llama-cli misbehaving (changed?)

[0wwafa]

I have a colab notebook here to quantize and the test models:
https://colab.research.google.com/drive/1TcyGL60GQzsxEHu-Xlos5u8bb_6SxMa3

The simple test has always been this line:

prompt="""
Tell me the difference between thinking in humans and in LLMs.
"""
m=f'{model_name}.{q_type}.gguf'
!./build/bin/llama-cli --ignore-eos -c 4096 -m /content/$m -t $(nproc) -ngl 999 -p "User: Hi\nBot:Hi\nUser: {prompt}\nBot:"

Usually, after the initialization the models start answering. (and then even continuing on their own... which is fine).

Now ( b4762 ) instead it does this:

build: 4762 (af7747c9) 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 39 key-value pairs and 464 tensors from /content/gemma-2-Ifable-9B.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.type str              = model
llama_model_loader: - kv   2:                               general.name str              = Gemma 2 Ifable 9B
llama_model_loader: - kv   3:                       general.organization str              = Ifable
llama_model_loader: - kv   4:                           general.basename str              = gemma-2-Ifable
llama_model_loader: - kv   5:                         general.size_label str              = 9B
llama_model_loader: - kv   6:                            general.license str              = gemma
llama_model_loader: - kv   7:                      general.dataset.count u32              = 1
llama_model_loader: - kv   8:                     general.dataset.0.name str              = Gutenberg Dpo v0.1
llama_model_loader: - kv   9:                  general.dataset.0.version str              = v0.1
llama_model_loader: - kv  10:             general.dataset.0.organization str              = Jondurbin
llama_model_loader: - kv  11:                 general.dataset.0.repo_url str              = https://huggingface.co/jondurbin/gute...
llama_model_loader: - kv  12:                      gemma2.context_length u32              = 8192
llama_model_loader: - kv  13:                    gemma2.embedding_length u32              = 3584
llama_model_loader: - kv  14:                         gemma2.block_count u32              = 42
llama_model_loader: - kv  15:                 gemma2.feed_forward_length u32              = 14336
llama_model_loader: - kv  16:                gemma2.attention.head_count u32              = 16
llama_model_loader: - kv  17:             gemma2.attention.head_count_kv u32              = 8
llama_model_loader: - kv  18:    gemma2.attention.layer_norm_rms_epsilon f32              = 0.000001
llama_model_loader: - kv  19:                gemma2.attention.key_length u32              = 256
llama_model_loader: - kv  20:              gemma2.attention.value_length u32              = 256
llama_model_loader: - kv  21:              gemma2.attn_logit_softcapping f32              = 50.000000
llama_model_loader: - kv  22:             gemma2.final_logit_softcapping f32              = 30.000000
llama_model_loader: - kv  23:            gemma2.attention.sliding_window u32              = 4096
llama_model_loader: - kv  24:                       tokenizer.ggml.model str              = llama
llama_model_loader: - kv  25:                         tokenizer.ggml.pre str              = default
llama_model_loader: - kv  26:                      tokenizer.ggml.tokens arr[str,256000]  = ["<pad>", "<eos>", "<bos>", "<unk>", ...
llama_model_loader: - kv  27:                      tokenizer.ggml.scores arr[f32,256000]  = [-1000.000000, -1000.000000, -1000.00...
llama_model_loader: - kv  28:                  tokenizer.ggml.token_type arr[i32,256000]  = [3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, ...
llama_model_loader: - kv  29:                tokenizer.ggml.bos_token_id u32              = 2
llama_model_loader: - kv  30:                tokenizer.ggml.eos_token_id u32              = 1
llama_model_loader: - kv  31:            tokenizer.ggml.unknown_token_id u32              = 3
llama_model_loader: - kv  32:            tokenizer.ggml.padding_token_id u32              = 0
llama_model_loader: - kv  33:               tokenizer.ggml.add_bos_token bool             = true
llama_model_loader: - kv  34:               tokenizer.ggml.add_eos_token bool             = false
llama_model_loader: - kv  35:                    tokenizer.chat_template str              = {{ '<bos>' }}{% if messages[0]['role'...
llama_model_loader: - kv  36:            tokenizer.ggml.add_space_prefix bool             = false
llama_model_loader: - kv  37:               general.quantization_version u32              = 2
llama_model_loader: - kv  38:                          general.file_type u32              = 7
llama_model_loader: - type  f32:  169 tensors
llama_model_loader: - type  f16:    1 tensors
llama_model_loader: - type q8_0:  294 tensors
print_info: file format = GGUF V3 (latest)
print_info: file type   = Q8_0
print_info: file size   = 9.95 GiB (9.25 BPW) 
load: special_eos_id is not in special_eog_ids - the tokenizer config may be incorrect
load: special tokens cache size = 217
load: token to piece cache size = 1.6014 MB
print_info: arch             = gemma2
print_info: vocab_only       = 0
print_info: n_ctx_train      = 8192
print_info: n_embd           = 3584
print_info: n_layer          = 42
print_info: n_head           = 16
print_info: n_head_kv        = 8
print_info: n_rot            = 256
print_info: n_swa            = 4096
print_info: n_embd_head_k    = 256
print_info: n_embd_head_v    = 256
print_info: n_gqa            = 2
print_info: n_embd_k_gqa     = 2048
print_info: n_embd_v_gqa     = 2048
print_info: f_norm_eps       = 0.0e+00
print_info: f_norm_rms_eps   = 1.0e-06
print_info: f_clamp_kqv      = 0.0e+00
print_info: f_max_alibi_bias = 0.0e+00
print_info: f_logit_scale    = 0.0e+00
print_info: n_ff             = 14336
print_info: n_expert         = 0
print_info: n_expert_used    = 0
print_info: causal attn      = 1
print_info: pooling type     = 0
print_info: rope type        = 2
print_info: rope scaling     = linear
print_info: freq_base_train  = 10000.0
print_info: freq_scale_train = 1
print_info: n_ctx_orig_yarn  = 8192
print_info: rope_finetuned   = unknown
print_info: ssm_d_conv       = 0
print_info: ssm_d_inner      = 0
print_info: ssm_d_state      = 0
print_info: ssm_dt_rank      = 0
print_info: ssm_dt_b_c_rms   = 0
print_info: model type       = 9B
print_info: model params     = 9.24 B
print_info: general.name     = Gemma 2 Ifable 9B
print_info: vocab type       = SPM
print_info: n_vocab          = 256000
print_info: n_merges         = 0
print_info: BOS token        = 2 '<bos>'
print_info: EOS token        = 1 '<eos>'
print_info: EOT token        = 107 '<end_of_turn>'
print_info: UNK token        = 3 '<unk>'
print_info: PAD token        = 0 '<pad>'
print_info: LF token         = 227 '<0x0A>'
print_info: EOG token        = 1 '<eos>'
print_info: EOG token        = 107 '<end_of_turn>'
print_info: max token length = 48
load_tensors: loading model tensors, this can take a while... (mmap = true)
load_tensors: offloading 42 repeating layers to GPU
load_tensors: offloading output layer to GPU
load_tensors: offloaded 43/43 layers to GPU
load_tensors:   CPU_Mapped model buffer size = 10186.44 MiB
....................................................................................
llama_init_from_model: n_seq_max     = 1
llama_init_from_model: n_ctx         = 4096
llama_init_from_model: n_ctx_per_seq = 4096
llama_init_from_model: n_batch       = 2048
llama_init_from_model: n_ubatch      = 512
llama_init_from_model: flash_attn    = 0
llama_init_from_model: freq_base     = 10000.0
llama_init_from_model: freq_scale    = 1
llama_init_from_model: n_ctx_per_seq (4096) < n_ctx_train (8192) -- the full capacity of the model will not be utilized
llama_kv_cache_init: kv_size = 4096, offload = 1, type_k = 'f16', type_v = 'f16', n_layer = 42, can_shift = 1
llama_kv_cache_init:        CPU KV buffer size =  1344.00 MiB
llama_init_from_model: KV self size  = 1344.00 MiB, K (f16):  672.00 MiB, V (f16):  672.00 MiB
llama_init_from_model:        CPU  output buffer size =     0.98 MiB
llama_init_from_model:        CPU compute buffer size =   514.00 MiB
llama_init_from_model: graph nodes  = 1690
llama_init_from_model: graph splits = 1
common_init_from_params: added <eos> logit bias = -inf
common_init_from_params: added <end_of_turn> logit bias = -inf
common_init_from_params: setting dry_penalty_last_n to ctx_size = 4096
common_init_from_params: warming up the model with an empty run - please wait ... (--no-warmup to disable)
main: llama threadpool init, n_threads = 2
main: chat template is available, enabling conversation mode (disable it with -no-cnv)
main: chat template 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


system_info: n_threads = 2 (n_threads_batch = 2) / 2 | CPU : SSE3 = 1 | SSSE3 = 1 | AVX = 1 | AVX2 = 1 | F16C = 1 | FMA = 1 | LLAMAFILE = 1 | OPENMP = 1 | AARCH64_REPACK = 1 | 

main: interactive mode on.
sampler seed: 3895428166
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 = 4096
	top_k = 40, top_p = 0.950, min_p = 0.050, xtc_probability = 0.000, xtc_threshold = 0.100, typical_p = 1.000, top_n_sigma = -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 = -1, n_keep = 1

== Running in interactive mode. ==
 - Press Ctrl+C to interject at any time.
 - Press Return to return control to the AI.
 - To return control without starting a new line, end your input with '/'.
 - If you want to submit another line, end your input with '\'.


> 

Am I doing something wrong?

Note:
if I use b4000 everything works as usual.

[0wwafa]

For reference, this is the output I get with the same model using b4000:

build: 4000 (c02e5ab2) 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 39 key-value pairs and 464 tensors from /content/gemma-2-Ifable-9B.q8q4.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.type str              = model
llama_model_loader: - kv   2:                               general.name str              = Gemma 2 Ifable 9B
llama_model_loader: - kv   3:                       general.organization str              = Ifable
llama_model_loader: - kv   4:                           general.basename str              = gemma-2-Ifable
llama_model_loader: - kv   5:                         general.size_label str              = 9B
llama_model_loader: - kv   6:                            general.license str              = gemma
llama_model_loader: - kv   7:                      general.dataset.count u32              = 1
llama_model_loader: - kv   8:                     general.dataset.0.name str              = Gutenberg Dpo v0.1
llama_model_loader: - kv   9:                  general.dataset.0.version str              = v0.1
llama_model_loader: - kv  10:             general.dataset.0.organization str              = Jondurbin
llama_model_loader: - kv  11:                 general.dataset.0.repo_url str              = https://huggingface.co/jondurbin/gute...
llama_model_loader: - kv  12:                      gemma2.context_length u32              = 8192
llama_model_loader: - kv  13:                    gemma2.embedding_length u32              = 3584
llama_model_loader: - kv  14:                         gemma2.block_count u32              = 42
llama_model_loader: - kv  15:                 gemma2.feed_forward_length u32              = 14336
llama_model_loader: - kv  16:                gemma2.attention.head_count u32              = 16
llama_model_loader: - kv  17:             gemma2.attention.head_count_kv u32              = 8
llama_model_loader: - kv  18:    gemma2.attention.layer_norm_rms_epsilon f32              = 0.000001
llama_model_loader: - kv  19:                gemma2.attention.key_length u32              = 256
llama_model_loader: - kv  20:              gemma2.attention.value_length u32              = 256
llama_model_loader: - kv  21:              gemma2.attn_logit_softcapping f32              = 50.000000
llama_model_loader: - kv  22:             gemma2.final_logit_softcapping f32              = 30.000000
llama_model_loader: - kv  23:            gemma2.attention.sliding_window u32              = 4096
llama_model_loader: - kv  24:                       tokenizer.ggml.model str              = llama
llama_model_loader: - kv  25:                         tokenizer.ggml.pre str              = default
llama_model_loader: - kv  26:                      tokenizer.ggml.tokens arr[str,256000]  = ["<pad>", "<eos>", "<bos>", "<unk>", ...
llama_model_loader: - kv  27:                      tokenizer.ggml.scores arr[f32,256000]  = [-1000.000000, -1000.000000, -1000.00...
llama_model_loader: - kv  28:                  tokenizer.ggml.token_type arr[i32,256000]  = [3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, ...
llama_model_loader: - kv  29:                tokenizer.ggml.bos_token_id u32              = 2
llama_model_loader: - kv  30:                tokenizer.ggml.eos_token_id u32              = 1
llama_model_loader: - kv  31:            tokenizer.ggml.unknown_token_id u32              = 3
llama_model_loader: - kv  32:            tokenizer.ggml.padding_token_id u32              = 0
llama_model_loader: - kv  33:               tokenizer.ggml.add_bos_token bool             = true
llama_model_loader: - kv  34:               tokenizer.ggml.add_eos_token bool             = false
llama_model_loader: - kv  35:                    tokenizer.chat_template str              = {{ '<bos>' }}{% if messages[0]['role'...
llama_model_loader: - kv  36:            tokenizer.ggml.add_space_prefix bool             = false
llama_model_loader: - kv  37:               general.quantization_version u32              = 2
llama_model_loader: - kv  38:                          general.file_type u32              = 15
llama_model_loader: - type  f32:  169 tensors
llama_model_loader: - type q8_0:    1 tensors
llama_model_loader: - type q4_K:  252 tensors
llama_model_loader: - type q6_K:   42 tensors
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 = 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           = 3584
llm_load_print_meta: n_layer          = 42
llm_load_print_meta: n_head           = 16
llm_load_print_meta: n_head_kv        = 8
llm_load_print_meta: n_rot            = 256
llm_load_print_meta: n_swa            = 4096
llm_load_print_meta: n_embd_head_k    = 256
llm_load_print_meta: n_embd_head_v    = 256
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             = 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        = 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       = 9B
llm_load_print_meta: model ftype      = Q4_K - Medium
llm_load_print_meta: model params     = 9.24 B
llm_load_print_meta: model size       = 5.57 GiB (5.17 BPW) 
llm_load_print_meta: general.name     = Gemma 2 Ifable 9B
llm_load_print_meta: BOS token        = 2 '<bos>'
llm_load_print_meta: EOS token        = 1 '<eos>'
llm_load_print_meta: EOT token        = 107 '<end_of_turn>'
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: EOG token        = 1 '<eos>'
llm_load_print_meta: EOG token        = 107 '<end_of_turn>'
llm_load_print_meta: max token length = 48
llm_load_tensors: offloading 42 repeating layers to GPU
llm_load_tensors: offloading output layer to GPU
llm_load_tensors: offloaded 43/43 layers to GPU
llm_load_tensors: CPU_Mapped model buffer size =  5700.31 MiB
.....................................................................................
llama_new_context_with_model: n_ctx      = 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_kv_cache_init:        CPU KV buffer size =  1344.00 MiB
llama_new_context_with_model: KV self size  = 1344.00 MiB, K (f16):  672.00 MiB, V (f16):  672.00 MiB
llama_new_context_with_model:        CPU  output buffer size =     0.98 MiB
llama_new_context_with_model:        CPU compute buffer size =   514.00 MiB
llama_new_context_with_model: graph nodes  = 1690
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 = 2

system_info: n_threads = 2 (n_threads_batch = 2) / 2 | AVX = 1 | AVX_VNNI = 0 | AVX2 = 1 | AVX512 = 0 | AVX512_VBMI = 0 | AVX512_VNNI = 0 | AVX512_BF16 = 0 | AMX_INT8 = 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: 172490367
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 = -1, n_keep = 1

User: Hi
Bot:Hi
User: 
Tell me the difference between thinking in humans and in LLMs.

Bot:Here's a breakdown of the key differences between human thinking and how Large Language Models (LLMs) like me "think":

**Human Thinking:**

* **Biological & Subconscious:**  Rooted in complex neural networks in the brain, much of human thought is subconscious, emergent, and influenced by emotions, experiences, and bodily sensations.
* **Intuitive & Creative:** Humans excel at making leaps of logic,  
llama_perf_sampler_print:    sampling time =      36.64 ms /   116 runs   (    0.32 ms per token,  3166.11 tokens per second)
llama_perf_context_print:        load time =   30829.81 ms
llama_perf_context_print: prompt eval time =   14709.99 ms /    29 tokens (  507.24 ms per token,     1.97 tokens per second)
llama_perf_context_print:        eval time =   70962.65 ms /    86 runs   (  825.15 ms per token,     1.21 tokens per second)
llama_perf_context_print:       total time =   85905.09 ms /   115 tokens
Interrupted by user

[DarkTyger]

git clone https://github.com/ggerganov/llama.cpp
cd llama.cpp
cmake -B build
cmake --build build --config Release -j$(nproc)
wget https://huggingface.co/ggml-org/gemma-1.1-7b-it-Q4_K_M-GGUF/resolve/main/gemma-1.1-7b-it.Q4_K_M.gguf -O gemma-7b-q4.gguf
./build/bin/llama-cli -m gemma-7b-q4.gguf -c 1024 -p "Once upon a time"

Output:

main: llama threadpool init, n_threads = 6
main: chat template is available, enabling conversation mode (disable it with -no-cnv)
main: chat template 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


system_info: n_threads = 6 (n_threads_batch = 6) / 12 | CPU : SSE3 = 1 | SSSE3 = 1 | AVX = 1 | AVX2 = 1 | F16C = 1 | FMA = 1 | AVX512 = 1 | AVX512_VBMI = 1 | AVX512_VNNI = 1 | AVX512_BF16 = 1 | LLAMAFILE = 1 | OPENMP = 1 | AARCH64_REPACK = 1 | 

main: interactive mode on.
sampler seed: 3615269160
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 = 1024
	top_k = 40, top_p = 0.950, min_p = 0.050, xtc_probability = 0.000, xtc_threshold = 0.100, typical_p = 1.000, top_n_sigma = -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 = 1024, n_batch = 2048, n_predict = -1, n_keep = 1

== Running in interactive mode. ==
 - Press Ctrl+C to interject at any time.
 - Press Return to return control to the AI.
 - To return control without starting a new line, end your input with '/'.
 - If you want to submit another line, end your input with '\'.

>

Is the documentation outdated?

https://github.com/ggml-org/llama.cpp/blob/master/examples/main/README.md