This is the home for llama-cpp-2. It also contains the llama-cpp-sys bindings which are updated regularly and in sync with llama-cpp-2.
This project was created with the explict goal of staying as up to date as possible with llama.cpp, as a result it is dead simple, very close to raw bindings, and does not follow semver meaningfully.
Check out the docs.rs for crate documentation or the readme for high level information about the project.
We maintain a super simple example of using the library:
Clone the repo
git clone --recursive https://github.com/utilityai/llama-cpp-rs
cd llama-cpp-rsRun the simple example (add --featues cuda if you have a cuda gpu)
cargo run --release --bin simple -- --prompt "The way to kill a linux process is" hf-model TheBloke/Llama-2-7B-GGUF llama-2-7b.Q4_K_M.ggufOutput
ggml_init_cublas: GGML_CUDA_FORCE_MMQ: no
ggml_init_cublas: CUDA_USE_TENSOR_CORES: yes
ggml_init_cublas: found 1 CUDA devices:
Device 0: NVIDIA GeForce RTX 3090, compute capability 8.6, VMM: yes
llama_model_params { n_gpu_layers: 1000, split_mode: 1, main_gpu: 0, tensor_split: 0x0, progress_callback: None, progress_callback_user_data: 0x0, kv_overrides: 0x0, vocab_only: false, use_mmap: true, use_mlock: false }
llama_model_loader: loaded meta data with 19 key-value pairs and 291 tensors from /home/marcus/.cache/huggingface/hub/models--TheBloke--Llama-2-7B-GGUF/snapshots/b4e04e128f421c93a5f1e34ac4d7ca9b0af47b80/llama-2-7b.Q4_K_M.gguf (version GGUF V2)
llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
llama_model_loader: - kv 0: general.architecture str = llama
llama_model_loader: - kv 1: general.name str = LLaMA v2
llama_model_loader: - kv 2: llama.context_length u32 = 4096
llama_model_loader: - kv 3: llama.embedding_length u32 = 4096
llama_model_loader: - kv 4: llama.block_count u32 = 32
llama_model_loader: - kv 5: llama.feed_forward_length u32 = 11008
llama_model_loader: - kv 6: llama.rope.dimension_count u32 = 128
llama_model_loader: - kv 7: llama.attention.head_count u32 = 32
llama_model_loader: - kv 8: llama.attention.head_count_kv u32 = 32
llama_model_loader: - kv 9: llama.attention.layer_norm_rms_epsilon f32 = 0.000010
llama_model_loader: - kv 10: general.file_type u32 = 15
llama_model_loader: - kv 11: tokenizer.ggml.model str = llama
llama_model_loader: - kv 12: tokenizer.ggml.tokens arr[str,32000] = ["", "", "", "<0x00>", "<...
llama_model_loader: - kv 13: tokenizer.ggml.scores arr[f32,32000] = [0.000000, 0.000000, 0.000000, 0.0000...
llama_model_loader: - kv 14: tokenizer.ggml.token_type arr[i32,32000] = [2, 3, 3, 6, 6, 6, 6, 6, 6, 6, 6, 6, ...
llama_model_loader: - kv 15: tokenizer.ggml.bos_token_id u32 = 1
llama_model_loader: - kv 16: tokenizer.ggml.eos_token_id u32 = 2
llama_model_loader: - kv 17: tokenizer.ggml.unknown_token_id u32 = 0
llama_model_loader: - kv 18: general.quantization_version u32 = 2
llama_model_loader: - type f32: 65 tensors
llama_model_loader: - type q4_K: 193 tensors
llama_model_loader: - type q6_K: 33 tensors
llm_load_vocab: special tokens definition check successful ( 259/32000 ).
llm_load_print_meta: format = GGUF V2
llm_load_print_meta: arch = llama
llm_load_print_meta: vocab type = SPM
llm_load_print_meta: n_vocab = 32000
llm_load_print_meta: n_merges = 0
llm_load_print_meta: n_ctx_train = 4096
llm_load_print_meta: n_embd = 4096
llm_load_print_meta: n_head = 32
llm_load_print_meta: n_head_kv = 32
llm_load_print_meta: n_layer = 32
llm_load_print_meta: n_rot = 128
llm_load_print_meta: n_embd_head_k = 128
llm_load_print_meta: n_embd_head_v = 128
llm_load_print_meta: n_gqa = 1
llm_load_print_meta: n_embd_k_gqa = 4096
llm_load_print_meta: n_embd_v_gqa = 4096
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: n_ff = 11008
llm_load_print_meta: n_expert = 0
llm_load_print_meta: n_expert_used = 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_yarn_orig_ctx = 4096
llm_load_print_meta: rope_finetuned = unknown
llm_load_print_meta: model type = 7B
llm_load_print_meta: model ftype = Q4_K - Medium
llm_load_print_meta: model params = 6.74 B
llm_load_print_meta: model size = 3.80 GiB (4.84 BPW)
llm_load_print_meta: general.name = LLaMA v2
llm_load_print_meta: BOS token = 1 ''
llm_load_print_meta: EOS token = 2 ''
llm_load_print_meta: UNK token = 0 ''
llm_load_print_meta: LF token = 13 '<0x0A>'
llm_load_tensors: ggml ctx size = 0.22 MiB
llm_load_tensors: offloading 32 repeating layers to GPU
llm_load_tensors: offloading non-repeating layers to GPU
llm_load_tensors: offloaded 33/33 layers to GPU
llm_load_tensors: CUDA0 buffer size = 3820.94 MiB
llm_load_tensors: CPU buffer size = 70.31 MiB
..................................................................................................
Loaded "/home/marcus/.cache/huggingface/hub/models--TheBloke--Llama-2-7B-GGUF/snapshots/b4e04e128f421c93a5f1e34ac4d7ca9b0af47b80/llama-2-7b.Q4_K_M.gguf"
llama_new_context_with_model: n_ctx = 2048
llama_new_context_with_model: freq_base = 10000.0
llama_new_context_with_model: freq_scale = 1
llama_kv_cache_init: CUDA0 KV buffer size = 1024.00 MiB
llama_new_context_with_model: KV self size = 1024.00 MiB, K (f16): 512.00 MiB, V (f16): 512.00 MiB
llama_new_context_with_model: CUDA_Host input buffer size = 13.02 MiB
ggml_gallocr_reserve_n: reallocating CUDA0 buffer from size 0.00 MiB to 164.01 MiB
ggml_gallocr_reserve_n: reallocating CUDA_Host buffer from size 0.00 MiB to 8.00 MiB
llama_new_context_with_model: CUDA0 compute buffer size = 164.01 MiB
llama_new_context_with_model: CUDA_Host compute buffer size = 8.00 MiB
llama_new_context_with_model: graph splits (measure): 3
n_len = 32, n_ctx = 2048, k_kv_req = 32
The way to kill a linux process is to send it a SIGKILL signal.
The way to kill a windows process is to send it a S
decoded 24 tokens in 0.23 s, speed 105.65 t/s
load time = 727.50 ms
sample time = 0.46 ms / 24 runs (0.02 ms per token, 51835.85 tokens per second)
prompt eval time = 68.52 ms / 9 tokens (7.61 ms per token, 131.35 tokens per second)
eval time = 225.70 ms / 24 runs (9.40 ms per token, 106.34 tokens per second)
total time = 954.18 ms
Ensure that when you clone this project you also clone the submodules. This can be done with the following command:
git clone --recursive https://github.com/utilityai/llama-cpp-rsor if you have already cloned the project you can run:
git submodule update --init --recursive