Prepare
- Download llama2 gguf format model
- https://huggingface.co/TheBloke/Llama-2-7B-GGUF
- and put it into “./models/llama-2–7b/”
ls -lht models/llama-2-7b
total 7997112
-rw-r--r--@ 1 user staff 3.8G Jan 4 13:12 llama-2-7b.Q4_K_M.gguf- 2. Build the llama.cpp
git clone https://github.com/ggerganov/llama.cpp
cd llama.cpp
make -j- 3. Prepare the fine tuning data
mkdir tinystories && cd tinystories
wget https://huggingface.co/datasets/roneneldan/TinyStories/resolve/main/TinyStories_all_data.tar.gz
tar xf TinyStories_all_data.tar.gz- 4. Then run the fine tune command
./finetune --model-base ./models/llama-2-7b/ggml-model-q4_0.gguf --train-data tinystories/data49.txt --threads 26 --sample-start "<s>" --ctx 512
./finetune --model-base ./models/llama-2-7b/llama-2-7b.Q4_K_M.gguf --train-data data/allinone.md --threads 26 --sample-start "<s>" --ctx 512./finetune --model-base ./models/llama-2-7b/llama-2-7b.Q4_K_M.gguf --train-data data/allinone.md --threads 26 --sample-start "<s>" --ctx 512
main: seed: 1704345676
main: model base = './models/llama-2-7b/llama-2-7b.Q4_K_M.gguf'
llama_model_loader: loaded meta data with 19 key-value pairs and 291 tensors from ./models/llama-2-7b/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] = ["<unk>", "<s>", "</s>", "<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 '<s>'
llm_load_print_meta: EOS token = 2 '</s>'
llm_load_print_meta: UNK token = 0 '<unk>'
llm_load_print_meta: LF token = 13 '<0x0A>'
llm_load_tensors: ggml ctx size = 0.11 MiB
llm_load_tensors: system memory used = 3891.35 MiB
..................................................................................................
llama_new_context_with_model: n_ctx = 512
llama_new_context_with_model: freq_base = 10000.0
llama_new_context_with_model: freq_scale = 1
llama_new_context_with_model: KV self size = 256.00 MiB, K (f16): 128.00 MiB, V (f16): 128.00 MiB
llama_build_graph: non-view tensors processed: 676/676
llama_new_context_with_model: compute buffer total size = 73.69 MiB
main: init model
print_params: n_vocab : 32000
print_params: n_ctx : 512
print_params: n_embd : 4096
print_params: n_ff : 11008
print_params: n_head : 32
print_params: n_head_kv : 32
print_params: n_layer : 32
print_params: norm_rms_eps : 0.000010
print_params: rope_freq_base : 10000.000000
print_params: rope_freq_scale : 1.000000
print_lora_params: n_rank_attention_norm : 1
print_lora_params: n_rank_wq : 4
print_lora_params: n_rank_wk : 4
print_lora_params: n_rank_wv : 4
print_lora_params: n_rank_wo : 4
print_lora_params: n_rank_ffn_norm : 1
print_lora_params: n_rank_w1 : 4
print_lora_params: n_rank_w2 : 4
print_lora_params: n_rank_w3 : 4
print_lora_params: n_rank_tok_embeddings : 4
print_lora_params: n_rank_norm : 1
print_lora_params: n_rank_output : 4
main: total train_iterations 0
main: seen train_samples 0
main: seen train_tokens 0
main: completed train_epochs 0
main: lora_size = 84863776 bytes (80.9 MB)
main: opt_size = 126593008 bytes (120.7 MB)
main: opt iter 0
main: input_size = 524304416 bytes (500.0 MB)
main: compute_size = 24360536160 bytes (23232.0 MB)
main: evaluation order = LEFT_TO_RIGHT
main: tokenize training data
tokenize_file: warning: sample start pattern '<s>' not found. inserting single sample at data begin
tokenize_file: warning: found 1 samples (max length 30173) that exceed context length of 512. samples will be cut off.
tokenize_file: total number of samples: 1
main: number of training tokens: 30173
main: number of unique tokens: 1460
main: train data seems to have changed. restarting shuffled epoch.
main: begin training
main: work_size = 3329736 bytes (3.2 MB)
train_opt_callback: iter= 0 sample=1/1 sched=0.000000 loss=0.000000 |->
train_opt_callback: reshuffle samples. completed epochs: 1
train_opt_callback: iter= 1 sample=1/1 sched=0.010000 loss=2.079438 dt=00:04:04 eta=17:20:26 |->
train_opt_callback: reshuffle samples. completed epochs: 2
- 5. How to run the fine tuned model
./main --interactive --model ./models/llama-2-7b/llama-2-7b.Q4_K_M.gguf --lora ggml-lora-LATEST-f32.gguf --prompt"Can you please write a children's story with 200 words about father and son and friendship and bravery?"
Reference
- https://blog.gopenai.com/how-to-fine-tune-llama-2-on-mac-studio-4b42f317c975
- https://github.com/huggingface/autotrain-advanced
Fine tune llama2 on Mac