Misc. bug: inconsistent locale for printing GGUF kv data across examples #10613
Description
Name and Version
./build/bin/llama-cli --version
version: 4232 (6acce39)
built with cc (GCC) 14.2.1 20240910 for x86_64-pc-linux-gnu
Operating systems
Linux
Which llama.cpp modules do you know to be affected?
llama-cli, Other (Please specify in the next section)
Problem description & steps to reproduce
I am using a Linux PC with the locale set like this:
> locale
LANG=en_US.UTF-8
LC_CTYPE="en_US.UTF-8"
LC_NUMERIC=de_DE.UTF-8
LC_TIME=de_DE.UTF-8
LC_COLLATE="en_US.UTF-8"
LC_MONETARY=de_DE.UTF-8
LC_MESSAGES="en_US.UTF-8"
LC_PAPER=de_DE.UTF-8
LC_NAME=de_DE.UTF-8
LC_ADDRESS=de_DE.UTF-8
LC_TELEPHONE=de_DE.UTF-8
LC_MEASUREMENT=de_DE.UTF-8
LC_IDENTIFICATION=de_DE.UTF-8
LC_ALL=The way floating point numbers from the model GGUF kv data are printed is inconsistent depending on which binary I run.
llama_cli prints them with a point, llama-perplexity prints them with a comma.
It may make sense to completely ignore any locale set by the user and just always use points.
Honestly this is a very minor issue though.
First Bad Commit
No response
Relevant log output
/home/johannesg/Projects/llama.cpp [git::master *] [johannesg@johannes-pc] [11:43]
> export model_name=stories-260k && export quantization=f32
/home/johannesg/Projects/llama.cpp [git::master *] [johannesg@johannes-pc] [11:43]
> build/bin/llama-cli --model models/opt/${model_name}-${quantization}.gguf -n 64
build: 4232 (6acce397) with cc (GCC) 14.2.1 20240910 for x86_64-pc-linux-gnu
main: llama backend init
main: load the model and apply lora adapter, if any
llama_model_loader: loaded meta data with 19 key-value pairs and 48 tensors from models/opt/stories-260k-f32.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: tokenizer.ggml.tokens arr[str,512] = ["<unk>", "<s>", "</s>", "<0x00>", "<...
llama_model_loader: - kv 1: tokenizer.ggml.scores arr[f32,512] = [0,000000, 0,000000, 0,000000, 0,0000...
llama_model_loader: - kv 2: tokenizer.ggml.token_type arr[i32,512] = [2, 3, 3, 6, 6, 6, 6, 6, 6, 6, 6, 6, ...
llama_model_loader: - kv 3: tokenizer.ggml.model str = llama
llama_model_loader: - kv 4: general.architecture str = llama
llama_model_loader: - kv 5: general.name str = llama
llama_model_loader: - kv 6: tokenizer.ggml.unknown_token_id u32 = 0
llama_model_loader: - kv 7: tokenizer.ggml.bos_token_id u32 = 1
llama_model_loader: - kv 8: tokenizer.ggml.eos_token_id u32 = 2
llama_model_loader: - kv 9: tokenizer.ggml.seperator_token_id u32 = 4294967295
llama_model_loader: - kv 10: tokenizer.ggml.padding_token_id u32 = 4294967295
llama_model_loader: - kv 11: llama.context_length u32 = 128
llama_model_loader: - kv 12: llama.embedding_length u32 = 64
llama_model_loader: - kv 13: llama.feed_forward_length u32 = 172
llama_model_loader: - kv 14: llama.attention.head_count u32 = 8
llama_model_loader: - kv 15: llama.attention.head_count_kv u32 = 4
llama_model_loader: - kv 16: llama.block_count u32 = 5
llama_model_loader: - kv 17: llama.rope.dimension_count u32 = 8
llama_model_loader: - kv 18: llama.attention.layer_norm_rms_epsilon f32 = 0,000010
llama_model_loader: - type f32: 48 tensors
llm_load_vocab: bad special token: 'tokenizer.ggml.seperator_token_id' = 4294967295d, using default id -1
llm_load_vocab: bad special token: 'tokenizer.ggml.padding_token_id' = 4294967295d, using default id -1
llm_load_vocab: special_eos_id is not in special_eog_ids - the tokenizer config may be incorrect
llm_load_vocab: special tokens cache size = 3
llm_load_vocab: token to piece cache size = 0,0008 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 = 512
llm_load_print_meta: n_merges = 0
llm_load_print_meta: vocab_only = 0
llm_load_print_meta: n_ctx_train = 128
llm_load_print_meta: n_embd = 64
llm_load_print_meta: n_layer = 5
llm_load_print_meta: n_head = 8
llm_load_print_meta: n_head_kv = 4
llm_load_print_meta: n_rot = 8
llm_load_print_meta: n_swa = 0
llm_load_print_meta: n_embd_head_k = 8
llm_load_print_meta: n_embd_head_v = 8
llm_load_print_meta: n_gqa = 2
llm_load_print_meta: n_embd_k_gqa = 32
llm_load_print_meta: n_embd_v_gqa = 32
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 = 172
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 = 10000,0
llm_load_print_meta: freq_scale_train = 1
llm_load_print_meta: n_ctx_orig_yarn = 128
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 = all F32 (guessed)
llm_load_print_meta: model params = 292,80 K
llm_load_print_meta: model size = 1,12 MiB (32,00 BPW)
llm_load_print_meta: general.name = llama
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 = 13 '<0x0A>'
llm_load_print_meta: EOG token = 2 '</s>'
llm_load_print_meta: max token length = 9
llm_load_tensors: CPU_Mapped model buffer size = 1,12 MiB
...................................
llama_new_context_with_model: n_seq_max = 1
llama_new_context_with_model: n_ctx = 4096
llama_new_context_with_model: n_ctx_per_seq = 4096
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_new_context_with_model: n_ctx_pre_seq (4096) > n_ctx_train (128) -- possible training context overflow
llama_kv_cache_init: CPU KV buffer size = 2,50 MiB
llama_new_context_with_model: KV self size = 2,50 MiB, K (f16): 1,25 MiB, V (f16): 1,25 MiB
llama_new_context_with_model: CPU output buffer size = 0,00 MiB
llama_new_context_with_model: CPU compute buffer size = 72,51 MiB
llama_new_context_with_model: graph nodes = 166
llama_new_context_with_model: graph splits = 1
common_init_from_params: warming up the model with an empty run - please wait ... (--no-warmup to disable)
main: llama threadpool init, n_threads = 16
main: model was trained on only 128 context tokens (4096 specified)
system_info: n_threads = 16 (n_threads_batch = 16) / 32 | CPU : SSE3 = 1 | SSSE3 = 1 | AVX = 1 | AVX2 = 1 | F16C = 1 | FMA = 1 | LLAMAFILE = 1 | AARCH64_REPACK = 1 |
sampler seed: 2284000892
sampler params:
repeat_last_n = 64, repeat_penalty = 1,000, frequency_penalty = 0,000, presence_penalty = 0,000
dry_multiplier = 0,000, dry_base = 1,750, dry_allowed_length = 2, dry_penalty_last_n = -1
top_k = 40, top_p = 0,950, min_p = 0,050, xtc_probability = 0,000, xtc_threshold = 0,100, typical_p = 1,000, temp = 0,800
mirostat = 0, mirostat_lr = 0,100, mirostat_ent = 5,000
sampler chain: logits -> logit-bias -> penalties -> dry -> top-k -> typical -> top-p -> min-p -> xtc -> temp-ext -> dist
generate: n_ctx = 4096, n_batch = 2048, n_predict = 64, n_keep = 1
Once upon a time, there was a little girl named Lily. She loved to play with her toys and explore the woods. One day, her mommy told her they were going on a big road with long ha
llama_perf_sampler_print: sampling time = 0,57 ms / 65 runs ( 0,01 ms per token, 114235,50 tokens per second)
llama_perf_context_print: load time = 4,01 ms
llama_perf_context_print: prompt eval time = 0,00 ms / 1 tokens ( 0,00 ms per token, inf tokens per second)
llama_perf_context_print: eval time = 15,61 ms / 64 runs ( 0,24 ms per token, 4099,67 tokens per second)
llama_perf_context_print: total time = 16,64 ms / 65 tokens
/home/johannesg/Projects/llama.cpp [git::master *] [johannesg@johannes-pc] [11:43]
> build/bin/llama-perplexity --model models/opt/${model_name}-${quantization}.gguf -f wikitext-2-raw/wiki.test.raw -c 128 --chunks 1
build: 4232 (6acce397) with cc (GCC) 14.2.1 20240910 for x86_64-pc-linux-gnu
llama_model_loader: loaded meta data with 19 key-value pairs and 48 tensors from models/opt/stories-260k-f32.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: tokenizer.ggml.tokens arr[str,512] = ["<unk>", "<s>", "</s>", "<0x00>", "<...
llama_model_loader: - kv 1: tokenizer.ggml.scores arr[f32,512] = [0.000000, 0.000000, 0.000000, 0.0000...
llama_model_loader: - kv 2: tokenizer.ggml.token_type arr[i32,512] = [2, 3, 3, 6, 6, 6, 6, 6, 6, 6, 6, 6, ...
llama_model_loader: - kv 3: tokenizer.ggml.model str = llama
llama_model_loader: - kv 4: general.architecture str = llama
llama_model_loader: - kv 5: general.name str = llama
llama_model_loader: - kv 6: tokenizer.ggml.unknown_token_id u32 = 0
llama_model_loader: - kv 7: tokenizer.ggml.bos_token_id u32 = 1
llama_model_loader: - kv 8: tokenizer.ggml.eos_token_id u32 = 2
llama_model_loader: - kv 9: tokenizer.ggml.seperator_token_id u32 = 4294967295
llama_model_loader: - kv 10: tokenizer.ggml.padding_token_id u32 = 4294967295
llama_model_loader: - kv 11: llama.context_length u32 = 128
llama_model_loader: - kv 12: llama.embedding_length u32 = 64
llama_model_loader: - kv 13: llama.feed_forward_length u32 = 172
llama_model_loader: - kv 14: llama.attention.head_count u32 = 8
llama_model_loader: - kv 15: llama.attention.head_count_kv u32 = 4
llama_model_loader: - kv 16: llama.block_count u32 = 5
llama_model_loader: - kv 17: llama.rope.dimension_count u32 = 8
llama_model_loader: - kv 18: llama.attention.layer_norm_rms_epsilon f32 = 0.000010
llama_model_loader: - type f32: 48 tensors
llm_load_vocab: bad special token: 'tokenizer.ggml.seperator_token_id' = 4294967295d, using default id -1
llm_load_vocab: bad special token: 'tokenizer.ggml.padding_token_id' = 4294967295d, using default id -1
llm_load_vocab: special_eos_id is not in special_eog_ids - the tokenizer config may be incorrect
llm_load_vocab: special tokens cache size = 3
llm_load_vocab: token to piece cache size = 0.0008 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 = 512
llm_load_print_meta: n_merges = 0
llm_load_print_meta: vocab_only = 0
llm_load_print_meta: n_ctx_train = 128
llm_load_print_meta: n_embd = 64
llm_load_print_meta: n_layer = 5
llm_load_print_meta: n_head = 8
llm_load_print_meta: n_head_kv = 4
llm_load_print_meta: n_rot = 8
llm_load_print_meta: n_swa = 0
llm_load_print_meta: n_embd_head_k = 8
llm_load_print_meta: n_embd_head_v = 8
llm_load_print_meta: n_gqa = 2
llm_load_print_meta: n_embd_k_gqa = 32
llm_load_print_meta: n_embd_v_gqa = 32
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 = 172
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 = 10000.0
llm_load_print_meta: freq_scale_train = 1
llm_load_print_meta: n_ctx_orig_yarn = 128
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 = all F32 (guessed)
llm_load_print_meta: model params = 292.80 K
llm_load_print_meta: model size = 1.12 MiB (32.00 BPW)
llm_load_print_meta: general.name = llama
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 = 13 '<0x0A>'
llm_load_print_meta: EOG token = 2 '</s>'
llm_load_print_meta: max token length = 9
llm_load_tensors: CPU_Mapped model buffer size = 1.12 MiB
...................................
llama_new_context_with_model: n_seq_max = 16
llama_new_context_with_model: n_ctx = 2048
llama_new_context_with_model: n_ctx_per_seq = 128
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: CPU KV buffer size = 1.25 MiB
llama_new_context_with_model: KV self size = 1.25 MiB, K (f16): 0.62 MiB, V (f16): 0.62 MiB
llama_new_context_with_model: CPU output buffer size = 0.03 MiB
llama_new_context_with_model: CPU compute buffer size = 36.51 MiB
llama_new_context_with_model: graph nodes = 166
llama_new_context_with_model: graph splits = 1
common_init_from_params: warming up the model with an empty run - please wait ... (--no-warmup to disable)
main: model was trained on only 128 context tokens (2048 specified)
system_info: n_threads = 16 (n_threads_batch = 16) / 32 | CPU : SSE3 = 1 | SSSE3 = 1 | AVX = 1 | AVX2 = 1 | F16C = 1 | FMA = 1 | LLAMAFILE = 1 | AARCH64_REPACK = 1 |
perplexity: tokenizing the input ..
perplexity: tokenization took 190.742 ms
perplexity: calculating perplexity over 1 chunks, n_ctx=128, batch_size=2048, n_seq=16
perplexity: 0.00 seconds per pass - ETA 0.00 minutes
[1]75.2432,
Final estimate: PPL = 75.2432 +/- 40.48787
llama_perf_context_print: load time = 4.57 ms
llama_perf_context_print: prompt eval time = 2.19 ms / 128 tokens ( 0.02 ms per token, 58500.91 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 195.21 ms / 129 tokens