Eval bug: GGGGGGGG... or garbled works

[gbdb71]

Name and Version

load_backend: loaded RPC backend from /mnt/10a35c9a-e885-401a-b71a-38f856f9bf0a/ai/llama.cpp_ggerganov/build/bin/libggml-rpc.so
ggml_vulkan: Found 1 Vulkan devices:
ggml_vulkan: 0 = NVIDIA GeForce GTX 1050 Ti (NVIDIA) | uma: 0 | fp16: 0 | bf16: 0 | warp size: 32 | shared memory: 49152 | int dot: 1 | matrix cores: none
load_backend: loaded Vulkan backend from /mnt/10a35c9a-e885-401a-b71a-38f856f9bf0a/ai/llama.cpp_ggerganov/build/bin/libggml-vulkan.so
load_backend: loaded CPU backend from /mnt/10a35c9a-e885-401a-b71a-38f856f9bf0a/ai/llama.cpp_ggerganov/build/bin/libggml-cpu-haswell.so
version: 6106 (5fd160b)
built with cc (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0 for x86_64-linux-gnu

Operating systems

Linux

GGML backends

Vulkan

Hardware

System Details Report

---

Report details

• Date generated: 2025-08-12 18:47:32

Hardware Information:

• Hardware Model: ASUSTeK COMPUTER INC. TUF B450M-PRO GAMING
• Memory: 64.0 GiB
• Processor: AMD Ryzen™ 7 2700 × 16
• Graphics: NVIDIA GeForce GTX 1050 Ti
• Disk Capacity: 6.5 TB

Software Information:

• Firmware Version: 2006
• OS Name: Ubuntu 25.04
• OS Build: (null)
• OS Type: 64-bit
• GNOME Version: 48
• Windowing System: Wayland
• Kernel Version: Linux 6.14.0-27-generic

Models

Unsloth/gpt-oss-20b: gpt-oss-20b-F16

Problem description & steps to reproduce

Answer: GGGGGGGG... or garbled works. it occours randomly, but rebooting doesn't solve it. Happens with Ollama too. Happens with every model tested.

First Bad Commit

No response

Relevant log output

#CTRL+C 
srv    operator(): operator(): cleaning up before exit...
[New LWP 68015]
[New LWP 68014]
[New LWP 68013]
[New LWP 68012]
[New LWP 68011]
[New LWP 68010]
[New LWP 68009]
[New LWP 67996]
[New LWP 67993]
[New LWP 67991]
[New LWP 67990]
[New LWP 67989]
[New LWP 67973]
[New LWP 67972]
[New LWP 67971]

This GDB supports auto-downloading debuginfo from the following URLs:
  <https://debuginfod.ubuntu.com>
Enable debuginfod for this session? (y or [n]) [answered N; input not from terminal]
Debuginfod has been disabled.
To make this setting permanent, add 'set debuginfod enabled off' to .gdbinit.
Function(s) ^std::(move|forward|as_const|(__)?addressof) will be skipped when stepping.
Function(s) ^std::(shared|unique)_ptr<.*>::(get|operator) will be skipped when stepping.
Function(s) ^std::(basic_string|vector|array|deque|(forward_)?list|(unordered_|flat_)?(multi)?(map|set)|span)<.*>::(c?r?(begin|end)|front|back|data|size|empty) will be skipped when stepping.
Function(s) ^std::(basic_string|vector|array|deque|span)<.*>::operator.] will be skipped when stepping.
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
__syscall_cancel_arch () at ../sysdeps/unix/sysv/linux/x86_64/syscall_cancel.S:56
warning: 56	../sysdeps/unix/sysv/linux/x86_64/syscall_cancel.S: File o directory non esistente
#0  __syscall_cancel_arch () at ../sysdeps/unix/sysv/linux/x86_64/syscall_cancel.S:56
56	in ../sysdeps/unix/sysv/linux/x86_64/syscall_cancel.S
#1  0x00007f82b669eae3 in __internal_syscall_cancel (a1=a1@entry=140199065545104, a2=<optimized out>, a3=<optimized out>, a4=a4@entry=0, a5=a5@entry=0, a6=a6@entry=4294967295, nr=202) at ./nptl/cancellation.c:49
warning: 49	./nptl/cancellation.c: File o directory non esistente
#2  0x00007f82b669f237 in __futex_abstimed_wait_common64 (private=128, futex_word=0x7f82a37fe990, expected=<optimized out>, op=265, abstime=0x0, cancel=true) at ./nptl/futex-internal.c:57
warning: 57	./nptl/futex-internal.c: File o directory non esistente
#3  __futex_abstimed_wait_common (futex_word=0x7f82a37fe990, expected=<optimized out>, clockid=0, abstime=0x0, private=128, cancel=true) at ./nptl/futex-internal.c:87
87	in ./nptl/futex-internal.c
#4  __GI___futex_abstimed_wait_cancelable64 (futex_word=futex_word@entry=0x7f82a37fe990, expected=<optimized out>, clockid=clockid@entry=0, abstime=abstime@entry=0x0, private=private@entry=128) at ./nptl/futex-internal.c:139
139	in ./nptl/futex-internal.c
#5  0x00007f82b66a4614 in __pthread_clockjoin_ex (threadid=140199065544384, thread_return=0x0, clockid=0, abstime=0x0, block=<optimized out>) at ./nptl/pthread_join_common.c:108
warning: 108	./nptl/pthread_join_common.c: File o directory non esistente
#6  0x00007f82b6af2393 in std::thread::join() () from /lib/x86_64-linux-gnu/libstdc++.so.6
#7  0x000061d15aac05d9 in main ()
[Inferior 1 (process 67970) detached]
terminate called without an active exception
Annullato (core dump creato)

./llama-server -c 16384 --host 127.0.0.1 --port 8080 --jinja -m ../../../llm/gpt/gpt-oss-20b-F16/gpt-oss-20b-F16.gguf
load_backend: loaded RPC backend from /mnt/10a35c9a-e885-401a-b71a-38f856f9bf0a/ai/llama.cpp_ggerganov/build/bin/libggml-rpc.so
ggml_vulkan: Found 1 Vulkan devices:
ggml_vulkan: 0 = NVIDIA GeForce GTX 1050 Ti (NVIDIA) | uma: 0 | fp16: 0 | bf16: 0 | warp size: 32 | shared memory: 49152 | int dot: 1 | matrix cores: none
load_backend: loaded Vulkan backend from /mnt/10a35c9a-e885-401a-b71a-38f856f9bf0a/ai/llama.cpp_ggerganov/build/bin/libggml-vulkan.so
load_backend: loaded CPU backend from /mnt/10a35c9a-e885-401a-b71a-38f856f9bf0a/ai/llama.cpp_ggerganov/build/bin/libggml-cpu-haswell.so
build: 6106 (5fd160bb) with cc (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0 for x86_64-linux-gnu
system info: n_threads = 8, n_threads_batch = 8, total_threads = 16

system_info: n_threads = 8 (n_threads_batch = 8) / 16 | CPU : SSE3 = 1 | SSSE3 = 1 | AVX = 1 | AVX2 = 1 | F16C = 1 | FMA = 1 | BMI2 = 1 | LLAMAFILE = 1 | OPENMP = 1 | REPACK = 1 | 

main: binding port with default address family
main: HTTP server is listening, hostname: 127.0.0.1, port: 8080, http threads: 15
main: loading model
srv    load_model: loading model '../../../llm/gpt/gpt-oss-20b-F16/gpt-oss-20b-F16.gguf'
llama_model_load_from_file_impl: using device Vulkan0 (NVIDIA GeForce GTX 1050 Ti) - 4096 MiB free
llama_model_loader: loaded meta data with 37 key-value pairs and 459 tensors from ../../../llm/gpt/gpt-oss-20b-F16/gpt-oss-20b-F16.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              = gpt-oss
llama_model_loader: - kv   1:                               general.type str              = model
llama_model_loader: - kv   2:                               general.name str              = Gpt-Oss-20B
llama_model_loader: - kv   3:                           general.basename str              = Gpt-Oss-20B
llama_model_loader: - kv   4:                       general.quantized_by str              = Unsloth
llama_model_loader: - kv   5:                         general.size_label str              = 20B
llama_model_loader: - kv   6:                            general.license str              = apache-2.0
llama_model_loader: - kv   7:                           general.repo_url str              = https://huggingface.co/unsloth
llama_model_loader: - kv   8:                               general.tags arr[str,2]       = ["vllm", "text-generation"]
llama_model_loader: - kv   9:                        gpt-oss.block_count u32              = 24
llama_model_loader: - kv  10:                     gpt-oss.context_length u32              = 131072
llama_model_loader: - kv  11:                   gpt-oss.embedding_length u32              = 2880
llama_model_loader: - kv  12:                gpt-oss.feed_forward_length u32              = 2880
llama_model_loader: - kv  13:               gpt-oss.attention.head_count u32              = 64
llama_model_loader: - kv  14:            gpt-oss.attention.head_count_kv u32              = 8
llama_model_loader: - kv  15:                     gpt-oss.rope.freq_base f32              = 150000.000000
llama_model_loader: - kv  16:   gpt-oss.attention.layer_norm_rms_epsilon f32              = 0.000010
llama_model_loader: - kv  17:                       gpt-oss.expert_count u32              = 32
llama_model_loader: - kv  18:                  gpt-oss.expert_used_count u32              = 4
llama_model_loader: - kv  19:               gpt-oss.attention.key_length u32              = 64
llama_model_loader: - kv  20:             gpt-oss.attention.value_length u32              = 64
llama_model_loader: - kv  21:                          general.file_type u32              = 1
llama_model_loader: - kv  22:           gpt-oss.attention.sliding_window u32              = 128
llama_model_loader: - kv  23:         gpt-oss.expert_feed_forward_length u32              = 2880
llama_model_loader: - kv  24:                  gpt-oss.rope.scaling.type str              = yarn
llama_model_loader: - kv  25:                gpt-oss.rope.scaling.factor f32              = 32.000000
llama_model_loader: - kv  26: gpt-oss.rope.scaling.original_context_length u32              = 4096
llama_model_loader: - kv  27:               general.quantization_version u32              = 2
llama_model_loader: - kv  28:                       tokenizer.ggml.model str              = gpt2
llama_model_loader: - kv  29:                         tokenizer.ggml.pre str              = gpt-4o
llama_model_loader: - kv  30:                      tokenizer.ggml.tokens arr[str,201088]  = ["!", "\"", "#", "$", "%", "&", "'", ...
llama_model_loader: - kv  31:                  tokenizer.ggml.token_type arr[i32,201088]  = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
llama_model_loader: - kv  32:                      tokenizer.ggml.merges arr[str,446189]  = ["Ġ Ġ", "Ġ ĠĠĠ", "ĠĠ ĠĠ", "...
llama_model_loader: - kv  33:                tokenizer.ggml.bos_token_id u32              = 199998
llama_model_loader: - kv  34:                tokenizer.ggml.eos_token_id u32              = 200002
llama_model_loader: - kv  35:            tokenizer.ggml.padding_token_id u32              = 200017
llama_model_loader: - kv  36:                    tokenizer.chat_template str              = {# Copyright 2025-present Unsloth. Ap...
llama_model_loader: - type  f32:  289 tensors
llama_model_loader: - type  f16:   98 tensors
llama_model_loader: - type mxfp4:   72 tensors
print_info: file format = GGUF V3 (latest)
print_info: file type   = F16
print_info: file size   = 12.83 GiB (5.27 BPW) 
load: printing all EOG tokens:
load:   - 199999 ('<|endoftext|>')
load:   - 200002 ('<|return|>')
load:   - 200007 ('<|end|>')
load:   - 200012 ('<|call|>')
load: special_eog_ids contains both '<|return|>' and '<|call|>' tokens, removing '<|end|>' token from EOG list
load: special tokens cache size = 21
load: token to piece cache size = 1.3332 MB
print_info: arch             = gpt-oss
print_info: vocab_only       = 0
print_info: n_ctx_train      = 131072
print_info: n_embd           = 2880
print_info: n_layer          = 24
print_info: n_head           = 64
print_info: n_head_kv        = 8
print_info: n_rot            = 64
print_info: n_swa            = 128
print_info: is_swa_any       = 1
print_info: n_embd_head_k    = 64
print_info: n_embd_head_v    = 64
print_info: n_gqa            = 8
print_info: n_embd_k_gqa     = 512
print_info: n_embd_v_gqa     = 512
print_info: f_norm_eps       = 0.0e+00
print_info: f_norm_rms_eps   = 1.0e-05
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: f_attn_scale     = 0.0e+00
print_info: n_ff             = 2880
print_info: n_expert         = 32
print_info: n_expert_used    = 4
print_info: causal attn      = 1
print_info: pooling type     = 0
print_info: rope type        = 2
print_info: rope scaling     = yarn
print_info: freq_base_train  = 150000.0
print_info: freq_scale_train = 0.03125
print_info: n_ctx_orig_yarn  = 4096
print_info: rope_finetuned   = unknown
print_info: model type       = ?B
print_info: model params     = 20.91 B
print_info: general.name     = Gpt-Oss-20B
print_info: n_ff_exp         = 2880
print_info: vocab type       = BPE
print_info: n_vocab          = 201088
print_info: n_merges         = 446189
print_info: BOS token        = 199998 '<|startoftext|>'
print_info: EOS token        = 200002 '<|return|>'
print_info: EOT token        = 199999 '<|endoftext|>'
print_info: PAD token        = 200017 '<|reserved_200017|>'
print_info: LF token         = 198 'Ċ'
print_info: EOG token        = 199999 '<|endoftext|>'
print_info: EOG token        = 200002 '<|return|>'
print_info: EOG token        = 200012 '<|call|>'
print_info: max token length = 256
load_tensors: loading model tensors, this can take a while... (mmap = true)
load_tensors: offloading 0 repeating layers to GPU
load_tensors: offloaded 0/25 layers to GPU
load_tensors:   CPU_Mapped model buffer size = 13141.28 MiB
....................................................................................
llama_context: constructing llama_context
llama_context: n_seq_max     = 1
llama_context: n_ctx         = 16384
llama_context: n_ctx_per_seq = 16384
llama_context: n_batch       = 2048
llama_context: n_ubatch      = 512
llama_context: causal_attn   = 1
llama_context: flash_attn    = 0
llama_context: kv_unified    = false
llama_context: freq_base     = 150000.0
llama_context: freq_scale    = 0.03125
llama_context: n_ctx_per_seq (16384) < n_ctx_train (131072) -- the full capacity of the model will not be utilized
llama_context:        CPU  output buffer size =     0.77 MiB
llama_kv_cache_unified_iswa: creating non-SWA KV cache, size = 16384 cells
llama_kv_cache_unified:        CPU KV buffer size =   384.00 MiB
llama_kv_cache_unified: size =  384.00 MiB ( 16384 cells,  12 layers,  1/1 seqs), K (f16):  192.00 MiB, V (f16):  192.00 MiB
llama_kv_cache_unified_iswa: creating     SWA KV cache, size = 640 cells
llama_kv_cache_unified:        CPU KV buffer size =    15.00 MiB
llama_kv_cache_unified: size =   15.00 MiB (   640 cells,  12 layers,  1/1 seqs), K (f16):    7.50 MiB, V (f16):    7.50 MiB
llama_context:    Vulkan0 compute buffer size =  2120.89 MiB
llama_context: Vulkan_Host compute buffer size =    82.26 MiB
llama_context: graph nodes  = 1446
llama_context: graph splits = 556 (with bs=512), 1 (with bs=1)
common_init_from_params: KV cache shifting is not supported for this context, disabling KV cache shifting
common_init_from_params: added <|endoftext|> logit bias = -inf
common_init_from_params: added <|return|> logit bias = -inf
common_init_from_params: added <|call|> logit bias = -inf
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)
srv          init: initializing slots, n_slots = 1
slot         init: id  0 | task -1 | new slot n_ctx_slot = 16384
main: model loaded

[BugBusterMax]

I had the same problem as you before. Maybe you are using a quantitative model. You could try to de-quantify the model first and then perform the reasoning.

[gbdb71]

I'm getting [edit: I was hoping to get] a taste of serenity by using flag --swa-full (weights normalization?) and by dropping caches with command: sudo sync && echo 3 | sudo tee /proc/sys/vm/drop_caches. But that doesn't work.
.

[jeffbolznv]

@gbdb71 Please try #15334