bf16 GGUF fails with GGML_ASSERT on CUDA

[ddh0]

Hi!

I have just cloned the latest llama.cpp repo, and I'm getting this error when trying to run a BF16 GGUF model with llama.cpp:

GGML_ASSERT: ggml-cuda.cu:1277: to_fp32_cuda != nullptr

I am using ddh0/Meta-Llama-3-8B-Instruct-bf16-GGUF, which I made myself, and I have confirmed to work on Metal.

This is the command I'm running:

./llama.cpp/main -m ./models/Meta-Llama-3-8B-Instruct-bf16.gguf -ngl 0 -nkvo -c 8192 -n 1 -p 'nfuiresnfiuesnfuisnfuiesnfiusnfiusenfiusenfiusnfiusenfuisenfiusenfiusenfuisenfiusenfuisenfiuenfisenfuesncuesnfaensifonqfnoeqiwnfiowinefioewnfowienfoiewtgowefeioncwncioenciencisncenicsnicneicnejfoiwejfoiwefoewifnenfoiewnfiohgewoiewhdcmioesiuoghscnklesioughsmkcwkhgiowenweiopdciuewfuowncoiwenfiuwefwnueoicwe'

Important note:

The program only crashes with batch size 32 or greater -- if I only do text generation and small prompts, it works fine.

The full output of the command is here:

Click to expand full llama.cpp output

./llama.cpp/main -m ./models/Meta-Llama-3-8B-Instruct-bf16.gguf -ngl 0 -nkvo -c 8192 -n 1 -p 'nfuiresnfiuesnfuisnfuiesnfiusnfiusenfiusenfiusnfiusenfuisenfiusenfiusenfuisenfiusenfuisenfiuenfisenfuesncuesnfaensifonqfnoeqiwnfiowinefioewnfowienfoiewtgowefeioncwncioenciencisncenicsnicneicnejfoiwejfoiwefoewifnenfoiewnfiohgewoiewhdcmioesiuoghscnklesioughsmkcwkhgiowenweiopdciuewfuowncoiwenfiuwefwnueoicwe'
Log start
main: build = 2852 (fae9d234)
main: built with cc (Debian 12.2.0-14) 12.2.0 for x86_64-linux-gnu
main: seed  = 1715439608
llama_model_loader: loaded meta data with 22 key-value pairs and 291 tensors from ./models/Meta-Llama-3-8B-Instruct-bf16.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              = Meta-Llama-3-8B-Instruct
llama_model_loader: - kv   2:                          llama.block_count u32              = 32
llama_model_loader: - kv   3:                       llama.context_length u32              = 8192
llama_model_loader: - kv   4:                     llama.embedding_length u32              = 4096
llama_model_loader: - kv   5:                  llama.feed_forward_length u32              = 14336
llama_model_loader: - kv   6:                 llama.attention.head_count u32              = 32
llama_model_loader: - kv   7:              llama.attention.head_count_kv u32              = 8
llama_model_loader: - kv   8:                       llama.rope.freq_base f32              = 500000.000000
llama_model_loader: - kv   9:     llama.attention.layer_norm_rms_epsilon f32              = 0.000010
llama_model_loader: - kv  10:                          general.file_type u32              = 32
llama_model_loader: - kv  11:                           llama.vocab_size u32              = 128256
llama_model_loader: - kv  12:                 llama.rope.dimension_count u32              = 128
llama_model_loader: - kv  13:                       tokenizer.ggml.model str              = gpt2
llama_model_loader: - kv  14:                         tokenizer.ggml.pre str              = llama-bpe
llama_model_loader: - kv  15:                      tokenizer.ggml.tokens arr[str,128256]  = ["!", "\"", "#", "$", "%", "&", "'", ...
llama_model_loader: - kv  16:                  tokenizer.ggml.token_type arr[i32,128256]  = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
llama_model_loader: - kv  17:                      tokenizer.ggml.merges arr[str,280147]  = ["Ġ Ġ", "Ġ ĠĠĠ", "ĠĠ ĠĠ", "...
llama_model_loader: - kv  18:                tokenizer.ggml.bos_token_id u32              = 128000
llama_model_loader: - kv  19:                tokenizer.ggml.eos_token_id u32              = 128009
llama_model_loader: - kv  20:                    tokenizer.chat_template str              = {% set loop_messages = messages %}{% ...
llama_model_loader: - kv  21:               general.quantization_version u32              = 2
llama_model_loader: - type  f32:   65 tensors
llama_model_loader: - type bf16:  226 tensors
llm_load_vocab: special tokens definition check successful ( 256/128256 ).
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: n_ctx_train      = 8192
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         = 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_yarn_orig_ctx  = 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: model type       = 8B
llm_load_print_meta: model ftype      = BF16
llm_load_print_meta: model params     = 8.03 B
llm_load_print_meta: model size       = 14.96 GiB (16.00 BPW) 
llm_load_print_meta: general.name     = Meta-Llama-3-8B-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|>'
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 4060 Ti, compute capability 8.9, VMM: yes
llm_load_tensors: ggml ctx size =    0.15 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 = 15317.02 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  = 500000.0
llama_new_context_with_model: freq_scale = 1
llama_kv_cache_init:  CUDA_Host KV buffer size =  1024.00 MiB
llama_new_context_with_model: KV self size  = 1024.00 MiB, K (f16):  512.00 MiB, V (f16):  512.00 MiB
llama_new_context_with_model:  CUDA_Host  output buffer size =     0.49 MiB
llama_new_context_with_model:      CUDA0 compute buffer size =  1260.50 MiB
llama_new_context_with_model:  CUDA_Host compute buffer size =   540.01 MiB
llama_new_context_with_model: graph nodes  = 1030
llama_new_context_with_model: graph splits = 324

system_info: n_threads = 8 / 16 | AVX = 1 | AVX_VNNI = 0 | AVX2 = 1 | AVX512 = 1 | AVX512_VBMI = 1 | AVX512_VNNI = 1 | 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 | LLAMAFILE = 1 | 
sampling: 
	repeat_last_n = 64, repeat_penalty = 1.000, frequency_penalty = 0.000, presence_penalty = 0.000
	top_k = 40, tfs_z = 1.000, 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
sampling order: 
CFG -> Penalties -> top_k -> tfs_z -> typical_p -> top_p -> min_p -> temperature 
generate: n_ctx = 8192, n_batch = 2048, n_predict = 1, n_keep = 0


<|begin_of_text|>nfuiresnfiuesnfuisnfuiesnfiusnfiusenfiusenfiusnfiusenfuisenfiusenfiusenfuisenfiusenfuisenfiuenfisenfuesncuesnfaensifonqfnoeqiwnfiowinefioewnfowienfoiewtgowefeioncwncioenciencisncenicsnicneicnejfoiwejfoiwefoewifnenfoiewnfiohgewoiewhdcmioesiuoghscnklesioughsmkcwkhgiowenweiopdciuewfuowncoiwenfiuwefwnueoicweGGML_ASSERT: ggml-cuda.cu:1277: to_fp32_cuda != nullptr
Aborted

I'd welcome any insight. Please let me know if I can provide any other information.

[dranger003]

I don't think BF16 works with CUDA, i.e. #7211.

[ddh0]

I will close this issue as a duplicate and refer to the existing issue #7211 - thanks.