LLAMA-IMATRIX(1)

📦
version 10438 · 2 versions
, in ubuntu stonking

NAME

llama-imatrix — it's giving llama-imatrix, periodt. That's the name and the command, bestie. No notes. 💅

DESCRIPTION

okay so bestie, this page is serving up the common params shared across llama.cpp tools. It's a long list, no cap, but we're gonna go through every single one together. Grab a boba 🧋 and let's get into it. ✨

✨ ----- common params ----- ✨

  • -h, --help, --usage — prints the usage info and dips out, slay. 💅
  • --version — shows you the version and build info, cute. ✨
  • -cl, --cache-list — lists all the models chillin' in your cache, lowkey useful. 🗂️
  • --completion-bash — prints a source-able bash completion script for llama.cpp, so your tab key can go feral. 🔥
  • -t, --threads N — how many CPU threads to use during generation, bestie. default is <num_cpus>, and the env var is LLAMA_ARG_THREADS. 🧵
  • -tb, --threads-batch N — threads for batch and prompt processing. default is same as --threads, iykyk. 📦
  • -C, --cpu-mask M — CPU affinity mask: arbitrarily long hex. it complements cpu-range, and default is "". 🎭
  • -Cr, --cpu-range lo-hi — range of CPUs for affinity, complements --cpu-mask. 🎯
  • --cpu-strict <0|1> — use strict CPU placement, default 0. 💅
  • --prio N — set process/thread priority: low(-1), normal(0), medium(1), high(2), realtime(3), default 0. ⚡
  • --poll <0...100> — polling level to wait for work (0 = no polling), default 50. ⏳
  • -Cb, --cpu-mask-batch M — CPU affinity mask for batch: arbitrarily long hex. complements cpu-range-batch, default is same as --cpu-mask. 🎭
  • -Crb, --cpu-range-batch lo-hi — ranges of CPUs for affinity, complements --cpu-mask-batch. 🎯
  • --cpu-strict-batch <0|1> — strict CPU placement for batch, default same as --cpu-strict. 💅
  • --prio-batch N — priority for batch: 0-normal, 1-medium, 2-high, 3-realtime, default 0. ⚡
  • --poll-batch <0|1> — use polling to wait for work, default same as --poll. ⏳
  • -c, --ctx-size N — size of the prompt context. default 512, 0 means loaded from model, env is LLAMA_ARG_CTX_SIZE. 📏
  • -n, --predict, --n-predict N — number of tokens to predict. default -1, and -1 means infinity, env is LLAMA_ARG_N_PREDICT. 🔮
  • -b, --batch-size N — logical maximum batch size, default 2048, env LLAMA_ARG_BATCH. 📦
  • -ub, --ubatch-size N — physical maximum batch size, default 512, env LLAMA_ARG_UBATCH. 📦
  • --keep N — number of tokens to keep from the initial prompt. default 0, -1 means all. 💾
  • --swa-full — use full-size SWA cache, default false. more info: https://github.com/ggml-org/llama.cpp/pull/13194#issuecomment-2868343055, env LLAMA_ARG_SWA_FULL. 💾
  • -fa, --flash-attn [on|off|auto] — set Flash Attention use ('on', 'off', or 'auto', default 'auto'), env LLAMA_ARG_FLASH_ATTN. ⚡
  • -p, --prompt PROMPT — prompt to start generation with; for system message, use -sys. ✍️
  • --perf, --no-perf — whether to enable internal libllama performance timings, default false, env LLAMA_ARG_PERF. ⏱️
  • -f, --file FNAME — a file containing the prompt, default none. 📄
  • -bf, --binary-file FNAME — binary file containing the prompt, default none. 🗂️
  • -e, --escape, --no-escape — whether to process escape sequences (\n, \r, \t, \', \", \\), default false. 🔄
  • --rope-scaling {none,linear,yarn} — RoPE frequency scaling method, defaults to linear unless specified by the model, env LLAMA_ARG_ROPE_SCALING_TYPE. 🧬
  • --rope-scale N — RoPE context scaling factor, expands context by a factor of N, env LLAMA_ARG_ROPE_SCALE. 📈
  • --rope-freq-base N — RoPE base frequency, used by NTK-aware scaling, default is loaded from model, env LLAMA_ARG_ROPE_FREQ_BASE. 🧬
  • --rope-freq-scale N — RoPE frequency scaling factor, expands context by a factor of 1/N, env LLAMA_ARG_ROPE_FREQ_SCALE. 📉
  • --yarn-orig-ctx N — YaRN: original context size of model, default 0 means model training context size, env LLAMA_ARG_YARN_ORIG_CTX. 🧶
  • --yarn-ext-factor N — YaRN: extrapolation mix factor, default -1.00, 0.0 means full interpolation, env LLAMA_ARG_YARN_EXT_FACTOR. 🧶
  • --yarn-attn-factor N — YaRN: scale sqrt(t) or attention magnitude, default -1.00, env LLAMA_ARG_YARN_ATTN_FACTOR. 🧶
  • --yarn-beta-slow N — YaRN: high correction dim or alpha, default -1.00, env LLAMA_ARG_YARN_BETA_SLOW. 🧶
  • --yarn-beta-fast N — YaRN: low correction dim or beta, default -1.00, env LLAMA_ARG_YARN_BETA_FAST. 🧶
  • -kvo, --kv-offload, -nkvo, --no-kv-offload — whether to enable KV cache offloading, default enabled, env LLAMA_ARG_KV_OFFLOAD. 💾
  • --repack, -nr, --no-repack — whether to enable weight repacking, default enabled, env LLAMA_ARG_REPACK. 📦
  • --no-host — bypass host buffer allowing extra buffers to be used, env LLAMA_ARG_NO_HOST. 🚫
  • -ctk, --cache-type-k TYPE — KV cache data type for K. allowed values: f32, f16, bf16, q8_0, q4_0, q4_1, iq4_nl, q5_0, q5_1, default f16, env LLAMA_ARG_CACHE_TYPE_K. 🧮
  • -ctv, --cache-type-v TYPE — KV cache data type for V. allowed values: f32, f16, bf16, q8_0, q4_0, q4_1, iq4_nl, q5_0, q5_1, default f16, env LLAMA_ARG_CACHE_TYPE_V. 🧮
  • -dt, --defrag-thold N — KV cache defragmentation threshold (DEPRECATED 🚨), env LLAMA_ARG_DEFRAG_THOLD. 💀
  • -np, --parallel N — number of parallel sequences to decode, default 1, env LLAMA_ARG_N_PARALLEL. 🧵
  • --mlock — DEPRECATED 🚨 in favor of --load-mode: force system to keep model in RAM rather than swapping or compressing, env LLAMA_ARG_MLOCK. 💀
  • --mmap, --no-mmap — DEPRECATED 🚨 in favor of --load-mode: whether to memory-map model. (if mmap disabled, slower load but may reduce pageouts if not using mlock), env LLAMA_ARG_MMAP. 💀
  • -dio, --direct-io, -ndio, --no-direct-io — DEPRECATED 🚨 in favor of --load-mode: use DirectIO if available, env LLAMA_ARG_DIO. 💀
  • -lm, --load-mode MODE — model loading mode, default auto. options are:
    • auto: mmap, unless a device does not support it
    • none: no special loading mode
    • mmap: memory-map model (if mmap disabled, slower load but may reduce pageouts if not using mlock)
    • mlock: force system to keep model in RAM rather than swapping or compressing
    • mmap+mlock: mmap + force system to keep model in RAM rather than swapping or compressing
    • dio: use DirectIO if available (env: LLAMA_ARG_LOAD_MODE) 🚀
  • --numa TYPE — attempt optimizations that help on some NUMA systems. options:
    • distribute: spread execution evenly over all nodes
    • isolate: only spawn threads on CPUs on the node that execution started on
    • numactl: use the CPU map provided by numactl if run without this previously, it is recommended to drop the system page cache before using this. see https://github.com/ggml-org/llama.cpp/issues/1437 (env: LLAMA_ARG_NUMA) 🧠
  • -dev, --device <dev1,dev2,..> — comma-separated list of devices to use for offloading (none = don't offload). use --list-devices to see a list of available devices, env LLAMA_ARG_DEVICE. 🎮
  • --list-devices — prints list of available devices and exits, slay. 📋
  • -ot, --override-tensor <tensor name pattern>=<buffer type>,... — override tensor buffer type, env LLAMA_ARG_OVERRIDE_TENSOR. 🧩
  • -cmoe, --cpu-moe — keep all Mixture of Experts (MoE) weights in the CPU, env LLAMA_ARG_CPU_MOE. 🧠
  • -ncmoe, --n-cpu-moe N — keep the Mixture of Experts (MoE) weights of the first N layers in the CPU, env LLAMA_ARG_N_CPU_MOE. 🧠
  • -ngl, --gpu-layers, --n-gpu-layers N — max. number of layers to store in VRAM, either an exact number, 'auto', or 'all', default auto, env LLAMA_ARG_N_GPU_LAYERS. 🎮
  • -sm, --split-mode {none,layer,row,tensor} — how to split the model across multiple GPUs, one of:
    • none: use one GPU only
    • layer (default): split layers and KV across GPUs (pipelined)
    • row: split weight across GPUs by rows (parallelized)
    • tensor: split weights and KV across GPUs (parallelized, EXPERIMENTAL 🧪) (env: LLAMA_ARG_SPLIT_MODE) 🎮
  • -ts, --tensor-split N0,N1,N2,... — fraction of the model to offload to each GPU, comma-separated list of proportions, e.g. 3,1, env LLAMA_ARG_TENSOR_SPLIT. 🎮
  • -mg, --main-gpu INDEX — the GPU to use for the model (with split-mode = none), or for intermediate results and KV (with split-mode = row), default 0, env LLAMA_ARG_MAIN_GPU. 🎮
  • -fit, --fit [on|off] — whether to adjust unset arguments to fit in device memory ('on' or 'off', default 'on'), env LLAMA_ARG_FIT. 🎯
  • -fitt, --fit-target MiB0,MiB1,MiB2,... — target margin per device for --fit, comma-separated list of values, single value is broadcast across all devices, default 1024, env LLAMA_ARG_FIT_TARGET. 🎯
  • -fitc, --fit-ctx N — minimum ctx size that can be set by --fit option, default 4096, env LLAMA_ARG_FIT_CTX. 📏
  • --check-tensors — check model tensor data for invalid values, default false. 🔍
  • --override-kv KEY=TYPE:VALUE,... — advanced option to override model metadata by key. to specify multiple overrides, either use comma-separated values. types: int, float, bool, str. example: --override-kv tokenizer.ggml.add_bos_token=bool:false,tokenizer.ggml.add_eos_token=bool:false. 🧠
  • --op-offload, --no-op-offload — whether to offload host tensor operations to device, default true. 🚀
  • --lora FNAME — path to LoRA adapter (use comma-separated values to load multiple adapters). 🧵
  • --lora-scaled FNAME:SCALE,... — path to LoRA adapter with user defined scaling (format: FNAME:SCALE,...) note: use comma-separated values. 🧵
  • --control-vector FNAME — add a control vector note: use comma-separated values to add multiple control vectors. 🎛️
  • --control-vector-scaled FNAME:SCALE,... — add a control vector with user defined scaling SCALE note: use comma-separated values (format: FNAME:SCALE,...). 🎛️
  • --control-vector-layer-range START END — layer range to apply the control vector(s) to, start and end inclusive. 🎛️
  • -m, --model FNAME — model path to load, env LLAMA_ARG_MODEL. 📁
  • -mu, --model-url MODEL_URL — model download url, default unused, env LLAMA_ARG_MODEL_URL. 🌐
  • -dr, --docker-repo [<repo>/]<model>[:quant]
  • Docker Hub model repository — repo is optional, defaults to ai/. quant is optional, defaults to :latest. Example: gemma3 (default: unused) (env: LLAMA_ARG_DOCKER_REPO)
  • -hf, -hfr, --hf-repo <user>/<model>[:quant] — pulls your model straight off Hugging Face 🦋. The quant bit is optional and case-insensitive, defaults to Q4_K_M, and if there's no Q4_K_M in the repo it gracefully falls back to the first file it finds. It also auto-downloads mmproj if the repo has one — if that's not the vibe, hit it with --no-mmproj. Example: ggml-org/GLM-4.7-Flash-GGUF:Q4_K_M (default: unused) (env: LLAMA_ARG_HF_REPO)
  • -hff, --hf-file FILE — the actual Hugging Face model file. Heads up bestie: if you specify this, it overrides the quant you set in --hf-repo (default: unused) (env: LLAMA_ARG_HF_FILE)
  • -hft, --hf-token TOKEN — your Hugging Face access token 🔑 (default: whatever's in the HF_TOKEN environment variable) (env: HF_TOKEN)
  • --log-disable — Log disable. Yeah. That's the whole description. It disables the log. 💀
  • --log-file FNAME — logs go to a file instead of screaming into your terminal (env: LLAMA_ARG_LOG_FILE)
  • --log-colors [on|off|auto] — sets colored logging. You've got 'on', 'off', or 'auto', and the default is 'auto'. What does 'auto' do? It turns colors on when the output is going to a terminal, like it's reading the room ✨ (env: LLAMA_ARG_LOG_COLORS)
  • -v, --verbose, --log-verbose — cranks the verbosity level to infinity, i.e. it logs every message. Chaotic but genuinely useful for debugging 🔥
  • --offline — offline mode, no cap. It forces the use of cache and flat-out prevents network access (env: LLAMA_ARG_OFFLINE)
  • -lv, --verbosity, --log-verbosity N — sets the verbosity threshold. Any message with a higher verbosity than your threshold gets ignored, sorry babe. The values are:
    • 0: generic output
    • 1: error
    • 2: warning
    • 3: info
    • 4: trace (more info)
    • 5: debug

(default: 3) (env: LLAMA_ARG_LOG_VERBOSITY)

  • --log-prefix, --no-log-prefix — enable the prefix in log messages (env: LLAMA_ARG_LOG_PREFIX)
  • --log-timestamps, --no-log-timestamps — enable timestamps in log messages (env: LLAMA_ARG_LOG_TIMESTAMPS)
  • --spec-draft-type-k, -ctkd, --cache-type-k-draft TYPE — the KV cache data type for K for the draft model. Allowed values: f32, f16, bf16, q8_0, q4_0, q4_1, iq4_nl, q5_0, q5_1 (default: f16) (env: LLAMA_ARG_SPEC_DRAFT_CACHE_TYPE_K)
  • --spec-draft-type-v, -ctvd, --cache-type-v-draft TYPE — same energy but for V, and for the draft model. Allowed values: f32, f16, bf16, q8_0, q4_0, q4_1, iq4_nl, q5_0, q5_1 (default: f16) (env: LLAMA_ARG_SPEC_DRAFT_CACHE_TYPE_V)

🍓 sampling params 🍓

  • --samplers SAMPLERS — the samplers used for generation, in order, separated by ; (default: penalties;dry;top_n_sigma;top_k;typ_p;top_p;min_p;xtc;temperature)
  • -s, --seed SEED — RNG seed (default: -1, and -1 means use a random seed for you 🎲)
  • --sampler-seq, --sampling-seq SEQUENCE — a simplified sequence for the samplers that will be used (default: edskypmxt)
  • --ignore-eos — ignores the end of stream token and just keeps generating, no thoughts head empty 😭 (this implies --logit-bias EOS-inf)
  • --temp, --temperature N — temperature (default: 0.80)
  • --top-k N — top-k sampling (default: 40, 0 = disabled) (env: LLAMA_ARG_TOP_K)
  • --top-p N — top-p sampling (default: 0.95, 1.0 = disabled)
  • --min-p N — min-p sampling (default: 0.05, 0.0 = disabled)
  • --top-nsigma, --top-n-sigma N — top-n-sigma sampling (default: -1.00, -1.0 = disabled)
  • --xtc-probability N — xtc probability (default: 0.00, 0.0 = disabled)
  • --xtc-threshold N — xtc threshold (default: 0.10, 1.0 = disabled)
  • --typical, --typical-p N — locally typical sampling, parameter p (default: 1.00, 1.0 = disabled)
  • --repeat-last-n N — the last n tokens to consider for penalize (default: 64, 0 = disabled)
  • --repeat-penalty N — penalize a repeat sequence of tokens (default: 1.00, 1.0 = disabled)
  • --presence-penalty N — repeat alpha presence penalty (default: 0.00, 0.0 = disabled)
  • --frequency-penalty N — repeat alpha frequency penalty (default: 0.00, 0.0 = disabled)
  • --dry-multiplier N — sets the DRY sampling multiplier (default: 0.00, 0.0 = disabled)
  • --dry-base N — sets the DRY sampling base value (default: 1.75)
  • --dry-allowed-length N — sets the allowed length for DRY sampling (default: 2)
  • --dry-penalty-last-n N — sets the DRY penalty for the last n tokens (default: 64, 0 = disable)
  • --dry-sequence-breaker STRING — adds a sequence breaker for DRY sampling, and when you do that it clears out the default breakers ('\n', ':', '"', '*') in the process. Use "none" if you don't want any sequence breakers at all 🚫
  • --adaptive-target N — adaptive-p: selects tokens near this probability (valid range 0.0 to 1.0; negative = disabled) (default: -1.00) — more info lives at https://github.com/ggml-org/llama.cpp/pull/17927 if you wanna go read the whole thing 👀
  • --adaptive-decay N — adaptive-p: the decay rate for target adaptation over time. Lower values are more reactive, higher values are more stable 📉📈 (valid range 0.0 to 0.99) (default: 0.90)
  • --dynatemp-range N — dynamic temperature range (default: 0.00, 0.0 = disabled)
  • --dynatemp-exp N — dynamic temperature exponent (default: 1.00)
  • --mirostat N — use Mirostat sampling. Important: Top K, Nucleus and Locally Typical samplers are ignored if they're used alongside this. (default: 0, 0 = disabled, 1 = Mirostat, 2 = Mirostat 2.0)
  • --mirostat-lr N — Mirostat learning rate, parameter eta (default: 0.10)
  • --mirostat-ent N — Mirostat target entropy, parameter tau (default: 5.00)
  • -l, --logit-bias TOKEN_ID(+/-)BIAS — modifies the likelihood of a token showing up in the completion, i.e. --logit-bias 15043+1 to increase the likelihood of the token ' Hello', or --logit-bias 15043-1 to decrease the likelihood of the token ' Hello'
  • --grammar GRAMMAR — a BNF-like grammar to constrain generations (check the samples in the grammars/ dir) 🎀
  • --grammar-file FNAME — the file to read the grammar from
  • -j, --json-schema SCHEMA — a JSON schema to constrain generations (https://json-schema.org/), e.g. {} for any JSON object. For schemas with external $refs, use --grammar plus example/json_schema_to_grammar.py instead ✨
  • -jf, --json-schema-file FILE — a file containing a JSON schema to constrain generations (https://json-schema.org/), e.g. {} for any JSON object. Same deal: for schemas with external $refs, use --grammar plus example/json_schema_to_grammar.py instead ✨
  • -bs, --backend-sampling — enable backend sampling (experimental 👩‍🔬) (default: disabled) (env: LLAMA_ARG_BACKEND_SAMPLING)

🎀 example-specific params 🎀

  • --context-shift, --no-context-shift — whether to use context shift on infinite text generation (default: disabled) (env: LLAMA_ARG_CONTEXT_SHIFT)
  • --chunks N — max number of chunks to process (default: -1, -1 = all)
  • --in-file FNAME — an input file (use comma-separated values to specify multiple files)
  • -o, --output, --output-file FNAME — output file (default: 'imatrix.gguf')
  • -ofreq, --output-frequency N — outputs the imatrix every N iterations (default: 10)
  • --output-format {gguf,dat} — the output format for the imatrix file (default: gguf)
  • --save-frequency N — saves an imatrix copy every N iterations (default: 0)
  • --process-output — collects data for the output tensor (default: false)
  • --ppl, --no-ppl — whether to compute perplexity (default: true)
  • --chunk, --from-chunk N — starts processing the input from chunk N (default: 0)
  • --show-statistics — shows imatrix statistics and then peaces out ✌️ (default: false)
  • --parse-special — parses special tokens (chat, tool, etc) (default: false)

example usage:

llama-imatrix \
       -m model.gguf -f some-text.txt [-o imatrix.gguf] [--output-format {gguf,dat}] [--no-ppl] \
       [--process-output] [--chunk 123] [--save-frequency 0] [--output-frequency 10] \
       [--in-file imatrix-prev-0.gguf --in-file imatrix-prev-1.gguf ...] [--parse-special] \
       [--show-statistics] [...]

okay that's the whole page bestie, you ate that 💅 this one ships out of debian, dated September 2026, and it's LLAMA-IMATRIX(1) in the manual — go forth and build your imatrix, you've got this 🫶✨

📦 where this page comes from