Eval bug: phi 4 - input is empty

[mirek190]

Name and Version

build: 4451 (d9feae1) with MSVC 19.29.30157.0 for

Operating systems

Windows

GGML backends

CUDA

Hardware

Ryzen 7950x3d + RTX 3090

Models

phi 4

Problem description & steps to reproduce

phi 4 - input is empty

Just load model

First Bad Commit

No response

Relevant log output

llama-cli.exe --model models/new3/phi-4-Q8_0.gguf --color --threads 30 --keep -1 --n-predict -1 --ctx-size 16384 --interactive -ngl 99 --simple-io -e --multiline-input --no-display-prompt --conversation --no-mmap
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 3090, compute capability 8.6, VMM: yes
build: 4451 (d9feae1c) with MSVC 19.29.30157.0 for
main: llama backend init
main: load the model and apply lora adapter, if any
llama_model_load_from_file: using device CUDA0 (NVIDIA GeForce RTX 3090) - 23306 MiB free
llama_model_loader: loaded meta data with 37 key-value pairs and 243 tensors from models/new3/phi-4-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              = phi3
llama_model_loader: - kv   1:                               general.type str              = model
llama_model_loader: - kv   2:                               general.name str              = Phi 4
llama_model_loader: - kv   3:                            general.version str              = 4
llama_model_loader: - kv   4:                       general.organization str              = Microsoft
llama_model_loader: - kv   5:                           general.basename str              = phi
llama_model_loader: - kv   6:                         general.size_label str              = 15B
llama_model_loader: - kv   7:                            general.license str              = mit
llama_model_loader: - kv   8:                       general.license.link str              = https://huggingface.co/microsoft/phi-...
llama_model_loader: - kv   9:                               general.tags arr[str,7]       = ["phi", "nlp", "math", "code", "chat"...
llama_model_loader: - kv  10:                          general.languages arr[str,1]       = ["en"]
llama_model_loader: - kv  11:                        phi3.context_length u32              = 16384
llama_model_loader: - kv  12:  phi3.rope.scaling.original_context_length u32              = 16384
llama_model_loader: - kv  13:                      phi3.embedding_length u32              = 5120
llama_model_loader: - kv  14:                   phi3.feed_forward_length u32              = 17920
llama_model_loader: - kv  15:                           phi3.block_count u32              = 40
llama_model_loader: - kv  16:                  phi3.attention.head_count u32              = 40
llama_model_loader: - kv  17:               phi3.attention.head_count_kv u32              = 10
llama_model_loader: - kv  18:      phi3.attention.layer_norm_rms_epsilon f32              = 0.000010
llama_model_loader: - kv  19:                  phi3.rope.dimension_count u32              = 128
llama_model_loader: - kv  20:                        phi3.rope.freq_base f32              = 250000.000000
llama_model_loader: - kv  21:                          general.file_type u32              = 7
llama_model_loader: - kv  22:              phi3.attention.sliding_window u32              = 0
llama_model_loader: - kv  23:                       tokenizer.ggml.model str              = gpt2
llama_model_loader: - kv  24:                         tokenizer.ggml.pre str              = dbrx
llama_model_loader: - kv  25:                      tokenizer.ggml.tokens arr[str,100352]  = ["!", "\"", "#", "$", "%", "&", "'", ...
llama_model_loader: - kv  26:                  tokenizer.ggml.token_type arr[i32,100352]  = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
llama_model_loader: - kv  27:                      tokenizer.ggml.merges arr[str,100000]  = ["Ġ Ġ", "ĠĠ ĠĠ", "i n", "Ġ t",...
llama_model_loader: - kv  28:                tokenizer.ggml.bos_token_id u32              = 100257
llama_model_loader: - kv  29:                tokenizer.ggml.eos_token_id u32              = 100257
llama_model_loader: - kv  30:            tokenizer.ggml.padding_token_id u32              = 100257
llama_model_loader: - kv  31:                    tokenizer.chat_template str              = {% for message in messages %}{% if (m...
llama_model_loader: - kv  32:               general.quantization_version u32              = 2
llama_model_loader: - kv  33:                      quantize.imatrix.file str              = /models_out/phi-4-GGUF/phi-4.imatrix
llama_model_loader: - kv  34:                   quantize.imatrix.dataset str              = /training_dir/calibration_datav3.txt
llama_model_loader: - kv  35:             quantize.imatrix.entries_count i32              = 160
llama_model_loader: - kv  36:              quantize.imatrix.chunks_count i32              = 127
llama_model_loader: - type  f32:   81 tensors
llama_model_loader: - type q8_0:  162 tensors
llm_load_vocab: special tokens cache size = 96
llm_load_vocab: token to piece cache size = 0.6151 MB
llm_load_print_meta: format           = GGUF V3 (latest)
llm_load_print_meta: arch             = phi3
llm_load_print_meta: vocab type       = BPE
llm_load_print_meta: n_vocab          = 100352
llm_load_print_meta: n_merges         = 100000
llm_load_print_meta: vocab_only       = 0
llm_load_print_meta: n_ctx_train      = 16384
llm_load_print_meta: n_embd           = 5120
llm_load_print_meta: n_layer          = 40
llm_load_print_meta: n_head           = 40
llm_load_print_meta: n_head_kv        = 10
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            = 4
llm_load_print_meta: n_embd_k_gqa     = 1280
llm_load_print_meta: n_embd_v_gqa     = 1280
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             = 17920
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  = 250000.0
llm_load_print_meta: freq_scale_train = 1
llm_load_print_meta: n_ctx_orig_yarn  = 16384
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       = 14B
llm_load_print_meta: model ftype      = Q8_0
llm_load_print_meta: model params     = 14.66 B
llm_load_print_meta: model size       = 14.51 GiB (8.50 BPW)
llm_load_print_meta: general.name     = Phi 4
llm_load_print_meta: BOS token        = 100257 '<|endoftext|>'
llm_load_print_meta: EOS token        = 100257 '<|endoftext|>'
llm_load_print_meta: EOT token        = 100265 '<|im_end|>'
llm_load_print_meta: PAD token        = 100257 '<|endoftext|>'
llm_load_print_meta: LF token         = 128 'Ä'
llm_load_print_meta: FIM PRE token    = 100258 '<|fim_prefix|>'
llm_load_print_meta: FIM SUF token    = 100260 '<|fim_suffix|>'
llm_load_print_meta: FIM MID token    = 100259 '<|fim_middle|>'
llm_load_print_meta: EOG token        = 100257 '<|endoftext|>'
llm_load_print_meta: EOG token        = 100265 '<|im_end|>'
llm_load_print_meta: max token length = 256
llm_load_tensors: offloading 40 repeating layers to GPU
llm_load_tensors: offloading output layer to GPU
llm_load_tensors: offloaded 41/41 layers to GPU
llm_load_tensors:    CUDA_Host model buffer size =   520.62 MiB
llm_load_tensors:        CUDA0 model buffer size = 14334.71 MiB
.....................................................................................
llama_new_context_with_model: n_seq_max     = 1
llama_new_context_with_model: n_ctx         = 16384
llama_new_context_with_model: n_ctx_per_seq = 16384
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     = 250000.0
llama_new_context_with_model: freq_scale    = 1
llama_kv_cache_init: kv_size = 16384, offload = 1, type_k = 'f16', type_v = 'f16', n_layer = 40, can_shift = 1
llama_kv_cache_init:      CUDA0 KV buffer size =  3200.00 MiB
llama_new_context_with_model: KV self size  = 3200.00 MiB, K (f16): 1600.00 MiB, V (f16): 1600.00 MiB
llama_new_context_with_model:  CUDA_Host  output buffer size =     0.38 MiB
llama_new_context_with_model:      CUDA0 compute buffer size =  1357.00 MiB
llama_new_context_with_model:  CUDA_Host compute buffer size =    42.01 MiB
llama_new_context_with_model: graph nodes  = 1606
llama_new_context_with_model: graph splits = 2
common_init_from_params: setting dry_penalty_last_n to ctx_size = 16384
common_init_from_params: warming up the model with an empty run - please wait ... (--no-warmup to disable)
main: llama threadpool init, n_threads = 30
main: chat template example:
<|im_start|>system<|im_sep|>You are a helpful assistant<|im_end|><|im_start|>user<|im_sep|>Hello<|im_end|><|im_start|>assistant<|im_sep|>Hi there<|im_end|><|im_start|>user<|im_sep|>How are you?<|im_end|><|im_start|>assistant<|im_sep|>

system_info: n_threads = 30 (n_threads_batch = 30) / 32 | CUDA : ARCHS = 520,610,700,750 | USE_GRAPHS = 1 | PEER_MAX_BATCH_SIZE = 128 | CPU : SSE3 = 1 | SSSE3 = 1 | AVX = 1 | AVX2 = 1 | F16C = 1 | FMA = 1 | LLAMAFILE = 1 | OPENMP = 1 | AARCH64_REPACK = 1 |

input is empty

[ngxson]

You need to add system message: -p "You are a helpful assistant"