta Llama                                                       
llama_model_loader: - kv  10:              general.base_model.0.repo_url str              = https://huggingface.co/meta-llama/Met...                         
llama_model_loader: - kv  11:                               general.tags arr[str,2]       = ["mergekit", "merge"]                                            
llama_model_loader: - kv  12:                          llama.block_count u32              = 210                                                              
llama_model_loader: - kv  13:                       llama.context_length u32              = 131072                                                           
llama_model_loader: - kv  14:                     llama.embedding_length u32              = 16384                                                            
llama_model_loader: - kv  15:                  llama.feed_forward_length u32              = 53248                                                            
llama_model_loader: - kv  16:                 llama.attention.head_count u32              = 128                                                              
llama_model_loader: - kv  17:              llama.attention.head_count_kv u32              = 16                                                               
llama_model_loader: - kv  18:                       llama.rope.freq_base f32              = 500000.000000                                                    
llama_model_loader: - kv  19:     llama.attention.layer_norm_rms_epsilon f32              = 0.000010                                                         
llama_model_loader: - kv  20:                          general.file_type u32              = 14                                                               
llama_model_loader: - kv  21:                           llama.vocab_size u32              = 128256                                                           
llama_model_loader: - kv  22:                 llama.rope.dimension_count u32              = 128                                                              
llama_model_loader: - kv  23:                       tokenizer.ggml.model str              = gpt2                                                             
llama_model_loader: - kv  24:                         tokenizer.ggml.pre str              = llama-bpe                                                        
llama_model_loader: - kv  25:                      tokenizer.ggml.tokens arr[str,128256]  = ["!", "\"", "#", "$", "%", "&", "'", ...                         
llama_model_loader: - kv  26:                  tokenizer.ggml.token_type arr[i32,128256]  = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...                         
llama_model_loader: - kv  27:                      tokenizer.ggml.merges arr[str,280147]  = ["Ġ Ġ", "Ġ ĠĠĠ", "ĠĠ ĠĠ", ...                                   
llama_model_loader: - kv  28:                tokenizer.ggml.bos_token_id u32              = 128000                                                           
llama_model_loader: - kv  29:                tokenizer.ggml.eos_token_id u32              = 128009                                                           
llama_model_loader: - kv  30:                    tokenizer.chat_template str              = {{- bos_token }}\n{%- if custom_tools ...                        
llama_model_loader: - kv  31:               general.quantization_version u32              = 2                                                                
llama_model_loader: - type  f32:  422 tensors                                                                                                                
llama_model_loader: - type q4_K: 1441 tensors                                                                                                                
llama_model_loader: - type q5_K:   30 tensors                                                                                                                
llama_model_loader: - type q6_K:    1 tensors                                                                                                                
llm_load_vocab: special tokens cache size = 256                                                                                                              
llm_load_vocab: token to piece cache size = 0.7999 MB
llm_load_print_meta: format           = GGUF V3 (latest)                                                                                                     
llm_load_print_meta: arch             = llama                                                                                                                
llm_load_print_meta: vocab type       = BPE                                                                                                                  
llm_load_print_meta: n_vocab          = 128256                                                                                                               
llm_load_print_meta: n_merges         = 280147                                                                                                               
llm_load_print_meta: vocab_only       = 0                                                                                                                    
llm_load_print_meta: n_ctx_train      = 131072                                                                                                               
llm_load_print_meta: n_embd           = 16384                                                                                                                
llm_load_print_meta: n_layer          = 210                                                                                                                  
llm_load_print_meta: n_head           = 128                                                                                                                  
llm_load_print_meta: n_head_kv        = 16                                                                                                                   
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            = 8                                                                                                                    
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-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             = 53248                                                                                                                
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  = 500000.0                                                                                                             
llm_load_print_meta: freq_scale_train = 1                                                                                                                    
llm_load_print_meta: n_ctx_orig_yarn  = 131072                                                                                                               
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       = ?B
llm_load_print_meta: model ftype      = Q4_K - Small
llm_load_print_meta: model params     = 680.67 B
llm_load_print_meta: model size       = 359.76 GiB (4.54 BPW) 
llm_load_print_meta: general.name     = Meta Llama 3.1 405B Instruct
llm_load_print_meta: BOS token        = 128000 '<|begin_of_text|>'
llm_load_print_meta: EOS token        = 128009 '<|eot_id|>'
llm_load_print_meta: LF token         = 128 'Ä'
llm_load_print_meta: EOT token        = 128009 '<|eot_id|>'
llm_load_print_meta: max token length = 256
llm_load_tensors: ggml ctx size =    0.89 MiB
llm_load_tensors:        CPU buffer size = 368397.47 MiB
....................................................................................................
llama_new_context_with_model: n_ctx      = 128
llama_new_context_with_model: n_batch    = 128
llama_new_context_with_model: n_ubatch   = 128
llama_new_context_with_model: flash_attn = 0
llama_new_context_with_model: freq_base  = 500000.0
llama_new_context_with_model: freq_scale = 1
llama_kv_cache_init:        CPU KV buffer size =   210.00 MiB
llama_new_context_with_model: KV self size  =  210.00 MiB, K (f16):  105.00 MiB, V (f16):  105.00 MiB
llama_new_context_with_model:        CPU  output buffer size =     0.49 MiB
ggml/src/ggml-backend.c:1936: GGML_ASSERT((int)sched->hash_set.size >= measure_graph->n_nodes + measure_graph->n_leafs) failed
./llama-cli(+0x5e208)[0x5b4ba469d208]
./llama-cli(+0x60655)[0x5b4ba469f655]
./llama-cli(+0xad7c6)[0x5b4ba46ec7c6]
./llama-cli(+0x10a400)[0x5b4ba4749400]
./llama-cli(+0x1d160c)[0x5b4ba481060c]
./llama-cli(+0x3ec0e)[0x5b4ba467dc0e]
/lib/x86_64-linux-gnu/libc.so.6(+0x2724a)[0x7b12ad92f24a]
/lib/x86_64-linux-gnu/libc.so.6(__libc_start_main+0x85)[0x7b12ad92f305]
./llama-cli(+0x45521)[0x5b4ba4684521]

[slaren]

Something like model.tensors_by_name.size()*5 would likely work well with every model.

[eriktrom]

Maybe a good approach would be having 0-200 return 8192, >200 return 16384 and >400 return 32768. To play around with this model I made the llama_model_max_nodes function always return 16384 which fixed the issue.

thanks for sharing, saved a ton of time 👍