root@ub2: llama-server --model Qwen3.6-27B-Uncensored-HauhauCS-Aggressive-IQ3_M.gguf --port 10000 --host 0.0.0.0 -c 10240 --mmproj mmproj-Qwen3.6-27B-Uncensored-HauhauCS-Aggressive-f16.gguf --jinja -ngl 99 --flash-attn on --no-mmap --cache-ram 8192 --cache-type-k turbo3 --cache-type-v turbo3
ggml_cuda_init: found 1 CUDA devices (Total VRAM: 15841 MiB):
Device 0: NVIDIA GeForce RTX 5080, compute capability 12.0, VMM: yes, VRAM: 15841 MiB
main: n_parallel is set to auto, using n_parallel = 4 and kv_unified = true
build_info: b9064-67559e58
system_info: n_threads = 12 (n_threads_batch = 12) / 12 | CUDA : ARCHS = 1200 | USE_GRAPHS = 1 | PEER_MAX_BATCH_SIZE = 128 | BLACKWELL_NATIVE_FP4 = 1 | CPU : SSE3 = 1 | SSSE3 = 1 | AVX = 1 | AVX2 = 1 | F16C = 1 | FMA = 1 | BMI2 = 1 | LLAMAFILE = 1 | OPENMP = 1 | REPACK = 1 |
Running without SSL
init: using 11 threads for HTTP server
start: binding port with default address family
main: loading model
srv load_model: loading model '/opt/llama-cpp-turboquant-guide/dl/Qwen3.6-27B-Uncensored-HauhauCS-Aggressive-IQ3_M.gguf'
common_init_result: fitting params to device memory, for bugs during this step try to reproduce them with -fit off, or provide --verbose logs if the bug only occurs with -fit on
common_params_fit_impl: getting device memory data for initial parameters:
common_memory_breakdown_print: | memory breakdown [MiB] | total free self model context compute unaccounted |
common_memory_breakdown_print: | - CUDA0 (RTX 5080) | 15841 = 14962 + (12695 = 11466 + 723 + 505) + 17592186032600 |
common_memory_breakdown_print: | - Host | 561 = 520 + 0 + 40 |
common_params_fit_impl: projected to use 12695 MiB of device memory vs. 14962 MiB of free device memory
common_params_fit_impl: will leave 2267 >= 1024 MiB of free device memory, no changes needed
common_fit_params: successfully fit params to free device memory
common_fit_params: fitting params to free memory took 0.46 seconds
llama_model_load_from_file_impl: using device CUDA0 (NVIDIA GeForce RTX 5080) (0000:00:11.0) - 15562 MiB free
llama_model_loader: loaded meta data with 43 key-value pairs and 851 tensors from /opt/llama-cpp-turboquant-guide/dl/Qwen3.6-27B-Uncensored-HauhauCS-Aggressive-IQ3_M.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 = qwen35
llama_model_loader: - kv 1: general.type str = model
llama_model_loader: - kv 2: general.sampling.top_k i32 = 20
llama_model_loader: - kv 3: general.sampling.top_p f32 = 0.950000
llama_model_loader: - kv 4: general.sampling.temp f32 = 1.000000
llama_model_loader: - kv 5: general.name str = Qwen3.6-27B-Uncensored-HauhauCS-Aggre...
llama_model_loader: - kv 6: general.finetune str = Aggressive
llama_model_loader: - kv 7: general.basename str = Qwen3.6
llama_model_loader: - kv 8: general.size_label str = 27B
llama_model_loader: - kv 9: qwen35.block_count u32 = 64
llama_model_loader: - kv 10: qwen35.context_length u32 = 262144
llama_model_loader: - kv 11: qwen35.embedding_length u32 = 5120
llama_model_loader: - kv 12: qwen35.feed_forward_length u32 = 17408
llama_model_loader: - kv 13: qwen35.attention.head_count u32 = 24
llama_model_loader: - kv 14: qwen35.attention.head_count_kv u32 = 4
llama_model_loader: - kv 15: qwen35.rope.dimension_sections arr[i32,4] = [11, 11, 10, 0]
llama_model_loader: - kv 16: qwen35.rope.freq_base f32 = 10000000.000000
llama_model_loader: - kv 17: qwen35.attention.layer_norm_rms_epsilon f32 = 0.000001
llama_model_loader: - kv 18: qwen35.attention.key_length u32 = 256
llama_model_loader: - kv 19: qwen35.attention.value_length u32 = 256
llama_model_loader: - kv 20: qwen35.ssm.conv_kernel u32 = 4
llama_model_loader: - kv 21: qwen35.ssm.state_size u32 = 128
llama_model_loader: - kv 22: qwen35.ssm.group_count u32 = 16
llama_model_loader: - kv 23: qwen35.ssm.time_step_rank u32 = 48
llama_model_loader: - kv 24: qwen35.ssm.inner_size u32 = 6144
llama_model_loader: - kv 25: qwen35.full_attention_interval u32 = 4
llama_model_loader: - kv 26: qwen35.rope.dimension_count u32 = 64
llama_model_loader: - kv 27: tokenizer.ggml.model str = gpt2
llama_model_loader: - kv 28: tokenizer.ggml.pre str = qwen35
llama_model_loader: - kv 29: tokenizer.ggml.tokens arr[str,248320] = ["!", "\"", "#", "$", "%", "&", "'", ...
llama_model_loader: - kv 30: tokenizer.ggml.token_type arr[i32,248320] = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
llama_model_loader: - kv 31: tokenizer.ggml.merges arr[str,247587] = ["Ġ Ġ", "ĠĠ ĠĠ", "i n", "Ġ t",...
llama_model_loader: - kv 32: tokenizer.ggml.eos_token_id u32 = 248046
llama_model_loader: - kv 33: tokenizer.ggml.padding_token_id u32 = 248044
llama_model_loader: - kv 34: tokenizer.ggml.bos_token_id u32 = 248044
llama_model_loader: - kv 35: tokenizer.chat_template str = {%- set image_count = namespace(value...
llama_model_loader: - kv 36: general.quantization_version u32 = 2
llama_model_loader: - kv 37: general.file_type u32 = 27
llama_model_loader: - kv 38: general.author str = HauhauCS
llama_model_loader: - kv 39: quantize.imatrix.file str = Qwen3.6-27B-Uncensored-HauhauCS-Aggre...
llama_model_loader: - kv 40: quantize.imatrix.dataset str = groups_merged.txt
llama_model_loader: - kv 41: quantize.imatrix.entries_count u32 = 496
llama_model_loader: - kv 42: quantize.imatrix.chunks_count u32 = 93
llama_model_loader: - type f32: 353 tensors
llama_model_loader: - type q4_K: 88 tensors
llama_model_loader: - type q6_K: 1 tensors
llama_model_loader: - type iq3_s: 409 tensors
print_info: file format = GGUF V3 (latest)
print_info: file type = IQ3_S mix - 3.66 bpw
print_info: file size = 11.71 GiB (3.74 BPW)
load: 0 unused tokens
load: printing all EOG tokens:
load: - 248044 ('<|endoftext|>')
load: - 248046 ('<|im_end|>')
load: - 248063 ('<|fim_pad|>')
load: - 248064 ('<|repo_name|>')
load: - 248065 ('<|file_sep|>')
load: special tokens cache size = 33
load: token to piece cache size = 1.7581 MB
print_info: arch = qwen35
print_info: vocab_only = 0
print_info: no_alloc = 0
print_info: n_ctx_train = 262144
print_info: n_embd = 5120
print_info: n_embd_inp = 5120
print_info: n_layer = 64
print_info: n_head = 24
print_info: n_head_kv = 4
print_info: n_rot = 64
print_info: n_swa = 0
print_info: is_swa_any = 0
print_info: n_embd_head_k = 256
print_info: n_embd_head_v = 256
print_info: n_gqa = 6
print_info: n_embd_k_gqa = 1024
print_info: n_embd_v_gqa = 1024
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: f_attn_scale = 0.0e+00
print_info: n_ff = 17408
print_info: n_expert = 0
print_info: n_expert_used = 0
print_info: n_expert_groups = 0
print_info: n_group_used = 0
print_info: causal attn = 1
print_info: pooling type = -1
print_info: rope type = 40
print_info: rope scaling = linear
print_info: freq_base_train = 10000000.0
print_info: freq_scale_train = 1
print_info: n_ctx_orig_yarn = 262144
print_info: rope_yarn_log_mul = 0.0000
print_info: rope_finetuned = unknown
print_info: mrope sections = [11, 11, 10, 0]
print_info: ssm_d_conv = 4
print_info: ssm_d_inner = 6144
print_info: ssm_d_state = 128
print_info: ssm_dt_rank = 48
print_info: ssm_n_group = 16
print_info: ssm_dt_b_c_rms = 0
print_info: model type = 27B
print_info: model params = 26.90 B
print_info: general.name = Qwen3.6-27B-Uncensored-HauhauCS-Aggressive
print_info: vocab type = BPE
print_info: n_vocab = 248320
print_info: n_merges = 247587
print_info: BOS token = 248044 '<|endoftext|>'
print_info: EOS token = 248046 '<|im_end|>'
print_info: EOT token = 248046 '<|im_end|>'
print_info: PAD token = 248044 '<|endoftext|>'
print_info: LF token = 198 'Ċ'
print_info: FIM PRE token = 248060 '<|fim_prefix|>'
print_info: FIM SUF token = 248062 '<|fim_suffix|>'
print_info: FIM MID token = 248061 '<|fim_middle|>'
print_info: FIM PAD token = 248063 '<|fim_pad|>'
print_info: FIM REP token = 248064 '<|repo_name|>'
print_info: FIM SEP token = 248065 '<|file_sep|>'
print_info: EOG token = 248044 '<|endoftext|>'
print_info: EOG token = 248046 '<|im_end|>'
print_info: EOG token = 248063 '<|fim_pad|>'
print_info: EOG token = 248064 '<|repo_name|>'
print_info: EOG token = 248065 '<|file_sep|>'
print_info: max token length = 256
load_tensors: loading model tensors, this can take a while... (mmap = false, direct_io = false)
load_tensors: offloading output layer to GPU
load_tensors: offloading 63 repeating layers to GPU
load_tensors: offloaded 65/65 layers to GPU
load_tensors: CPU model buffer size = 521.00 MiB
load_tensors: CUDA0 model buffer size = 11466.58 MiB
..........................................................................................
common_init_result: added <|endoftext|> logit bias = -inf
common_init_result: added <|im_end|> logit bias = -inf
common_init_result: added <|fim_pad|> logit bias = -inf
common_init_result: added <|repo_name|> logit bias = -inf
common_init_result: added <|file_sep|> logit bias = -inf
llama_context: constructing llama_context
llama_context: n_seq_max = 4
llama_context: n_ctx = 10240
llama_context: n_ctx_seq = 10240
llama_context: n_batch = 2048
llama_context: n_ubatch = 512
llama_context: causal_attn = 1
llama_context: flash_attn = enabled
llama_context: kv_unified = true
llama_context: freq_base = 10000000.0
llama_context: freq_scale = 1
llama_context: n_ctx_seq (10240) < n_ctx_train (262144) -- the full capacity of the model will not be utilized
llama_context: CUDA_Host output buffer size = 3.79 MiB
llama_kv_cache: CUDA0 KV buffer size = 125.13 MiB
llama_kv_cache: TurboQuant rotation matrices initialized (128x128)
llama_kv_cache: size = 125.00 MiB ( 10240 cells, 16 layers, 4/1 seqs), K (turbo3): 62.50 MiB, V (turbo3): 62.50 MiB
llama_kv_cache: upstream attention rotation disabled (TurboQuant uses kernel-level WHT)
llama_kv_cache: attn_rot_k = 0, n_embd_head_k_all = 256
llama_kv_cache: attn_rot_v = 0, n_embd_head_k_all = 256
llama_memory_recurrent: CUDA0 RS buffer size = 598.50 MiB
llama_memory_recurrent: size = 598.50 MiB ( 4 cells, 64 layers, 4 seqs), R (f32): 22.50 MiB, S (f32): 576.00 MiB
sched_reserve: reserving ...
sched_reserve: resolving fused Gated Delta Net support:
sched_reserve: fused Gated Delta Net (autoregressive) enabled
sched_reserve: fused Gated Delta Net (chunked) enabled
sched_reserve: CUDA0 compute buffer size = 505.00 MiB
sched_reserve: CUDA_Host compute buffer size = 40.02 MiB
sched_reserve: graph nodes = 3689
sched_reserve: graph splits = 2
sched_reserve: reserve took 20.55 ms, sched copies = 1
common_init_from_params: warming up the model with an empty run - please wait ... (--no-warmup to disable)
clip_model_loader: model name: Qwen3.6-27B-Uncensored-HauhauCS-Aggressive
clip_model_loader: description:
clip_model_loader: GGUF version: 3
clip_model_loader: alignment: 32
clip_model_loader: n_tensors: 334
clip_model_loader: n_kv: 26
clip_model_loader: has vision encoder
clip_ctx: CLIP using CUDA0 backend
load_hparams: Qwen-VL models require at minimum 1024 image tokens to function correctly on grounding tasks
load_hparams: if you encounter problems with accuracy, try adding --image-min-tokens 1024
load_hparams: more info: https://github.com/ggml-org/llama.cpp/issues/16842
load_hparams: projector: qwen3vl_merger
load_hparams: n_embd: 1152
load_hparams: n_head: 16
load_hparams: n_ff: 4304
load_hparams: n_layer: 27
load_hparams: ffn_op: gelu
load_hparams: projection_dim: 5120
--- vision hparams ---
load_hparams: image_size: 768
load_hparams: patch_size: 16
load_hparams: has_llava_proj: 0
load_hparams: minicpmv_version: 0
load_hparams: n_merge: 2
load_hparams: n_wa_pattern: 0
load_hparams: image_min_pixels: 8192
load_hparams: image_max_pixels: 4194304
load_hparams: model size: 884.62 MiB
load_hparams: metadata size: 0.12 MiB
warmup: warmup with image size = 1472 x 1472
alloc_compute_meta: CUDA0 compute buffer size = 248.10 MiB
alloc_compute_meta: CPU compute buffer size = 24.93 MiB
alloc_compute_meta: graph splits = 1, nodes = 823
warmup: flash attention is enabled
srv load_model: loaded multimodal model, '/opt/llama-cpp-turboquant-guide/dl/mmproj-Qwen3.6-27B-Uncensored-HauhauCS-Aggressive-f16.gguf'
srv load_model: initializing slots, n_slots = 4
common_context_can_seq_rm: the target context does not support partial sequence removal
srv load_model: speculative decoding will use checkpoints
no implementations specified for speculative decoding
slot load_model: id 0 | task -1 | new slot, n_ctx = 10240
no implementations specified for speculative decoding
slot load_model: id 1 | task -1 | new slot, n_ctx = 10240
no implementations specified for speculative decoding
slot load_model: id 2 | task -1 | new slot, n_ctx = 10240
no implementations specified for speculative decoding
slot load_model: id 3 | task -1 | new slot, n_ctx = 10240
srv load_model: prompt cache is enabled, size limit: 8192 MiB
srv load_model: use `--cache-ram 0` to disable the prompt cache
srv load_model: for more info see https://github.com/ggml-org/llama.cpp/pull/16391
srv init: init: idle slots will be saved to prompt cache and cleared upon starting a new task
init: chat template, example_format: '<|im_start|>system
You are a helpful assistant<|im_end|>
<|im_start|>user
Hello<|im_end|>
<|im_start|>assistant
Hi there<|im_end|>
<|im_start|>user
How are you?<|im_end|>
<|im_start|>assistant
<think>
'
srv init: init: chat template, thinking = 1
main: model loaded
main: server is listening on http://0.0.0.0:10000
main: starting the main loop...
srv update_slots: all slots are idle
srv log_server_r: done request: GET / 192.168.120.3 200
srv log_server_r: done request: GET /bundle.css 192.168.120.3 200
srv log_server_r: done request: GET /bundle.js 192.168.120.3 200
srv log_server_r: done request: HEAD /cors-proxy 192.168.120.3 404
srv params_from_: Chat format: peg-native
slot get_availabl: id 3 | task -1 | selected slot by LRU, t_last = -1
srv get_availabl: updating prompt cache
srv load: - looking for better prompt, base f_keep = -1.000, sim = 0.000
srv update: - cache state: 0 prompts, 0.000 MiB (limits: 8192.000 MiB, 10240 tokens, 8589934592 est)
srv get_availabl: prompt cache update took 0.00 ms
slot launch_slot_: id 3 | task -1 | sampler chain: logits -> ?penalties -> ?dry -> ?top-n-sigma -> top-k -> ?typical -> top-p -> min-p -> ?xtc -> ?temp-ext -> dist
slot launch_slot_: id 3 | task 0 | processing task, is_child = 0
slot update_slots: id 3 | task 0 | new prompt, n_ctx_slot = 10240, n_keep = 0, task.n_tokens = 11
slot update_slots: id 3 | task 0 | n_tokens = 0, memory_seq_rm [0, end)
slot update_slots: id 3 | task 0 | prompt processing progress, n_tokens = 7, batch.n_tokens = 7, progress = 0.636364
srv log_server_r: done request: POST /v1/chat/completions 192.168.120.3 200
slot update_slots: id 3 | task 0 | n_tokens = 7, memory_seq_rm [7, end)
slot init_sampler: id 3 | task 0 | init sampler, took 0.01 ms, tokens: text = 11, total = 11
slot update_slots: id 3 | task 0 | prompt processing done, n_tokens = 11, batch.n_tokens = 4
srv stop: cancel task, id_task = 0
slot release: id 3 | task 0 | stop processing: n_tokens = 3560, truncated = 0
srv update_slots: all slots are idle
srv params_from_: Chat format: peg-native
slot get_availabl: id 3 | task -1 | selected slot by LCP similarity, sim_best = 1.000 (> 0.100 thold), f_keep = 0.003
srv get_availabl: updating prompt cache
srv prompt_save: - saving prompt with length 3560, total state size = 193.151 MiB
srv load: - looking for better prompt, base f_keep = 0.003, sim = 1.000
srv update: - cache state: 1 prompts, 193.151 MiB (limits: 8192.000 MiB, 10240 tokens, 150987 est)
srv update: - prompt 0x5e919ffcef30: 3560 tokens, checkpoints: 0, 193.151 MiB
srv get_availabl: prompt cache update took 84.99 ms
slot launch_slot_: id 3 | task -1 | sampler chain: logits -> ?penalties -> ?dry -> ?top-n-sigma -> top-k -> ?typical -> top-p -> min-p -> ?xtc -> ?temp-ext -> dist
slot launch_slot_: id 3 | task 3553 | processing task, is_child = 0
slot update_slots: id 3 | task 3553 | new prompt, n_ctx_slot = 10240, n_keep = 0, task.n_tokens = 11
slot update_slots: id 3 | task 3553 | n_past = 11, slot.prompt.tokens.size() = 3560, seq_id = 3, pos_min = 3559, n_swa = 0
slot update_slots: id 3 | task 3553 | forcing full prompt re-processing due to lack of cache data (likely due to SWA or hybrid/recurrent memory, see https://github.com/ggml-org/llama.cpp/pull/13194#issuecomment-2868343055)
slot update_slots: id 3 | task 3553 | n_tokens = 0, memory_seq_rm [0, end)
srv log_server_r: done request: POST /v1/chat/completions 192.168.120.3 200
slot update_slots: id 3 | task 3553 | prompt processing progress, n_tokens = 7, batch.n_tokens = 7, progress = 0.636364
slot update_slots: id 3 | task 3553 | n_tokens = 7, memory_seq_rm [7, end)
slot init_sampler: id 3 | task 3553 | init sampler, took 0.00 ms, tokens: text = 11, total = 11
slot update_slots: id 3 | task 3553 | prompt processing done, n_tokens = 11, batch.n_tokens = 4
srv stop: cancel task, id_task = 3553
slot release: id 3 | task 3553 | stop processing: n_tokens = 79, truncated = 0
srv update_slots: all slots are idle
srv params_from_: Chat format: peg-native
slot get_availabl: id 3 | task -1 | selected slot by LCP similarity, sim_best = 1.000 (> 0.100 thold), f_keep = 0.139
srv get_availabl: updating prompt cache
srv prompt_save: - saving prompt with length 79, total state size = 150.592 MiB
srv load: - looking for better prompt, base f_keep = 0.139, sim = 1.000
srv update: - cache state: 2 prompts, 343.744 MiB (limits: 8192.000 MiB, 10240 tokens, 86723 est)
srv update: - prompt 0x5e919ffcef30: 3560 tokens, checkpoints: 0, 193.151 MiB
srv update: - prompt 0x5e919fbdcc90: 79 tokens, checkpoints: 0, 150.592 MiB
srv get_availabl: prompt cache update took 56.77 ms
slot launch_slot_: id 3 | task -1 | sampler chain: logits -> ?penalties -> ?dry -> ?top-n-sigma -> top-k -> ?typical -> top-p -> min-p -> ?xtc -> ?temp-ext -> dist
slot launch_slot_: id 3 | task 3625 | processing task, is_child = 0
slot update_slots: id 3 | task 3625 | new prompt, n_ctx_slot = 10240, n_keep = 0, task.n_tokens = 11
slot update_slots: id 3 | task 3625 | n_past = 11, slot.prompt.tokens.size() = 79, seq_id = 3, pos_min = 78, n_swa = 0
slot update_slots: id 3 | task 3625 | forcing full prompt re-processing due to lack of cache data (likely due to SWA or hybrid/recurrent memory, see https://github.com/ggml-org/llama.cpp/pull/13194#issuecomment-2868343055)
slot update_slots: id 3 | task 3625 | n_tokens = 0, memory_seq_rm [0, end)
slot update_slots: id 3 | task 3625 | prompt processing progress, n_tokens = 7, batch.n_tokens = 7, progress = 0.636364
srv log_server_r: done request: POST /v1/chat/completions 192.168.120.3 200
slot update_slots: id 3 | task 3625 | n_tokens = 7, memory_seq_rm [7, end)
slot init_sampler: id 3 | task 3625 | init sampler, took 0.00 ms, tokens: text = 11, total = 11
slot update_slots: id 3 | task 3625 | prompt processing done, n_tokens = 11, batch.n_tokens = 4
srv stop: cancel task, id_task = 3625
slot release: id 3 | task 3625 | stop processing: n_tokens = 259, truncated = 0
srv update_slots: all slots are idle
^Csrv operator(): operator(): cleaning up before exit...
common_memory_breakdown_print: | memory breakdown [MiB] | total free self model context compute unaccounted |
common_memory_breakdown_print: | - CUDA0 (RTX 5080) | 15841 = 1682 + (12695 = 11466 + 723 + 505) + 1464 |
common_memory_breakdown_print: | - Host | 561 = 520 + 0 + 40 |
Name and Version
Operating systems
Linux
GGML backends
CUDA
Hardware
RTX 5080 / 16G + CUDA 13.2
Models
Qwen3.6-27B-Uncensored-HauhauCS-Aggressive-IQ3_M
Problem description & steps to reproduce
I tried two IQ model all failed with a rave response. (Qwen3.6-27B-Uncensored-HauhauCS-Aggressive-IQ3_M.gguf , Qwen3.6-35B-A3B-UD-IQ4_NL.gguf)
While the Q4 model works perfectly.
I tried build locally or at github action (https://github.com/starskyzheng/llamaup/releases/tag/feature-turboquant-kv-cache__thetom-llama-cpp-turboquant), but all fail to response.
First Bad Commit
No response
Relevant log output
Webui:
Server Log: