add example output for easier documentation for those trying to replicate
Browse files- README.md +3 -239
- convert_hf_to_gguf.output.txt +582 -0
- llamafile-creation.sh +2 -2
- llamafile_output_example.output.txt +806 -0
README.md
CHANGED
|
@@ -14,8 +14,6 @@ tags:
|
|
| 14 |
- Model creator: [Maykeye](https://huggingface.co/Maykeye)
|
| 15 |
- Original model: [TinyLLama-v0](https://huggingface.co/Maykeye/TinyLLama-v0)
|
| 16 |
|
| 17 |
-
If interested in the internal content of this model you can check [Tinyllama-4.6M-v0.0-F16.dump.md](./Tinyllama-4.6M-v0.0-F16.dump.md) included in this repo.
|
| 18 |
-
|
| 19 |
## Description
|
| 20 |
|
| 21 |
* This repo is targeted towards:
|
|
@@ -54,7 +52,7 @@ chmod +x Tinyllama-5M-v0.2-F16.llamafile
|
|
| 54 |
|
| 55 |
## About llamafile
|
| 56 |
|
| 57 |
-
llamafile is a new format introduced by Mozilla Ocho on Nov 20th 2023. It uses Cosmopolitan Libc to turn LLM weights into runnable llama.cpp binaries that run on the stock installs of six OSes for both ARM64 and AMD64.
|
| 58 |
|
| 59 |
## Replication Steps Assumption
|
| 60 |
|
|
@@ -66,240 +64,6 @@ llamafile is a new format introduced by Mozilla Ocho on Nov 20th 2023. It uses C
|
|
| 66 |
|
| 67 |
For the most current replication steps, refer to the bash script `llamafile-creation.sh` in this repo.
|
| 68 |
|
| 69 |
-
|
| 70 |
-
$ ./llamafile-creation.sh
|
| 71 |
-
== Prep Enviroment ==
|
| 72 |
-
== Build and prep the llamafile engine execuable ==
|
| 73 |
-
~/huggingface/TinyLLama-v0-5M-F16-llamafile/llamafile ~/huggingface/TinyLLama-v0-5M-F16-llamafile
|
| 74 |
-
make: Nothing to be done for 'all'.
|
| 75 |
-
make: Nothing to be done for 'all'.
|
| 76 |
-
~/huggingface/TinyLLama-v0-5M-F16-llamafile
|
| 77 |
-
== What is our llamafile name going to be? ==
|
| 78 |
-
maykeye_tinyllama/Tinyllama-4.6M-v0.0-F16.gguf
|
| 79 |
-
We will be aiming to generate Tinyllama-4.6M-v0.0-F16.llamafile
|
| 80 |
-
== Convert from safetensor to gguf ==
|
| 81 |
-
INFO:hf-to-gguf:Loading model: maykeye_tinyllama
|
| 82 |
-
INFO:gguf.gguf_writer:gguf: This GGUF file is for Little Endian only
|
| 83 |
-
INFO:hf-to-gguf:Exporting model...
|
| 84 |
-
INFO:hf-to-gguf:gguf: loading model part 'model.safetensors'
|
| 85 |
-
INFO:hf-to-gguf:output.weight, torch.bfloat16 --> F16, shape = {64, 32000}
|
| 86 |
-
INFO:hf-to-gguf:token_embd.weight, torch.bfloat16 --> F16, shape = {64, 32000}
|
| 87 |
-
INFO:hf-to-gguf:blk.0.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 88 |
-
INFO:hf-to-gguf:blk.0.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 89 |
-
INFO:hf-to-gguf:blk.0.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 90 |
-
INFO:hf-to-gguf:blk.0.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 91 |
-
INFO:hf-to-gguf:blk.0.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 92 |
-
INFO:hf-to-gguf:blk.0.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 93 |
-
INFO:hf-to-gguf:blk.0.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 94 |
-
INFO:hf-to-gguf:blk.0.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 95 |
-
INFO:hf-to-gguf:blk.0.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 96 |
-
INFO:hf-to-gguf:blk.1.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 97 |
-
INFO:hf-to-gguf:blk.1.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 98 |
-
INFO:hf-to-gguf:blk.1.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 99 |
-
INFO:hf-to-gguf:blk.1.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 100 |
-
INFO:hf-to-gguf:blk.1.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 101 |
-
INFO:hf-to-gguf:blk.1.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 102 |
-
INFO:hf-to-gguf:blk.1.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 103 |
-
INFO:hf-to-gguf:blk.1.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 104 |
-
INFO:hf-to-gguf:blk.1.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 105 |
-
INFO:hf-to-gguf:blk.2.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 106 |
-
INFO:hf-to-gguf:blk.2.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 107 |
-
INFO:hf-to-gguf:blk.2.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 108 |
-
INFO:hf-to-gguf:blk.2.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 109 |
-
INFO:hf-to-gguf:blk.2.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 110 |
-
INFO:hf-to-gguf:blk.2.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 111 |
-
INFO:hf-to-gguf:blk.2.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 112 |
-
INFO:hf-to-gguf:blk.2.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 113 |
-
INFO:hf-to-gguf:blk.2.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 114 |
-
INFO:hf-to-gguf:blk.3.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 115 |
-
INFO:hf-to-gguf:blk.3.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 116 |
-
INFO:hf-to-gguf:blk.3.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 117 |
-
INFO:hf-to-gguf:blk.3.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 118 |
-
INFO:hf-to-gguf:blk.3.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 119 |
-
INFO:hf-to-gguf:blk.3.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 120 |
-
INFO:hf-to-gguf:blk.3.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 121 |
-
INFO:hf-to-gguf:blk.3.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 122 |
-
INFO:hf-to-gguf:blk.3.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 123 |
-
INFO:hf-to-gguf:blk.4.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 124 |
-
INFO:hf-to-gguf:blk.4.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 125 |
-
INFO:hf-to-gguf:blk.4.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 126 |
-
INFO:hf-to-gguf:blk.4.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 127 |
-
INFO:hf-to-gguf:blk.4.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 128 |
-
INFO:hf-to-gguf:blk.4.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 129 |
-
INFO:hf-to-gguf:blk.4.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 130 |
-
INFO:hf-to-gguf:blk.4.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 131 |
-
INFO:hf-to-gguf:blk.4.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 132 |
-
INFO:hf-to-gguf:blk.5.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 133 |
-
INFO:hf-to-gguf:blk.5.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 134 |
-
INFO:hf-to-gguf:blk.5.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 135 |
-
INFO:hf-to-gguf:blk.5.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 136 |
-
INFO:hf-to-gguf:blk.5.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 137 |
-
INFO:hf-to-gguf:blk.5.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 138 |
-
INFO:hf-to-gguf:blk.5.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 139 |
-
INFO:hf-to-gguf:blk.5.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 140 |
-
INFO:hf-to-gguf:blk.5.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 141 |
-
INFO:hf-to-gguf:blk.6.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 142 |
-
INFO:hf-to-gguf:blk.6.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 143 |
-
INFO:hf-to-gguf:blk.6.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 144 |
-
INFO:hf-to-gguf:blk.6.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 145 |
-
INFO:hf-to-gguf:blk.6.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 146 |
-
INFO:hf-to-gguf:blk.6.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 147 |
-
INFO:hf-to-gguf:blk.6.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 148 |
-
INFO:hf-to-gguf:blk.6.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 149 |
-
INFO:hf-to-gguf:blk.6.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 150 |
-
INFO:hf-to-gguf:blk.7.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 151 |
-
INFO:hf-to-gguf:blk.7.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 152 |
-
INFO:hf-to-gguf:blk.7.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 153 |
-
INFO:hf-to-gguf:blk.7.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 154 |
-
INFO:hf-to-gguf:blk.7.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 155 |
-
INFO:hf-to-gguf:blk.7.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 156 |
-
INFO:hf-to-gguf:blk.7.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 157 |
-
INFO:hf-to-gguf:blk.7.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 158 |
-
INFO:hf-to-gguf:blk.7.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 159 |
-
INFO:hf-to-gguf:output_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 160 |
-
INFO:hf-to-gguf:Set meta model
|
| 161 |
-
INFO:hf-to-gguf:Set model parameters
|
| 162 |
-
INFO:hf-to-gguf:gguf: context length = 2048
|
| 163 |
-
INFO:hf-to-gguf:gguf: embedding length = 64
|
| 164 |
-
INFO:hf-to-gguf:gguf: feed forward length = 256
|
| 165 |
-
INFO:hf-to-gguf:gguf: head count = 16
|
| 166 |
-
INFO:hf-to-gguf:gguf: rms norm epsilon = 1e-06
|
| 167 |
-
INFO:hf-to-gguf:gguf: file type = 1
|
| 168 |
-
INFO:hf-to-gguf:Set model tokenizer
|
| 169 |
-
INFO:gguf.vocab:Setting special token type bos to 1
|
| 170 |
-
INFO:gguf.vocab:Setting special token type eos to 2
|
| 171 |
-
INFO:gguf.vocab:Setting special token type unk to 0
|
| 172 |
-
INFO:gguf.vocab:Setting special token type pad to 0
|
| 173 |
-
INFO:hf-to-gguf:Set model quantization version
|
| 174 |
-
INFO:gguf.gguf_writer:Writing the following files:
|
| 175 |
-
INFO:gguf.gguf_writer:maykeye_tinyllama/Tinyllama-4.6M-v0.0-F16.gguf: n_tensors = 75, total_size = 9.2M
|
| 176 |
-
Writing: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 9.24M/9.24M [00:00<00:00, 83.7Mbyte/s]
|
| 177 |
-
INFO:hf-to-gguf:Model successfully exported to maykeye_tinyllama/Tinyllama-4.6M-v0.0-F16.gguf
|
| 178 |
-
== Generating Llamafile ==
|
| 179 |
-
== Test Output ./Tinyllama-4.6M-v0.0-F16.llamafile ==
|
| 180 |
-
note: if you have an AMD or NVIDIA GPU then you need to pass -ngl 9999 to enable GPU offloading
|
| 181 |
-
main: llamafile version 0.8.9
|
| 182 |
-
main: seed = 1721461448
|
| 183 |
-
llama_model_loader: loaded meta data with 33 key-value pairs and 75 tensors from Tinyllama-4.6M-v0.0-F16.gguf (version GGUF V3 (latest))
|
| 184 |
-
llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
|
| 185 |
-
llama_model_loader: - kv 0: general.architecture str = llama
|
| 186 |
-
llama_model_loader: - kv 1: general.type str = model
|
| 187 |
-
llama_model_loader: - kv 2: general.name str = TinyLLama
|
| 188 |
-
llama_model_loader: - kv 3: general.author str = Maykeye
|
| 189 |
-
llama_model_loader: - kv 4: general.version str = v0.0
|
| 190 |
-
llama_model_loader: - kv 5: general.description str = This gguf is ported from a first vers...
|
| 191 |
-
llama_model_loader: - kv 6: general.quantized_by str = Mofosyne
|
| 192 |
-
llama_model_loader: - kv 7: general.size_label str = 4.6M
|
| 193 |
-
llama_model_loader: - kv 8: general.license str = apache-2.0
|
| 194 |
-
llama_model_loader: - kv 9: general.url str = https://huggingface.co/mofosyne/TinyL...
|
| 195 |
-
llama_model_loader: - kv 10: general.source.url str = https://huggingface.co/Maykeye/TinyLL...
|
| 196 |
-
llama_model_loader: - kv 11: general.tags arr[str,5] = ["text generation", "transformer", "l...
|
| 197 |
-
llama_model_loader: - kv 12: general.languages arr[str,1] = ["en"]
|
| 198 |
-
llama_model_loader: - kv 13: general.datasets arr[str,2] = ["https://huggingface.co/datasets/ron...
|
| 199 |
-
llama_model_loader: - kv 14: llama.block_count u32 = 8
|
| 200 |
-
llama_model_loader: - kv 15: llama.context_length u32 = 2048
|
| 201 |
-
llama_model_loader: - kv 16: llama.embedding_length u32 = 64
|
| 202 |
-
llama_model_loader: - kv 17: llama.feed_forward_length u32 = 256
|
| 203 |
-
llama_model_loader: - kv 18: llama.attention.head_count u32 = 16
|
| 204 |
-
llama_model_loader: - kv 19: llama.attention.layer_norm_rms_epsilon f32 = 0.000001
|
| 205 |
-
llama_model_loader: - kv 20: general.file_type u32 = 1
|
| 206 |
-
llama_model_loader: - kv 21: llama.vocab_size u32 = 32000
|
| 207 |
-
llama_model_loader: - kv 22: llama.rope.dimension_count u32 = 4
|
| 208 |
-
llama_model_loader: - kv 23: tokenizer.ggml.model str = llama
|
| 209 |
-
llama_model_loader: - kv 24: tokenizer.ggml.pre str = default
|
| 210 |
-
llama_model_loader: - kv 25: tokenizer.ggml.tokens arr[str,32000] = ["<unk>", "<s>", "</s>", "<0x00>", "<...
|
| 211 |
-
llama_model_loader: - kv 26: tokenizer.ggml.scores arr[f32,32000] = [0.000000, 0.000000, 0.000000, 0.0000...
|
| 212 |
-
llama_model_loader: - kv 27: tokenizer.ggml.token_type arr[i32,32000] = [2, 3, 3, 6, 6, 6, 6, 6, 6, 6, 6, 6, ...
|
| 213 |
-
llama_model_loader: - kv 28: tokenizer.ggml.bos_token_id u32 = 1
|
| 214 |
-
llama_model_loader: - kv 29: tokenizer.ggml.eos_token_id u32 = 2
|
| 215 |
-
llama_model_loader: - kv 30: tokenizer.ggml.unknown_token_id u32 = 0
|
| 216 |
-
llama_model_loader: - kv 31: tokenizer.ggml.padding_token_id u32 = 0
|
| 217 |
-
llama_model_loader: - kv 32: general.quantization_version u32 = 2
|
| 218 |
-
llama_model_loader: - type f32: 17 tensors
|
| 219 |
-
llama_model_loader: - type f16: 58 tensors
|
| 220 |
-
llm_load_vocab: special tokens definition check successful ( 259/32000 ).
|
| 221 |
-
llm_load_print_meta: format = GGUF V3 (latest)
|
| 222 |
-
llm_load_print_meta: arch = llama
|
| 223 |
-
llm_load_print_meta: vocab type = SPM
|
| 224 |
-
llm_load_print_meta: n_vocab = 32000
|
| 225 |
-
llm_load_print_meta: n_merges = 0
|
| 226 |
-
llm_load_print_meta: n_ctx_train = 2048
|
| 227 |
-
llm_load_print_meta: n_embd = 64
|
| 228 |
-
llm_load_print_meta: n_head = 16
|
| 229 |
-
llm_load_print_meta: n_head_kv = 16
|
| 230 |
-
llm_load_print_meta: n_layer = 8
|
| 231 |
-
llm_load_print_meta: n_rot = 4
|
| 232 |
-
llm_load_print_meta: n_swa = 0
|
| 233 |
-
llm_load_print_meta: n_embd_head_k = 4
|
| 234 |
-
llm_load_print_meta: n_embd_head_v = 4
|
| 235 |
-
llm_load_print_meta: n_gqa = 1
|
| 236 |
-
llm_load_print_meta: n_embd_k_gqa = 64
|
| 237 |
-
llm_load_print_meta: n_embd_v_gqa = 64
|
| 238 |
-
llm_load_print_meta: f_norm_eps = 0.0e+00
|
| 239 |
-
llm_load_print_meta: f_norm_rms_eps = 1.0e-06
|
| 240 |
-
llm_load_print_meta: f_clamp_kqv = 0.0e+00
|
| 241 |
-
llm_load_print_meta: f_max_alibi_bias = 0.0e+00
|
| 242 |
-
llm_load_print_meta: f_logit_scale = 0.0e+00
|
| 243 |
-
llm_load_print_meta: n_ff = 256
|
| 244 |
-
llm_load_print_meta: n_expert = 0
|
| 245 |
-
llm_load_print_meta: n_expert_used = 0
|
| 246 |
-
llm_load_print_meta: causal attn = 1
|
| 247 |
-
llm_load_print_meta: pooling type = 0
|
| 248 |
-
llm_load_print_meta: rope type = 0
|
| 249 |
-
llm_load_print_meta: rope scaling = linear
|
| 250 |
-
llm_load_print_meta: freq_base_train = 10000.0
|
| 251 |
-
llm_load_print_meta: freq_scale_train = 1
|
| 252 |
-
llm_load_print_meta: n_yarn_orig_ctx = 2048
|
| 253 |
-
llm_load_print_meta: rope_finetuned = unknown
|
| 254 |
-
llm_load_print_meta: ssm_d_conv = 0
|
| 255 |
-
llm_load_print_meta: ssm_d_inner = 0
|
| 256 |
-
llm_load_print_meta: ssm_d_state = 0
|
| 257 |
-
llm_load_print_meta: ssm_dt_rank = 0
|
| 258 |
-
llm_load_print_meta: model type = ?B
|
| 259 |
-
llm_load_print_meta: model ftype = F16
|
| 260 |
-
llm_load_print_meta: model params = 4.62 M
|
| 261 |
-
llm_load_print_meta: model size = 8.82 MiB (16.00 BPW)
|
| 262 |
-
llm_load_print_meta: general.name = TinyLLama
|
| 263 |
-
llm_load_print_meta: BOS token = 1 '<s>'
|
| 264 |
-
llm_load_print_meta: EOS token = 2 '</s>'
|
| 265 |
-
llm_load_print_meta: UNK token = 0 '<unk>'
|
| 266 |
-
llm_load_print_meta: PAD token = 0 '<unk>'
|
| 267 |
-
llm_load_print_meta: LF token = 13 '<0x0A>'
|
| 268 |
-
llm_load_tensors: ggml ctx size = 0.04 MiB
|
| 269 |
-
llm_load_tensors: CPU buffer size = 8.82 MiB
|
| 270 |
-
..............
|
| 271 |
-
llama_new_context_with_model: n_ctx = 512
|
| 272 |
-
llama_new_context_with_model: n_batch = 512
|
| 273 |
-
llama_new_context_with_model: n_ubatch = 512
|
| 274 |
-
llama_new_context_with_model: flash_attn = 0
|
| 275 |
-
llama_new_context_with_model: freq_base = 10000.0
|
| 276 |
-
llama_new_context_with_model: freq_scale = 1
|
| 277 |
-
llama_kv_cache_init: CPU KV buffer size = 1.00 MiB
|
| 278 |
-
llama_new_context_with_model: KV self size = 1.00 MiB, K (f16): 0.50 MiB, V (f16): 0.50 MiB
|
| 279 |
-
llama_new_context_with_model: CPU output buffer size = 0.12 MiB
|
| 280 |
-
llama_new_context_with_model: CPU compute buffer size = 62.75 MiB
|
| 281 |
-
llama_new_context_with_model: graph nodes = 262
|
| 282 |
-
llama_new_context_with_model: graph splits = 1
|
| 283 |
-
|
| 284 |
-
system_info: n_threads = 4 / 8 | AVX = 1 | AVX_VNNI = 0 | AVX2 = 1 | AVX512 = 0 | AVX512_VBMI = 0 | AVX512_VNNI = 0 | AVX512_BF16 = 0 | FMA = 1 | NEON = 0 | ARM_FMA = 0 | F16C = 1 | FP16_VA = 0 | WASM_SIMD = 0 | BLAS = 0 | SSE3 = 1 | SSSE3 = 1 | VSX = 0 | MATMUL_INT8 = 0 | LLAMAFILE = 1 |
|
| 285 |
-
sampling:
|
| 286 |
-
repeat_last_n = 64, repeat_penalty = 1.000, frequency_penalty = 0.000, presence_penalty = 0.000
|
| 287 |
-
top_k = 40, tfs_z = 1.000, top_p = 0.950, min_p = 0.050, typical_p = 1.000, temp = 0.800
|
| 288 |
-
mirostat = 0, mirostat_lr = 0.100, mirostat_ent = 5.000
|
| 289 |
-
sampling order:
|
| 290 |
-
CFG -> Penalties -> top_k -> tfs_z -> typical_p -> top_p -> min_p -> temperature
|
| 291 |
-
generate: n_ctx = 512, n_batch = 2048, n_predict = -1, n_keep = 1
|
| 292 |
-
|
| 293 |
|
| 294 |
-
|
| 295 |
-
One day, the man decided to go in and he took off his own new home. He gave the bird a little bit of his friend. He said he had to find a way to hide it in his woods. The man was very happy, but he knew he needed to make it in the yard.
|
| 296 |
-
The man was very sad and he could not find the bird. He didn't want to get to the park and his friend was very sad. They could not find the bird and his friend. But the man was too sad. He had no friends and no friends. [end of text]
|
| 297 |
-
|
| 298 |
-
|
| 299 |
-
llama_print_timings: load time = 10.26 ms
|
| 300 |
-
llama_print_timings: sample time = 6.03 ms / 156 runs ( 0.04 ms per token, 25879.23 tokens per second)
|
| 301 |
-
llama_print_timings: prompt eval time = 2.16 ms / 8 tokens ( 0.27 ms per token, 3696.86 tokens per second)
|
| 302 |
-
llama_print_timings: eval time = 748.08 ms / 155 runs ( 4.83 ms per token, 207.20 tokens per second)
|
| 303 |
-
llama_print_timings: total time = 800.80 ms / 163 tokens
|
| 304 |
-
Log end
|
| 305 |
-
```
|
|
|
|
| 14 |
- Model creator: [Maykeye](https://huggingface.co/Maykeye)
|
| 15 |
- Original model: [TinyLLama-v0](https://huggingface.co/Maykeye/TinyLLama-v0)
|
| 16 |
|
|
|
|
|
|
|
| 17 |
## Description
|
| 18 |
|
| 19 |
* This repo is targeted towards:
|
|
|
|
| 52 |
|
| 53 |
## About llamafile
|
| 54 |
|
| 55 |
+
[llamafile](https://github.com/Mozilla-Ocho/llamafile) is a new format introduced by Mozilla Ocho on Nov 20th 2023. It uses [Cosmopolitan Libc](https://github.com/jart/cosmopolitan) to turn LLM weights into runnable [llama.cpp](https://github.com/ggerganov/llama.cpp) binaries that run on the stock installs of six OSes for both ARM64 and AMD64.
|
| 56 |
|
| 57 |
## Replication Steps Assumption
|
| 58 |
|
|
|
|
| 64 |
|
| 65 |
For the most current replication steps, refer to the bash script `llamafile-creation.sh` in this repo.
|
| 66 |
|
| 67 |
+
You may want to also check these [output of convert_hf_to_gguf](convert_hf_to_gguf.output.txt) and [output of the generated llamafile](./llamafile_output_example.output.txt) to sanity check that you got the right process if you feel like your output doesn't quite make sense.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 68 |
|
| 69 |
+
If interested in the internal content of this model you can check [Tinyllama-4.6M-v0.0-F16.dump.md](./Tinyllama-4.6M-v0.0-F16.dump.md) included in this repo as well which would be helpful if you are curious about the structure of a gguf file.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
convert_hf_to_gguf.output.txt
ADDED
|
@@ -0,0 +1,582 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
INFO:hf-to-gguf:Loading model: maykeye_tinyllama
|
| 2 |
+
INFO:gguf.gguf_writer:gguf: This GGUF file is for Little Endian only
|
| 3 |
+
INFO:hf-to-gguf:Exporting model...
|
| 4 |
+
INFO:hf-to-gguf:gguf: loading model part 'model.safetensors'
|
| 5 |
+
INFO:hf-to-gguf:output.weight, torch.bfloat16 --> F16, shape = {64, 32000}
|
| 6 |
+
INFO:hf-to-gguf:token_embd.weight, torch.bfloat16 --> F16, shape = {64, 32000}
|
| 7 |
+
INFO:hf-to-gguf:blk.0.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 8 |
+
INFO:hf-to-gguf:blk.0.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 9 |
+
INFO:hf-to-gguf:blk.0.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 10 |
+
INFO:hf-to-gguf:blk.0.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 11 |
+
INFO:hf-to-gguf:blk.0.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 12 |
+
INFO:hf-to-gguf:blk.0.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 13 |
+
INFO:hf-to-gguf:blk.0.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 14 |
+
INFO:hf-to-gguf:blk.0.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 15 |
+
INFO:hf-to-gguf:blk.0.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 16 |
+
INFO:hf-to-gguf:blk.1.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 17 |
+
INFO:hf-to-gguf:blk.1.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 18 |
+
INFO:hf-to-gguf:blk.1.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 19 |
+
INFO:hf-to-gguf:blk.1.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 20 |
+
INFO:hf-to-gguf:blk.1.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 21 |
+
INFO:hf-to-gguf:blk.1.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 22 |
+
INFO:hf-to-gguf:blk.1.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 23 |
+
INFO:hf-to-gguf:blk.1.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 24 |
+
INFO:hf-to-gguf:blk.1.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 25 |
+
INFO:hf-to-gguf:blk.2.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 26 |
+
INFO:hf-to-gguf:blk.2.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 27 |
+
INFO:hf-to-gguf:blk.2.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 28 |
+
INFO:hf-to-gguf:blk.2.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 29 |
+
INFO:hf-to-gguf:blk.2.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 30 |
+
INFO:hf-to-gguf:blk.2.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 31 |
+
INFO:hf-to-gguf:blk.2.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 32 |
+
INFO:hf-to-gguf:blk.2.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 33 |
+
INFO:hf-to-gguf:blk.2.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 34 |
+
INFO:hf-to-gguf:blk.3.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 35 |
+
INFO:hf-to-gguf:blk.3.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 36 |
+
INFO:hf-to-gguf:blk.3.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 37 |
+
INFO:hf-to-gguf:blk.3.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 38 |
+
INFO:hf-to-gguf:blk.3.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 39 |
+
INFO:hf-to-gguf:blk.3.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 40 |
+
INFO:hf-to-gguf:blk.3.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 41 |
+
INFO:hf-to-gguf:blk.3.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 42 |
+
INFO:hf-to-gguf:blk.3.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 43 |
+
INFO:hf-to-gguf:blk.4.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 44 |
+
INFO:hf-to-gguf:blk.4.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 45 |
+
INFO:hf-to-gguf:blk.4.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 46 |
+
INFO:hf-to-gguf:blk.4.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 47 |
+
INFO:hf-to-gguf:blk.4.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 48 |
+
INFO:hf-to-gguf:blk.4.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 49 |
+
INFO:hf-to-gguf:blk.4.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 50 |
+
INFO:hf-to-gguf:blk.4.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 51 |
+
INFO:hf-to-gguf:blk.4.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 52 |
+
INFO:hf-to-gguf:blk.5.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 53 |
+
INFO:hf-to-gguf:blk.5.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 54 |
+
INFO:hf-to-gguf:blk.5.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 55 |
+
INFO:hf-to-gguf:blk.5.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 56 |
+
INFO:hf-to-gguf:blk.5.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 57 |
+
INFO:hf-to-gguf:blk.5.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 58 |
+
INFO:hf-to-gguf:blk.5.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 59 |
+
INFO:hf-to-gguf:blk.5.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 60 |
+
INFO:hf-to-gguf:blk.5.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 61 |
+
INFO:hf-to-gguf:blk.6.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 62 |
+
INFO:hf-to-gguf:blk.6.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 63 |
+
INFO:hf-to-gguf:blk.6.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 64 |
+
INFO:hf-to-gguf:blk.6.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 65 |
+
INFO:hf-to-gguf:blk.6.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 66 |
+
INFO:hf-to-gguf:blk.6.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 67 |
+
INFO:hf-to-gguf:blk.6.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 68 |
+
INFO:hf-to-gguf:blk.6.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 69 |
+
INFO:hf-to-gguf:blk.6.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 70 |
+
INFO:hf-to-gguf:blk.7.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 71 |
+
INFO:hf-to-gguf:blk.7.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 72 |
+
INFO:hf-to-gguf:blk.7.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 73 |
+
INFO:hf-to-gguf:blk.7.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 74 |
+
INFO:hf-to-gguf:blk.7.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 75 |
+
INFO:hf-to-gguf:blk.7.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 76 |
+
INFO:hf-to-gguf:blk.7.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 77 |
+
INFO:hf-to-gguf:blk.7.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 78 |
+
INFO:hf-to-gguf:blk.7.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 79 |
+
INFO:hf-to-gguf:output_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 80 |
+
INFO:hf-to-gguf:Set meta model
|
| 81 |
+
INFO:hf-to-gguf:Set model parameters
|
| 82 |
+
INFO:hf-to-gguf:gguf: context length = 2048
|
| 83 |
+
INFO:hf-to-gguf:gguf: embedding length = 64
|
| 84 |
+
INFO:hf-to-gguf:gguf: feed forward length = 256
|
| 85 |
+
INFO:hf-to-gguf:gguf: head count = 16
|
| 86 |
+
INFO:hf-to-gguf:gguf: rms norm epsilon = 1e-06
|
| 87 |
+
INFO:hf-to-gguf:gguf: file type = 1
|
| 88 |
+
INFO:hf-to-gguf:Set model tokenizer
|
| 89 |
+
INFO:gguf.vocab:Setting special token type bos to 1
|
| 90 |
+
INFO:gguf.vocab:Setting special token type eos to 2
|
| 91 |
+
INFO:gguf.vocab:Setting special token type unk to 0
|
| 92 |
+
INFO:gguf.vocab:Setting special token type pad to 0
|
| 93 |
+
INFO:hf-to-gguf:Set model quantization version
|
| 94 |
+
INFO:gguf.gguf_writer:Writing the following files:
|
| 95 |
+
INFO:gguf.gguf_writer:maykeye_tinyllama/TinyLLama-4.6M-v0.0-F16.gguf: n_tensors = 75, total_size = 9.2M
|
| 96 |
+
|
| 97 |
+
INFO:hf-to-gguf:Model successfully exported to maykeye_tinyllama/TinyLLama-4.6M-v0.0-F16.gguf
|
| 98 |
+
INFO:hf-to-gguf:Loading model: maykeye_tinyllama
|
| 99 |
+
INFO:gguf.gguf_writer:gguf: This GGUF file is for Little Endian only
|
| 100 |
+
INFO:hf-to-gguf:Exporting model...
|
| 101 |
+
INFO:hf-to-gguf:gguf: loading model part 'model.safetensors'
|
| 102 |
+
INFO:hf-to-gguf:output.weight, torch.bfloat16 --> F16, shape = {64, 32000}
|
| 103 |
+
INFO:hf-to-gguf:token_embd.weight, torch.bfloat16 --> F16, shape = {64, 32000}
|
| 104 |
+
INFO:hf-to-gguf:blk.0.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 105 |
+
INFO:hf-to-gguf:blk.0.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 106 |
+
INFO:hf-to-gguf:blk.0.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 107 |
+
INFO:hf-to-gguf:blk.0.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 108 |
+
INFO:hf-to-gguf:blk.0.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 109 |
+
INFO:hf-to-gguf:blk.0.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 110 |
+
INFO:hf-to-gguf:blk.0.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 111 |
+
INFO:hf-to-gguf:blk.0.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 112 |
+
INFO:hf-to-gguf:blk.0.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 113 |
+
INFO:hf-to-gguf:blk.1.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 114 |
+
INFO:hf-to-gguf:blk.1.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 115 |
+
INFO:hf-to-gguf:blk.1.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 116 |
+
INFO:hf-to-gguf:blk.1.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 117 |
+
INFO:hf-to-gguf:blk.1.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 118 |
+
INFO:hf-to-gguf:blk.1.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 119 |
+
INFO:hf-to-gguf:blk.1.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 120 |
+
INFO:hf-to-gguf:blk.1.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 121 |
+
INFO:hf-to-gguf:blk.1.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 122 |
+
INFO:hf-to-gguf:blk.2.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 123 |
+
INFO:hf-to-gguf:blk.2.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 124 |
+
INFO:hf-to-gguf:blk.2.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 125 |
+
INFO:hf-to-gguf:blk.2.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 126 |
+
INFO:hf-to-gguf:blk.2.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 127 |
+
INFO:hf-to-gguf:blk.2.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 128 |
+
INFO:hf-to-gguf:blk.2.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 129 |
+
INFO:hf-to-gguf:blk.2.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 130 |
+
INFO:hf-to-gguf:blk.2.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 131 |
+
INFO:hf-to-gguf:blk.3.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 132 |
+
INFO:hf-to-gguf:blk.3.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 133 |
+
INFO:hf-to-gguf:blk.3.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 134 |
+
INFO:hf-to-gguf:blk.3.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 135 |
+
INFO:hf-to-gguf:blk.3.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 136 |
+
INFO:hf-to-gguf:blk.3.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 137 |
+
INFO:hf-to-gguf:blk.3.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 138 |
+
INFO:hf-to-gguf:blk.3.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 139 |
+
INFO:hf-to-gguf:blk.3.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 140 |
+
INFO:hf-to-gguf:blk.4.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 141 |
+
INFO:hf-to-gguf:blk.4.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 142 |
+
INFO:hf-to-gguf:blk.4.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 143 |
+
INFO:hf-to-gguf:blk.4.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 144 |
+
INFO:hf-to-gguf:blk.4.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 145 |
+
INFO:hf-to-gguf:blk.4.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 146 |
+
INFO:hf-to-gguf:blk.4.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 147 |
+
INFO:hf-to-gguf:blk.4.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 148 |
+
INFO:hf-to-gguf:blk.4.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 149 |
+
INFO:hf-to-gguf:blk.5.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 150 |
+
INFO:hf-to-gguf:blk.5.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 151 |
+
INFO:hf-to-gguf:blk.5.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 152 |
+
INFO:hf-to-gguf:blk.5.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 153 |
+
INFO:hf-to-gguf:blk.5.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 154 |
+
INFO:hf-to-gguf:blk.5.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 155 |
+
INFO:hf-to-gguf:blk.5.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 156 |
+
INFO:hf-to-gguf:blk.5.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 157 |
+
INFO:hf-to-gguf:blk.5.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 158 |
+
INFO:hf-to-gguf:blk.6.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 159 |
+
INFO:hf-to-gguf:blk.6.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 160 |
+
INFO:hf-to-gguf:blk.6.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 161 |
+
INFO:hf-to-gguf:blk.6.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 162 |
+
INFO:hf-to-gguf:blk.6.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 163 |
+
INFO:hf-to-gguf:blk.6.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 164 |
+
INFO:hf-to-gguf:blk.6.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 165 |
+
INFO:hf-to-gguf:blk.6.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 166 |
+
INFO:hf-to-gguf:blk.6.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 167 |
+
INFO:hf-to-gguf:blk.7.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 168 |
+
INFO:hf-to-gguf:blk.7.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 169 |
+
INFO:hf-to-gguf:blk.7.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 170 |
+
INFO:hf-to-gguf:blk.7.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 171 |
+
INFO:hf-to-gguf:blk.7.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 172 |
+
INFO:hf-to-gguf:blk.7.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 173 |
+
INFO:hf-to-gguf:blk.7.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 174 |
+
INFO:hf-to-gguf:blk.7.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 175 |
+
INFO:hf-to-gguf:blk.7.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 176 |
+
INFO:hf-to-gguf:output_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 177 |
+
INFO:hf-to-gguf:Set meta model
|
| 178 |
+
INFO:hf-to-gguf:Set model parameters
|
| 179 |
+
INFO:hf-to-gguf:gguf: context length = 2048
|
| 180 |
+
INFO:hf-to-gguf:gguf: embedding length = 64
|
| 181 |
+
INFO:hf-to-gguf:gguf: feed forward length = 256
|
| 182 |
+
INFO:hf-to-gguf:gguf: head count = 16
|
| 183 |
+
INFO:hf-to-gguf:gguf: rms norm epsilon = 1e-06
|
| 184 |
+
INFO:hf-to-gguf:gguf: file type = 1
|
| 185 |
+
INFO:hf-to-gguf:Set model tokenizer
|
| 186 |
+
INFO:gguf.vocab:Setting special token type bos to 1
|
| 187 |
+
INFO:gguf.vocab:Setting special token type eos to 2
|
| 188 |
+
INFO:gguf.vocab:Setting special token type unk to 0
|
| 189 |
+
INFO:gguf.vocab:Setting special token type pad to 0
|
| 190 |
+
INFO:hf-to-gguf:Set model quantization version
|
| 191 |
+
INFO:gguf.gguf_writer:Writing the following files:
|
| 192 |
+
INFO:gguf.gguf_writer:maykeye_tinyllama/TinyLLama-4.6M-v0.0-F16.gguf: n_tensors = 75, total_size = 9.2M
|
| 193 |
+
|
| 194 |
+
INFO:hf-to-gguf:Model successfully exported to maykeye_tinyllama/TinyLLama-4.6M-v0.0-F16.gguf
|
| 195 |
+
INFO:hf-to-gguf:Loading model: maykeye_tinyllama
|
| 196 |
+
INFO:gguf.gguf_writer:gguf: This GGUF file is for Little Endian only
|
| 197 |
+
INFO:hf-to-gguf:Exporting model...
|
| 198 |
+
INFO:hf-to-gguf:gguf: loading model part 'model.safetensors'
|
| 199 |
+
INFO:hf-to-gguf:output.weight, torch.bfloat16 --> F16, shape = {64, 32000}
|
| 200 |
+
INFO:hf-to-gguf:token_embd.weight, torch.bfloat16 --> F16, shape = {64, 32000}
|
| 201 |
+
INFO:hf-to-gguf:blk.0.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 202 |
+
INFO:hf-to-gguf:blk.0.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 203 |
+
INFO:hf-to-gguf:blk.0.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 204 |
+
INFO:hf-to-gguf:blk.0.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 205 |
+
INFO:hf-to-gguf:blk.0.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 206 |
+
INFO:hf-to-gguf:blk.0.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 207 |
+
INFO:hf-to-gguf:blk.0.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 208 |
+
INFO:hf-to-gguf:blk.0.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 209 |
+
INFO:hf-to-gguf:blk.0.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 210 |
+
INFO:hf-to-gguf:blk.1.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 211 |
+
INFO:hf-to-gguf:blk.1.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 212 |
+
INFO:hf-to-gguf:blk.1.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 213 |
+
INFO:hf-to-gguf:blk.1.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 214 |
+
INFO:hf-to-gguf:blk.1.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 215 |
+
INFO:hf-to-gguf:blk.1.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 216 |
+
INFO:hf-to-gguf:blk.1.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 217 |
+
INFO:hf-to-gguf:blk.1.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 218 |
+
INFO:hf-to-gguf:blk.1.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 219 |
+
INFO:hf-to-gguf:blk.2.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 220 |
+
INFO:hf-to-gguf:blk.2.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 221 |
+
INFO:hf-to-gguf:blk.2.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 222 |
+
INFO:hf-to-gguf:blk.2.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 223 |
+
INFO:hf-to-gguf:blk.2.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 224 |
+
INFO:hf-to-gguf:blk.2.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 225 |
+
INFO:hf-to-gguf:blk.2.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 226 |
+
INFO:hf-to-gguf:blk.2.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 227 |
+
INFO:hf-to-gguf:blk.2.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 228 |
+
INFO:hf-to-gguf:blk.3.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 229 |
+
INFO:hf-to-gguf:blk.3.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 230 |
+
INFO:hf-to-gguf:blk.3.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 231 |
+
INFO:hf-to-gguf:blk.3.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 232 |
+
INFO:hf-to-gguf:blk.3.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 233 |
+
INFO:hf-to-gguf:blk.3.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 234 |
+
INFO:hf-to-gguf:blk.3.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 235 |
+
INFO:hf-to-gguf:blk.3.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 236 |
+
INFO:hf-to-gguf:blk.3.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 237 |
+
INFO:hf-to-gguf:blk.4.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 238 |
+
INFO:hf-to-gguf:blk.4.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 239 |
+
INFO:hf-to-gguf:blk.4.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 240 |
+
INFO:hf-to-gguf:blk.4.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 241 |
+
INFO:hf-to-gguf:blk.4.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 242 |
+
INFO:hf-to-gguf:blk.4.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 243 |
+
INFO:hf-to-gguf:blk.4.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 244 |
+
INFO:hf-to-gguf:blk.4.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 245 |
+
INFO:hf-to-gguf:blk.4.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 246 |
+
INFO:hf-to-gguf:blk.5.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 247 |
+
INFO:hf-to-gguf:blk.5.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 248 |
+
INFO:hf-to-gguf:blk.5.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 249 |
+
INFO:hf-to-gguf:blk.5.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 250 |
+
INFO:hf-to-gguf:blk.5.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 251 |
+
INFO:hf-to-gguf:blk.5.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 252 |
+
INFO:hf-to-gguf:blk.5.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 253 |
+
INFO:hf-to-gguf:blk.5.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 254 |
+
INFO:hf-to-gguf:blk.5.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 255 |
+
INFO:hf-to-gguf:blk.6.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 256 |
+
INFO:hf-to-gguf:blk.6.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 257 |
+
INFO:hf-to-gguf:blk.6.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 258 |
+
INFO:hf-to-gguf:blk.6.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 259 |
+
INFO:hf-to-gguf:blk.6.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 260 |
+
INFO:hf-to-gguf:blk.6.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 261 |
+
INFO:hf-to-gguf:blk.6.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 262 |
+
INFO:hf-to-gguf:blk.6.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 263 |
+
INFO:hf-to-gguf:blk.6.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 264 |
+
INFO:hf-to-gguf:blk.7.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 265 |
+
INFO:hf-to-gguf:blk.7.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 266 |
+
INFO:hf-to-gguf:blk.7.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 267 |
+
INFO:hf-to-gguf:blk.7.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 268 |
+
INFO:hf-to-gguf:blk.7.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 269 |
+
INFO:hf-to-gguf:blk.7.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 270 |
+
INFO:hf-to-gguf:blk.7.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 271 |
+
INFO:hf-to-gguf:blk.7.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 272 |
+
INFO:hf-to-gguf:blk.7.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 273 |
+
INFO:hf-to-gguf:output_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 274 |
+
INFO:hf-to-gguf:Set meta model
|
| 275 |
+
INFO:hf-to-gguf:Set model parameters
|
| 276 |
+
INFO:hf-to-gguf:gguf: context length = 2048
|
| 277 |
+
INFO:hf-to-gguf:gguf: embedding length = 64
|
| 278 |
+
INFO:hf-to-gguf:gguf: feed forward length = 256
|
| 279 |
+
INFO:hf-to-gguf:gguf: head count = 16
|
| 280 |
+
INFO:hf-to-gguf:gguf: rms norm epsilon = 1e-06
|
| 281 |
+
INFO:hf-to-gguf:gguf: file type = 1
|
| 282 |
+
INFO:hf-to-gguf:Set model tokenizer
|
| 283 |
+
INFO:gguf.vocab:Setting special token type bos to 1
|
| 284 |
+
INFO:gguf.vocab:Setting special token type eos to 2
|
| 285 |
+
INFO:gguf.vocab:Setting special token type unk to 0
|
| 286 |
+
INFO:gguf.vocab:Setting special token type pad to 0
|
| 287 |
+
INFO:hf-to-gguf:Set model quantization version
|
| 288 |
+
INFO:gguf.gguf_writer:Writing the following files:
|
| 289 |
+
INFO:gguf.gguf_writer:maykeye_tinyllama/TinyLLama-4.6M-v0.0-F16.gguf: n_tensors = 75, total_size = 9.2M
|
| 290 |
+
|
| 291 |
+
INFO:hf-to-gguf:Model successfully exported to maykeye_tinyllama/TinyLLama-4.6M-v0.0-F16.gguf
|
| 292 |
+
INFO:hf-to-gguf:Loading model: maykeye_tinyllama
|
| 293 |
+
INFO:gguf.gguf_writer:gguf: This GGUF file is for Little Endian only
|
| 294 |
+
INFO:hf-to-gguf:Exporting model...
|
| 295 |
+
INFO:hf-to-gguf:gguf: loading model part 'model.safetensors'
|
| 296 |
+
INFO:hf-to-gguf:output.weight, torch.bfloat16 --> F16, shape = {64, 32000}
|
| 297 |
+
INFO:hf-to-gguf:token_embd.weight, torch.bfloat16 --> F16, shape = {64, 32000}
|
| 298 |
+
INFO:hf-to-gguf:blk.0.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 299 |
+
INFO:hf-to-gguf:blk.0.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 300 |
+
INFO:hf-to-gguf:blk.0.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 301 |
+
INFO:hf-to-gguf:blk.0.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 302 |
+
INFO:hf-to-gguf:blk.0.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 303 |
+
INFO:hf-to-gguf:blk.0.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 304 |
+
INFO:hf-to-gguf:blk.0.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 305 |
+
INFO:hf-to-gguf:blk.0.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 306 |
+
INFO:hf-to-gguf:blk.0.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 307 |
+
INFO:hf-to-gguf:blk.1.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 308 |
+
INFO:hf-to-gguf:blk.1.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 309 |
+
INFO:hf-to-gguf:blk.1.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 310 |
+
INFO:hf-to-gguf:blk.1.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 311 |
+
INFO:hf-to-gguf:blk.1.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 312 |
+
INFO:hf-to-gguf:blk.1.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 313 |
+
INFO:hf-to-gguf:blk.1.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 314 |
+
INFO:hf-to-gguf:blk.1.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 315 |
+
INFO:hf-to-gguf:blk.1.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 316 |
+
INFO:hf-to-gguf:blk.2.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 317 |
+
INFO:hf-to-gguf:blk.2.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 318 |
+
INFO:hf-to-gguf:blk.2.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 319 |
+
INFO:hf-to-gguf:blk.2.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 320 |
+
INFO:hf-to-gguf:blk.2.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 321 |
+
INFO:hf-to-gguf:blk.2.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 322 |
+
INFO:hf-to-gguf:blk.2.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 323 |
+
INFO:hf-to-gguf:blk.2.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 324 |
+
INFO:hf-to-gguf:blk.2.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 325 |
+
INFO:hf-to-gguf:blk.3.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 326 |
+
INFO:hf-to-gguf:blk.3.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 327 |
+
INFO:hf-to-gguf:blk.3.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 328 |
+
INFO:hf-to-gguf:blk.3.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 329 |
+
INFO:hf-to-gguf:blk.3.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 330 |
+
INFO:hf-to-gguf:blk.3.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 331 |
+
INFO:hf-to-gguf:blk.3.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 332 |
+
INFO:hf-to-gguf:blk.3.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 333 |
+
INFO:hf-to-gguf:blk.3.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 334 |
+
INFO:hf-to-gguf:blk.4.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 335 |
+
INFO:hf-to-gguf:blk.4.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 336 |
+
INFO:hf-to-gguf:blk.4.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 337 |
+
INFO:hf-to-gguf:blk.4.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 338 |
+
INFO:hf-to-gguf:blk.4.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 339 |
+
INFO:hf-to-gguf:blk.4.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 340 |
+
INFO:hf-to-gguf:blk.4.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 341 |
+
INFO:hf-to-gguf:blk.4.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 342 |
+
INFO:hf-to-gguf:blk.4.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 343 |
+
INFO:hf-to-gguf:blk.5.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 344 |
+
INFO:hf-to-gguf:blk.5.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 345 |
+
INFO:hf-to-gguf:blk.5.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 346 |
+
INFO:hf-to-gguf:blk.5.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 347 |
+
INFO:hf-to-gguf:blk.5.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 348 |
+
INFO:hf-to-gguf:blk.5.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 349 |
+
INFO:hf-to-gguf:blk.5.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 350 |
+
INFO:hf-to-gguf:blk.5.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 351 |
+
INFO:hf-to-gguf:blk.5.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 352 |
+
INFO:hf-to-gguf:blk.6.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 353 |
+
INFO:hf-to-gguf:blk.6.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 354 |
+
INFO:hf-to-gguf:blk.6.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 355 |
+
INFO:hf-to-gguf:blk.6.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 356 |
+
INFO:hf-to-gguf:blk.6.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 357 |
+
INFO:hf-to-gguf:blk.6.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 358 |
+
INFO:hf-to-gguf:blk.6.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 359 |
+
INFO:hf-to-gguf:blk.6.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 360 |
+
INFO:hf-to-gguf:blk.6.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 361 |
+
INFO:hf-to-gguf:blk.7.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 362 |
+
INFO:hf-to-gguf:blk.7.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 363 |
+
INFO:hf-to-gguf:blk.7.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 364 |
+
INFO:hf-to-gguf:blk.7.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 365 |
+
INFO:hf-to-gguf:blk.7.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 366 |
+
INFO:hf-to-gguf:blk.7.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 367 |
+
INFO:hf-to-gguf:blk.7.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 368 |
+
INFO:hf-to-gguf:blk.7.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 369 |
+
INFO:hf-to-gguf:blk.7.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 370 |
+
INFO:hf-to-gguf:output_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 371 |
+
INFO:hf-to-gguf:Set meta model
|
| 372 |
+
INFO:hf-to-gguf:Set model parameters
|
| 373 |
+
INFO:hf-to-gguf:gguf: context length = 2048
|
| 374 |
+
INFO:hf-to-gguf:gguf: embedding length = 64
|
| 375 |
+
INFO:hf-to-gguf:gguf: feed forward length = 256
|
| 376 |
+
INFO:hf-to-gguf:gguf: head count = 16
|
| 377 |
+
INFO:hf-to-gguf:gguf: rms norm epsilon = 1e-06
|
| 378 |
+
INFO:hf-to-gguf:gguf: file type = 1
|
| 379 |
+
INFO:hf-to-gguf:Set model tokenizer
|
| 380 |
+
INFO:gguf.vocab:Setting special token type bos to 1
|
| 381 |
+
INFO:gguf.vocab:Setting special token type eos to 2
|
| 382 |
+
INFO:gguf.vocab:Setting special token type unk to 0
|
| 383 |
+
INFO:gguf.vocab:Setting special token type pad to 0
|
| 384 |
+
INFO:hf-to-gguf:Set model quantization version
|
| 385 |
+
INFO:gguf.gguf_writer:Writing the following files:
|
| 386 |
+
INFO:gguf.gguf_writer:maykeye_tinyllama/TinyLLama-4.6M-v0.0-F16.gguf: n_tensors = 75, total_size = 9.2M
|
| 387 |
+
|
| 388 |
+
INFO:hf-to-gguf:Model successfully exported to maykeye_tinyllama/TinyLLama-4.6M-v0.0-F16.gguf
|
| 389 |
+
INFO:hf-to-gguf:Loading model: maykeye_tinyllama
|
| 390 |
+
INFO:gguf.gguf_writer:gguf: This GGUF file is for Little Endian only
|
| 391 |
+
INFO:hf-to-gguf:Exporting model...
|
| 392 |
+
INFO:hf-to-gguf:gguf: loading model part 'model.safetensors'
|
| 393 |
+
INFO:hf-to-gguf:output.weight, torch.bfloat16 --> F16, shape = {64, 32000}
|
| 394 |
+
INFO:hf-to-gguf:token_embd.weight, torch.bfloat16 --> F16, shape = {64, 32000}
|
| 395 |
+
INFO:hf-to-gguf:blk.0.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 396 |
+
INFO:hf-to-gguf:blk.0.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 397 |
+
INFO:hf-to-gguf:blk.0.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 398 |
+
INFO:hf-to-gguf:blk.0.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 399 |
+
INFO:hf-to-gguf:blk.0.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 400 |
+
INFO:hf-to-gguf:blk.0.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 401 |
+
INFO:hf-to-gguf:blk.0.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 402 |
+
INFO:hf-to-gguf:blk.0.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 403 |
+
INFO:hf-to-gguf:blk.0.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 404 |
+
INFO:hf-to-gguf:blk.1.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 405 |
+
INFO:hf-to-gguf:blk.1.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 406 |
+
INFO:hf-to-gguf:blk.1.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 407 |
+
INFO:hf-to-gguf:blk.1.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 408 |
+
INFO:hf-to-gguf:blk.1.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 409 |
+
INFO:hf-to-gguf:blk.1.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 410 |
+
INFO:hf-to-gguf:blk.1.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 411 |
+
INFO:hf-to-gguf:blk.1.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 412 |
+
INFO:hf-to-gguf:blk.1.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 413 |
+
INFO:hf-to-gguf:blk.2.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 414 |
+
INFO:hf-to-gguf:blk.2.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 415 |
+
INFO:hf-to-gguf:blk.2.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 416 |
+
INFO:hf-to-gguf:blk.2.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 417 |
+
INFO:hf-to-gguf:blk.2.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 418 |
+
INFO:hf-to-gguf:blk.2.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 419 |
+
INFO:hf-to-gguf:blk.2.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 420 |
+
INFO:hf-to-gguf:blk.2.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 421 |
+
INFO:hf-to-gguf:blk.2.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 422 |
+
INFO:hf-to-gguf:blk.3.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 423 |
+
INFO:hf-to-gguf:blk.3.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 424 |
+
INFO:hf-to-gguf:blk.3.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 425 |
+
INFO:hf-to-gguf:blk.3.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 426 |
+
INFO:hf-to-gguf:blk.3.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 427 |
+
INFO:hf-to-gguf:blk.3.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 428 |
+
INFO:hf-to-gguf:blk.3.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 429 |
+
INFO:hf-to-gguf:blk.3.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 430 |
+
INFO:hf-to-gguf:blk.3.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 431 |
+
INFO:hf-to-gguf:blk.4.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 432 |
+
INFO:hf-to-gguf:blk.4.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 433 |
+
INFO:hf-to-gguf:blk.4.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 434 |
+
INFO:hf-to-gguf:blk.4.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 435 |
+
INFO:hf-to-gguf:blk.4.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 436 |
+
INFO:hf-to-gguf:blk.4.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 437 |
+
INFO:hf-to-gguf:blk.4.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 438 |
+
INFO:hf-to-gguf:blk.4.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 439 |
+
INFO:hf-to-gguf:blk.4.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 440 |
+
INFO:hf-to-gguf:blk.5.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 441 |
+
INFO:hf-to-gguf:blk.5.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 442 |
+
INFO:hf-to-gguf:blk.5.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 443 |
+
INFO:hf-to-gguf:blk.5.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 444 |
+
INFO:hf-to-gguf:blk.5.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 445 |
+
INFO:hf-to-gguf:blk.5.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 446 |
+
INFO:hf-to-gguf:blk.5.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 447 |
+
INFO:hf-to-gguf:blk.5.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 448 |
+
INFO:hf-to-gguf:blk.5.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 449 |
+
INFO:hf-to-gguf:blk.6.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 450 |
+
INFO:hf-to-gguf:blk.6.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 451 |
+
INFO:hf-to-gguf:blk.6.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 452 |
+
INFO:hf-to-gguf:blk.6.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 453 |
+
INFO:hf-to-gguf:blk.6.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 454 |
+
INFO:hf-to-gguf:blk.6.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 455 |
+
INFO:hf-to-gguf:blk.6.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 456 |
+
INFO:hf-to-gguf:blk.6.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 457 |
+
INFO:hf-to-gguf:blk.6.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 458 |
+
INFO:hf-to-gguf:blk.7.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 459 |
+
INFO:hf-to-gguf:blk.7.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 460 |
+
INFO:hf-to-gguf:blk.7.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 461 |
+
INFO:hf-to-gguf:blk.7.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 462 |
+
INFO:hf-to-gguf:blk.7.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 463 |
+
INFO:hf-to-gguf:blk.7.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 464 |
+
INFO:hf-to-gguf:blk.7.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 465 |
+
INFO:hf-to-gguf:blk.7.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 466 |
+
INFO:hf-to-gguf:blk.7.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 467 |
+
INFO:hf-to-gguf:output_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 468 |
+
INFO:hf-to-gguf:Set meta model
|
| 469 |
+
INFO:hf-to-gguf:Set model parameters
|
| 470 |
+
INFO:hf-to-gguf:gguf: context length = 2048
|
| 471 |
+
INFO:hf-to-gguf:gguf: embedding length = 64
|
| 472 |
+
INFO:hf-to-gguf:gguf: feed forward length = 256
|
| 473 |
+
INFO:hf-to-gguf:gguf: head count = 16
|
| 474 |
+
INFO:hf-to-gguf:gguf: rms norm epsilon = 1e-06
|
| 475 |
+
INFO:hf-to-gguf:gguf: file type = 1
|
| 476 |
+
INFO:hf-to-gguf:Set model tokenizer
|
| 477 |
+
INFO:gguf.vocab:Setting special token type bos to 1
|
| 478 |
+
INFO:gguf.vocab:Setting special token type eos to 2
|
| 479 |
+
INFO:gguf.vocab:Setting special token type unk to 0
|
| 480 |
+
INFO:gguf.vocab:Setting special token type pad to 0
|
| 481 |
+
INFO:hf-to-gguf:Set model quantization version
|
| 482 |
+
INFO:gguf.gguf_writer:Writing the following files:
|
| 483 |
+
INFO:gguf.gguf_writer:maykeye_tinyllama/TinyLLama-4.6M-v0.0-F16.gguf: n_tensors = 75, total_size = 9.2M
|
| 484 |
+
|
| 485 |
+
INFO:hf-to-gguf:Model successfully exported to maykeye_tinyllama/TinyLLama-4.6M-v0.0-F16.gguf
|
| 486 |
+
INFO:hf-to-gguf:Loading model: maykeye_tinyllama
|
| 487 |
+
INFO:gguf.gguf_writer:gguf: This GGUF file is for Little Endian only
|
| 488 |
+
INFO:hf-to-gguf:Exporting model...
|
| 489 |
+
INFO:hf-to-gguf:gguf: loading model part 'model.safetensors'
|
| 490 |
+
INFO:hf-to-gguf:output.weight, torch.bfloat16 --> F16, shape = {64, 32000}
|
| 491 |
+
INFO:hf-to-gguf:token_embd.weight, torch.bfloat16 --> F16, shape = {64, 32000}
|
| 492 |
+
INFO:hf-to-gguf:blk.0.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 493 |
+
INFO:hf-to-gguf:blk.0.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 494 |
+
INFO:hf-to-gguf:blk.0.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 495 |
+
INFO:hf-to-gguf:blk.0.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 496 |
+
INFO:hf-to-gguf:blk.0.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 497 |
+
INFO:hf-to-gguf:blk.0.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 498 |
+
INFO:hf-to-gguf:blk.0.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 499 |
+
INFO:hf-to-gguf:blk.0.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 500 |
+
INFO:hf-to-gguf:blk.0.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 501 |
+
INFO:hf-to-gguf:blk.1.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 502 |
+
INFO:hf-to-gguf:blk.1.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 503 |
+
INFO:hf-to-gguf:blk.1.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 504 |
+
INFO:hf-to-gguf:blk.1.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 505 |
+
INFO:hf-to-gguf:blk.1.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 506 |
+
INFO:hf-to-gguf:blk.1.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 507 |
+
INFO:hf-to-gguf:blk.1.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 508 |
+
INFO:hf-to-gguf:blk.1.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 509 |
+
INFO:hf-to-gguf:blk.1.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 510 |
+
INFO:hf-to-gguf:blk.2.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 511 |
+
INFO:hf-to-gguf:blk.2.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 512 |
+
INFO:hf-to-gguf:blk.2.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 513 |
+
INFO:hf-to-gguf:blk.2.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 514 |
+
INFO:hf-to-gguf:blk.2.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 515 |
+
INFO:hf-to-gguf:blk.2.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 516 |
+
INFO:hf-to-gguf:blk.2.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 517 |
+
INFO:hf-to-gguf:blk.2.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 518 |
+
INFO:hf-to-gguf:blk.2.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 519 |
+
INFO:hf-to-gguf:blk.3.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 520 |
+
INFO:hf-to-gguf:blk.3.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 521 |
+
INFO:hf-to-gguf:blk.3.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 522 |
+
INFO:hf-to-gguf:blk.3.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 523 |
+
INFO:hf-to-gguf:blk.3.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 524 |
+
INFO:hf-to-gguf:blk.3.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 525 |
+
INFO:hf-to-gguf:blk.3.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 526 |
+
INFO:hf-to-gguf:blk.3.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 527 |
+
INFO:hf-to-gguf:blk.3.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 528 |
+
INFO:hf-to-gguf:blk.4.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 529 |
+
INFO:hf-to-gguf:blk.4.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 530 |
+
INFO:hf-to-gguf:blk.4.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 531 |
+
INFO:hf-to-gguf:blk.4.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 532 |
+
INFO:hf-to-gguf:blk.4.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 533 |
+
INFO:hf-to-gguf:blk.4.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 534 |
+
INFO:hf-to-gguf:blk.4.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 535 |
+
INFO:hf-to-gguf:blk.4.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 536 |
+
INFO:hf-to-gguf:blk.4.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 537 |
+
INFO:hf-to-gguf:blk.5.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 538 |
+
INFO:hf-to-gguf:blk.5.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 539 |
+
INFO:hf-to-gguf:blk.5.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 540 |
+
INFO:hf-to-gguf:blk.5.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 541 |
+
INFO:hf-to-gguf:blk.5.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 542 |
+
INFO:hf-to-gguf:blk.5.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 543 |
+
INFO:hf-to-gguf:blk.5.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 544 |
+
INFO:hf-to-gguf:blk.5.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 545 |
+
INFO:hf-to-gguf:blk.5.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 546 |
+
INFO:hf-to-gguf:blk.6.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 547 |
+
INFO:hf-to-gguf:blk.6.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 548 |
+
INFO:hf-to-gguf:blk.6.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 549 |
+
INFO:hf-to-gguf:blk.6.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 550 |
+
INFO:hf-to-gguf:blk.6.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 551 |
+
INFO:hf-to-gguf:blk.6.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 552 |
+
INFO:hf-to-gguf:blk.6.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 553 |
+
INFO:hf-to-gguf:blk.6.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 554 |
+
INFO:hf-to-gguf:blk.6.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 555 |
+
INFO:hf-to-gguf:blk.7.attn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 556 |
+
INFO:hf-to-gguf:blk.7.ffn_down.weight, torch.bfloat16 --> F16, shape = {256, 64}
|
| 557 |
+
INFO:hf-to-gguf:blk.7.ffn_gate.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 558 |
+
INFO:hf-to-gguf:blk.7.ffn_up.weight, torch.bfloat16 --> F16, shape = {64, 256}
|
| 559 |
+
INFO:hf-to-gguf:blk.7.ffn_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 560 |
+
INFO:hf-to-gguf:blk.7.attn_k.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 561 |
+
INFO:hf-to-gguf:blk.7.attn_output.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 562 |
+
INFO:hf-to-gguf:blk.7.attn_q.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 563 |
+
INFO:hf-to-gguf:blk.7.attn_v.weight, torch.bfloat16 --> F16, shape = {64, 64}
|
| 564 |
+
INFO:hf-to-gguf:output_norm.weight, torch.bfloat16 --> F32, shape = {64}
|
| 565 |
+
INFO:hf-to-gguf:Set meta model
|
| 566 |
+
INFO:hf-to-gguf:Set model parameters
|
| 567 |
+
INFO:hf-to-gguf:gguf: context length = 2048
|
| 568 |
+
INFO:hf-to-gguf:gguf: embedding length = 64
|
| 569 |
+
INFO:hf-to-gguf:gguf: feed forward length = 256
|
| 570 |
+
INFO:hf-to-gguf:gguf: head count = 16
|
| 571 |
+
INFO:hf-to-gguf:gguf: rms norm epsilon = 1e-06
|
| 572 |
+
INFO:hf-to-gguf:gguf: file type = 1
|
| 573 |
+
INFO:hf-to-gguf:Set model tokenizer
|
| 574 |
+
INFO:gguf.vocab:Setting special token type bos to 1
|
| 575 |
+
INFO:gguf.vocab:Setting special token type eos to 2
|
| 576 |
+
INFO:gguf.vocab:Setting special token type unk to 0
|
| 577 |
+
INFO:gguf.vocab:Setting special token type pad to 0
|
| 578 |
+
INFO:hf-to-gguf:Set model quantization version
|
| 579 |
+
INFO:gguf.gguf_writer:Writing the following files:
|
| 580 |
+
INFO:gguf.gguf_writer:maykeye_tinyllama/TinyLLama-4.6M-v0.0-F16.gguf: n_tensors = 75, total_size = 9.2M
|
| 581 |
+
|
| 582 |
+
INFO:hf-to-gguf:Model successfully exported to maykeye_tinyllama/TinyLLama-4.6M-v0.0-F16.gguf
|
llamafile-creation.sh
CHANGED
|
@@ -34,7 +34,7 @@ echo We will be aiming to generate $OUTFILE.llamafile
|
|
| 34 |
|
| 35 |
###############################################################################
|
| 36 |
echo == Convert from safetensor to gguf ==
|
| 37 |
-
./llama.cpp/convert_hf_to_gguf.py ${MODEL_DIR} --metadata ${METADATA_FILE} --outtype f16 --verbose
|
| 38 |
mv ${MODEL_DIR}/${OUTFILE}.gguf ${OUTFILE}.gguf
|
| 39 |
|
| 40 |
# Generate Diagnostics Dumpfile
|
|
@@ -55,4 +55,4 @@ EOF
|
|
| 55 |
|
| 56 |
###############################################################################
|
| 57 |
echo == Test Output ./${OUTFILE}.llamafile ==
|
| 58 |
-
./${OUTFILE}.llamafile --cli -p "hello world the gruff man said"
|
|
|
|
| 34 |
|
| 35 |
###############################################################################
|
| 36 |
echo == Convert from safetensor to gguf ==
|
| 37 |
+
./llama.cpp/convert_hf_to_gguf.py ${MODEL_DIR} --metadata ${METADATA_FILE} --outtype f16 --verbose &>> convert_hf_to_gguf.output.txt
|
| 38 |
mv ${MODEL_DIR}/${OUTFILE}.gguf ${OUTFILE}.gguf
|
| 39 |
|
| 40 |
# Generate Diagnostics Dumpfile
|
|
|
|
| 55 |
|
| 56 |
###############################################################################
|
| 57 |
echo == Test Output ./${OUTFILE}.llamafile ==
|
| 58 |
+
./${OUTFILE}.llamafile --cli -p "hello world the gruff man said" &>> llamafile_output_example.output.txt
|
llamafile_output_example.output.txt
ADDED
|
@@ -0,0 +1,806 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
note: if you have an AMD or NVIDIA GPU then you need to pass -ngl 9999 to enable GPU offloading
|
| 2 |
+
main: llamafile version 0.8.9
|
| 3 |
+
main: seed = 1721530767
|
| 4 |
+
llama_model_loader: loaded meta data with 37 key-value pairs and 75 tensors from TinyLLama-4.6M-v0.0-F16.gguf (version GGUF V3 (latest))
|
| 5 |
+
llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
|
| 6 |
+
llama_model_loader: - kv 0: general.architecture str = llama
|
| 7 |
+
llama_model_loader: - kv 1: general.type str = model
|
| 8 |
+
llama_model_loader: - kv 2: general.name str = TinyLLama
|
| 9 |
+
llama_model_loader: - kv 3: general.author str = Maykeye
|
| 10 |
+
llama_model_loader: - kv 4: general.version str = v0.0
|
| 11 |
+
llama_model_loader: - kv 5: general.description str = This gguf is ported from a first vers...
|
| 12 |
+
llama_model_loader: - kv 6: general.quantized_by str = Mofosyne
|
| 13 |
+
llama_model_loader: - kv 7: general.size_label str = 4.6M
|
| 14 |
+
llama_model_loader: - kv 8: general.license str = apache-2.0
|
| 15 |
+
llama_model_loader: - kv 9: general.license.name str = Apache License Version 2.0, January 2004
|
| 16 |
+
llama_model_loader: - kv 10: general.license.link str = https://huggingface.co/datasets/choos...
|
| 17 |
+
llama_model_loader: - kv 11: general.url str = https://huggingface.co/mofosyne/TinyL...
|
| 18 |
+
llama_model_loader: - kv 12: general.repo_url str = https://huggingface.co/mofosyne/TinyL...
|
| 19 |
+
llama_model_loader: - kv 13: general.source.url str = https://huggingface.co/Maykeye/TinyLL...
|
| 20 |
+
llama_model_loader: - kv 14: general.source.repo_url str = https://huggingface.co/Maykeye/TinyLL...
|
| 21 |
+
llama_model_loader: - kv 15: general.tags arr[str,5] = ["text generation", "transformer", "l...
|
| 22 |
+
llama_model_loader: - kv 16: general.languages arr[str,1] = ["en"]
|
| 23 |
+
llama_model_loader: - kv 17: general.datasets arr[str,2] = ["https://huggingface.co/datasets/ron...
|
| 24 |
+
llama_model_loader: - kv 18: llama.block_count u32 = 8
|
| 25 |
+
llama_model_loader: - kv 19: llama.context_length u32 = 2048
|
| 26 |
+
llama_model_loader: - kv 20: llama.embedding_length u32 = 64
|
| 27 |
+
llama_model_loader: - kv 21: llama.feed_forward_length u32 = 256
|
| 28 |
+
llama_model_loader: - kv 22: llama.attention.head_count u32 = 16
|
| 29 |
+
llama_model_loader: - kv 23: llama.attention.layer_norm_rms_epsilon f32 = 0.000001
|
| 30 |
+
llama_model_loader: - kv 24: general.file_type u32 = 1
|
| 31 |
+
llama_model_loader: - kv 25: llama.vocab_size u32 = 32000
|
| 32 |
+
llama_model_loader: - kv 26: llama.rope.dimension_count u32 = 4
|
| 33 |
+
llama_model_loader: - kv 27: tokenizer.ggml.model str = llama
|
| 34 |
+
llama_model_loader: - kv 28: tokenizer.ggml.pre str = default
|
| 35 |
+
llama_model_loader: - kv 29: tokenizer.ggml.tokens arr[str,32000] = ["<unk>", "<s>", "</s>", "<0x00>", "<...
|
| 36 |
+
llama_model_loader: - kv 30: tokenizer.ggml.scores arr[f32,32000] = [0.000000, 0.000000, 0.000000, 0.0000...
|
| 37 |
+
llama_model_loader: - kv 31: tokenizer.ggml.token_type arr[i32,32000] = [2, 3, 3, 6, 6, 6, 6, 6, 6, 6, 6, 6, ...
|
| 38 |
+
llama_model_loader: - kv 32: tokenizer.ggml.bos_token_id u32 = 1
|
| 39 |
+
llama_model_loader: - kv 33: tokenizer.ggml.eos_token_id u32 = 2
|
| 40 |
+
llama_model_loader: - kv 34: tokenizer.ggml.unknown_token_id u32 = 0
|
| 41 |
+
llama_model_loader: - kv 35: tokenizer.ggml.padding_token_id u32 = 0
|
| 42 |
+
llama_model_loader: - kv 36: general.quantization_version u32 = 2
|
| 43 |
+
llama_model_loader: - type f32: 17 tensors
|
| 44 |
+
llama_model_loader: - type f16: 58 tensors
|
| 45 |
+
llm_load_vocab: special tokens definition check successful ( 259/32000 ).
|
| 46 |
+
llm_load_print_meta: format = GGUF V3 (latest)
|
| 47 |
+
llm_load_print_meta: arch = llama
|
| 48 |
+
llm_load_print_meta: vocab type = SPM
|
| 49 |
+
llm_load_print_meta: n_vocab = 32000
|
| 50 |
+
llm_load_print_meta: n_merges = 0
|
| 51 |
+
llm_load_print_meta: n_ctx_train = 2048
|
| 52 |
+
llm_load_print_meta: n_embd = 64
|
| 53 |
+
llm_load_print_meta: n_head = 16
|
| 54 |
+
llm_load_print_meta: n_head_kv = 16
|
| 55 |
+
llm_load_print_meta: n_layer = 8
|
| 56 |
+
llm_load_print_meta: n_rot = 4
|
| 57 |
+
llm_load_print_meta: n_swa = 0
|
| 58 |
+
llm_load_print_meta: n_embd_head_k = 4
|
| 59 |
+
llm_load_print_meta: n_embd_head_v = 4
|
| 60 |
+
llm_load_print_meta: n_gqa = 1
|
| 61 |
+
llm_load_print_meta: n_embd_k_gqa = 64
|
| 62 |
+
llm_load_print_meta: n_embd_v_gqa = 64
|
| 63 |
+
llm_load_print_meta: f_norm_eps = 0.0e+00
|
| 64 |
+
llm_load_print_meta: f_norm_rms_eps = 1.0e-06
|
| 65 |
+
llm_load_print_meta: f_clamp_kqv = 0.0e+00
|
| 66 |
+
llm_load_print_meta: f_max_alibi_bias = 0.0e+00
|
| 67 |
+
llm_load_print_meta: f_logit_scale = 0.0e+00
|
| 68 |
+
llm_load_print_meta: n_ff = 256
|
| 69 |
+
llm_load_print_meta: n_expert = 0
|
| 70 |
+
llm_load_print_meta: n_expert_used = 0
|
| 71 |
+
llm_load_print_meta: causal attn = 1
|
| 72 |
+
llm_load_print_meta: pooling type = 0
|
| 73 |
+
llm_load_print_meta: rope type = 0
|
| 74 |
+
llm_load_print_meta: rope scaling = linear
|
| 75 |
+
llm_load_print_meta: freq_base_train = 10000.0
|
| 76 |
+
llm_load_print_meta: freq_scale_train = 1
|
| 77 |
+
llm_load_print_meta: n_yarn_orig_ctx = 2048
|
| 78 |
+
llm_load_print_meta: rope_finetuned = unknown
|
| 79 |
+
llm_load_print_meta: ssm_d_conv = 0
|
| 80 |
+
llm_load_print_meta: ssm_d_inner = 0
|
| 81 |
+
llm_load_print_meta: ssm_d_state = 0
|
| 82 |
+
llm_load_print_meta: ssm_dt_rank = 0
|
| 83 |
+
llm_load_print_meta: model type = ?B
|
| 84 |
+
llm_load_print_meta: model ftype = F16
|
| 85 |
+
llm_load_print_meta: model params = 4.62 M
|
| 86 |
+
llm_load_print_meta: model size = 8.82 MiB (16.00 BPW)
|
| 87 |
+
llm_load_print_meta: general.name = TinyLLama
|
| 88 |
+
llm_load_print_meta: BOS token = 1 '<s>'
|
| 89 |
+
llm_load_print_meta: EOS token = 2 '</s>'
|
| 90 |
+
llm_load_print_meta: UNK token = 0 '<unk>'
|
| 91 |
+
llm_load_print_meta: PAD token = 0 '<unk>'
|
| 92 |
+
llm_load_print_meta: LF token = 13 '<0x0A>'
|
| 93 |
+
llm_load_tensors: ggml ctx size = 0.04 MiB
|
| 94 |
+
llm_load_tensors: CPU buffer size = 8.82 MiB
|
| 95 |
+
..............
|
| 96 |
+
llama_new_context_with_model: n_ctx = 512
|
| 97 |
+
llama_new_context_with_model: n_batch = 512
|
| 98 |
+
llama_new_context_with_model: n_ubatch = 512
|
| 99 |
+
llama_new_context_with_model: flash_attn = 0
|
| 100 |
+
llama_new_context_with_model: freq_base = 10000.0
|
| 101 |
+
llama_new_context_with_model: freq_scale = 1
|
| 102 |
+
llama_kv_cache_init: CPU KV buffer size = 1.00 MiB
|
| 103 |
+
llama_new_context_with_model: KV self size = 1.00 MiB, K (f16): 0.50 MiB, V (f16): 0.50 MiB
|
| 104 |
+
llama_new_context_with_model: CPU output buffer size = 0.12 MiB
|
| 105 |
+
llama_new_context_with_model: CPU compute buffer size = 62.75 MiB
|
| 106 |
+
llama_new_context_with_model: graph nodes = 262
|
| 107 |
+
llama_new_context_with_model: graph splits = 1
|
| 108 |
+
|
| 109 |
+
system_info: n_threads = 4 / 8 | AVX = 1 | AVX_VNNI = 0 | AVX2 = 1 | AVX512 = 0 | AVX512_VBMI = 0 | AVX512_VNNI = 0 | AVX512_BF16 = 0 | FMA = 1 | NEON = 0 | ARM_FMA = 0 | F16C = 1 | FP16_VA = 0 | WASM_SIMD = 0 | BLAS = 0 | SSE3 = 1 | SSSE3 = 1 | VSX = 0 | MATMUL_INT8 = 0 | LLAMAFILE = 1 |
|
| 110 |
+
sampling:
|
| 111 |
+
repeat_last_n = 64, repeat_penalty = 1.000, frequency_penalty = 0.000, presence_penalty = 0.000
|
| 112 |
+
top_k = 40, tfs_z = 1.000, top_p = 0.950, min_p = 0.050, typical_p = 1.000, temp = 0.800
|
| 113 |
+
mirostat = 0, mirostat_lr = 0.100, mirostat_ent = 5.000
|
| 114 |
+
sampling order:
|
| 115 |
+
CFG -> Penalties -> top_k -> tfs_z -> typical_p -> top_p -> min_p -> temperature
|
| 116 |
+
generate: n_ctx = 512, n_batch = 2048, n_predict = -1, n_keep = 1
|
| 117 |
+
|
| 118 |
+
|
| 119 |
+
hello world the gruff man said they had a big heart. The man smiled and said he would be the best friends.
|
| 120 |
+
The man said no. He said he didn't want to take a nap and he was too sad.
|
| 121 |
+
The man was so sad he started to cry. He didn't know what the man was one of his friends.
|
| 122 |
+
The man saw that the man was not the children. He felt sad and said he wanted to come back. The man was mad and he was so sad.
|
| 123 |
+
The man felt bad for the people, and they both felt bad for the story.
|
| 124 |
+
The man told the kids about the end. He said he had to stay too slow and not take them. He was sad, but it was too late.
|
| 125 |
+
The people were very sad, but they knew it was not nice. They were never seen again. [end of text]
|
| 126 |
+
|
| 127 |
+
|
| 128 |
+
llama_print_timings: load time = 10.61 ms
|
| 129 |
+
llama_print_timings: sample time = 5.92 ms / 172 runs ( 0.03 ms per token, 29073.70 tokens per second)
|
| 130 |
+
llama_print_timings: prompt eval time = 2.03 ms / 8 tokens ( 0.25 ms per token, 3942.83 tokens per second)
|
| 131 |
+
llama_print_timings: eval time = 245.61 ms / 171 runs ( 1.44 ms per token, 696.21 tokens per second)
|
| 132 |
+
llama_print_timings: total time = 292.61 ms / 179 tokens
|
| 133 |
+
Log end
|
| 134 |
+
note: if you have an AMD or NVIDIA GPU then you need to pass -ngl 9999 to enable GPU offloading
|
| 135 |
+
main: llamafile version 0.8.9
|
| 136 |
+
main: seed = 1721531043
|
| 137 |
+
llama_model_loader: loaded meta data with 37 key-value pairs and 75 tensors from TinyLLama-4.6M-v0.0-F16.gguf (version GGUF V3 (latest))
|
| 138 |
+
llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
|
| 139 |
+
llama_model_loader: - kv 0: general.architecture str = llama
|
| 140 |
+
llama_model_loader: - kv 1: general.type str = model
|
| 141 |
+
llama_model_loader: - kv 2: general.name str = TinyLLama
|
| 142 |
+
llama_model_loader: - kv 3: general.author str = Maykeye
|
| 143 |
+
llama_model_loader: - kv 4: general.version str = v0.0
|
| 144 |
+
llama_model_loader: - kv 5: general.description str = This gguf is ported from a first vers...
|
| 145 |
+
llama_model_loader: - kv 6: general.quantized_by str = Mofosyne
|
| 146 |
+
llama_model_loader: - kv 7: general.size_label str = 4.6M
|
| 147 |
+
llama_model_loader: - kv 8: general.license str = apache-2.0
|
| 148 |
+
llama_model_loader: - kv 9: general.license.name str = Apache License Version 2.0, January 2004
|
| 149 |
+
llama_model_loader: - kv 10: general.license.link str = https://huggingface.co/datasets/choos...
|
| 150 |
+
llama_model_loader: - kv 11: general.url str = https://huggingface.co/mofosyne/TinyL...
|
| 151 |
+
llama_model_loader: - kv 12: general.repo_url str = https://huggingface.co/mofosyne/TinyL...
|
| 152 |
+
llama_model_loader: - kv 13: general.source.url str = https://huggingface.co/Maykeye/TinyLL...
|
| 153 |
+
llama_model_loader: - kv 14: general.source.repo_url str = https://huggingface.co/Maykeye/TinyLL...
|
| 154 |
+
llama_model_loader: - kv 15: general.tags arr[str,5] = ["text generation", "transformer", "l...
|
| 155 |
+
llama_model_loader: - kv 16: general.languages arr[str,1] = ["en"]
|
| 156 |
+
llama_model_loader: - kv 17: general.datasets arr[str,2] = ["https://huggingface.co/datasets/ron...
|
| 157 |
+
llama_model_loader: - kv 18: llama.block_count u32 = 8
|
| 158 |
+
llama_model_loader: - kv 19: llama.context_length u32 = 2048
|
| 159 |
+
llama_model_loader: - kv 20: llama.embedding_length u32 = 64
|
| 160 |
+
llama_model_loader: - kv 21: llama.feed_forward_length u32 = 256
|
| 161 |
+
llama_model_loader: - kv 22: llama.attention.head_count u32 = 16
|
| 162 |
+
llama_model_loader: - kv 23: llama.attention.layer_norm_rms_epsilon f32 = 0.000001
|
| 163 |
+
llama_model_loader: - kv 24: general.file_type u32 = 1
|
| 164 |
+
llama_model_loader: - kv 25: llama.vocab_size u32 = 32000
|
| 165 |
+
llama_model_loader: - kv 26: llama.rope.dimension_count u32 = 4
|
| 166 |
+
llama_model_loader: - kv 27: tokenizer.ggml.model str = llama
|
| 167 |
+
llama_model_loader: - kv 28: tokenizer.ggml.pre str = default
|
| 168 |
+
llama_model_loader: - kv 29: tokenizer.ggml.tokens arr[str,32000] = ["<unk>", "<s>", "</s>", "<0x00>", "<...
|
| 169 |
+
llama_model_loader: - kv 30: tokenizer.ggml.scores arr[f32,32000] = [0.000000, 0.000000, 0.000000, 0.0000...
|
| 170 |
+
llama_model_loader: - kv 31: tokenizer.ggml.token_type arr[i32,32000] = [2, 3, 3, 6, 6, 6, 6, 6, 6, 6, 6, 6, ...
|
| 171 |
+
llama_model_loader: - kv 32: tokenizer.ggml.bos_token_id u32 = 1
|
| 172 |
+
llama_model_loader: - kv 33: tokenizer.ggml.eos_token_id u32 = 2
|
| 173 |
+
llama_model_loader: - kv 34: tokenizer.ggml.unknown_token_id u32 = 0
|
| 174 |
+
llama_model_loader: - kv 35: tokenizer.ggml.padding_token_id u32 = 0
|
| 175 |
+
llama_model_loader: - kv 36: general.quantization_version u32 = 2
|
| 176 |
+
llama_model_loader: - type f32: 17 tensors
|
| 177 |
+
llama_model_loader: - type f16: 58 tensors
|
| 178 |
+
llm_load_vocab: special tokens definition check successful ( 259/32000 ).
|
| 179 |
+
llm_load_print_meta: format = GGUF V3 (latest)
|
| 180 |
+
llm_load_print_meta: arch = llama
|
| 181 |
+
llm_load_print_meta: vocab type = SPM
|
| 182 |
+
llm_load_print_meta: n_vocab = 32000
|
| 183 |
+
llm_load_print_meta: n_merges = 0
|
| 184 |
+
llm_load_print_meta: n_ctx_train = 2048
|
| 185 |
+
llm_load_print_meta: n_embd = 64
|
| 186 |
+
llm_load_print_meta: n_head = 16
|
| 187 |
+
llm_load_print_meta: n_head_kv = 16
|
| 188 |
+
llm_load_print_meta: n_layer = 8
|
| 189 |
+
llm_load_print_meta: n_rot = 4
|
| 190 |
+
llm_load_print_meta: n_swa = 0
|
| 191 |
+
llm_load_print_meta: n_embd_head_k = 4
|
| 192 |
+
llm_load_print_meta: n_embd_head_v = 4
|
| 193 |
+
llm_load_print_meta: n_gqa = 1
|
| 194 |
+
llm_load_print_meta: n_embd_k_gqa = 64
|
| 195 |
+
llm_load_print_meta: n_embd_v_gqa = 64
|
| 196 |
+
llm_load_print_meta: f_norm_eps = 0.0e+00
|
| 197 |
+
llm_load_print_meta: f_norm_rms_eps = 1.0e-06
|
| 198 |
+
llm_load_print_meta: f_clamp_kqv = 0.0e+00
|
| 199 |
+
llm_load_print_meta: f_max_alibi_bias = 0.0e+00
|
| 200 |
+
llm_load_print_meta: f_logit_scale = 0.0e+00
|
| 201 |
+
llm_load_print_meta: n_ff = 256
|
| 202 |
+
llm_load_print_meta: n_expert = 0
|
| 203 |
+
llm_load_print_meta: n_expert_used = 0
|
| 204 |
+
llm_load_print_meta: causal attn = 1
|
| 205 |
+
llm_load_print_meta: pooling type = 0
|
| 206 |
+
llm_load_print_meta: rope type = 0
|
| 207 |
+
llm_load_print_meta: rope scaling = linear
|
| 208 |
+
llm_load_print_meta: freq_base_train = 10000.0
|
| 209 |
+
llm_load_print_meta: freq_scale_train = 1
|
| 210 |
+
llm_load_print_meta: n_yarn_orig_ctx = 2048
|
| 211 |
+
llm_load_print_meta: rope_finetuned = unknown
|
| 212 |
+
llm_load_print_meta: ssm_d_conv = 0
|
| 213 |
+
llm_load_print_meta: ssm_d_inner = 0
|
| 214 |
+
llm_load_print_meta: ssm_d_state = 0
|
| 215 |
+
llm_load_print_meta: ssm_dt_rank = 0
|
| 216 |
+
llm_load_print_meta: model type = ?B
|
| 217 |
+
llm_load_print_meta: model ftype = F16
|
| 218 |
+
llm_load_print_meta: model params = 4.62 M
|
| 219 |
+
llm_load_print_meta: model size = 8.82 MiB (16.00 BPW)
|
| 220 |
+
llm_load_print_meta: general.name = TinyLLama
|
| 221 |
+
llm_load_print_meta: BOS token = 1 '<s>'
|
| 222 |
+
llm_load_print_meta: EOS token = 2 '</s>'
|
| 223 |
+
llm_load_print_meta: UNK token = 0 '<unk>'
|
| 224 |
+
llm_load_print_meta: PAD token = 0 '<unk>'
|
| 225 |
+
llm_load_print_meta: LF token = 13 '<0x0A>'
|
| 226 |
+
llm_load_tensors: ggml ctx size = 0.04 MiB
|
| 227 |
+
llm_load_tensors: CPU buffer size = 8.82 MiB
|
| 228 |
+
..............
|
| 229 |
+
llama_new_context_with_model: n_ctx = 512
|
| 230 |
+
llama_new_context_with_model: n_batch = 512
|
| 231 |
+
llama_new_context_with_model: n_ubatch = 512
|
| 232 |
+
llama_new_context_with_model: flash_attn = 0
|
| 233 |
+
llama_new_context_with_model: freq_base = 10000.0
|
| 234 |
+
llama_new_context_with_model: freq_scale = 1
|
| 235 |
+
llama_kv_cache_init: CPU KV buffer size = 1.00 MiB
|
| 236 |
+
llama_new_context_with_model: KV self size = 1.00 MiB, K (f16): 0.50 MiB, V (f16): 0.50 MiB
|
| 237 |
+
llama_new_context_with_model: CPU output buffer size = 0.12 MiB
|
| 238 |
+
llama_new_context_with_model: CPU compute buffer size = 62.75 MiB
|
| 239 |
+
llama_new_context_with_model: graph nodes = 262
|
| 240 |
+
llama_new_context_with_model: graph splits = 1
|
| 241 |
+
|
| 242 |
+
system_info: n_threads = 4 / 8 | AVX = 1 | AVX_VNNI = 0 | AVX2 = 1 | AVX512 = 0 | AVX512_VBMI = 0 | AVX512_VNNI = 0 | AVX512_BF16 = 0 | FMA = 1 | NEON = 0 | ARM_FMA = 0 | F16C = 1 | FP16_VA = 0 | WASM_SIMD = 0 | BLAS = 0 | SSE3 = 1 | SSSE3 = 1 | VSX = 0 | MATMUL_INT8 = 0 | LLAMAFILE = 1 |
|
| 243 |
+
sampling:
|
| 244 |
+
repeat_last_n = 64, repeat_penalty = 1.000, frequency_penalty = 0.000, presence_penalty = 0.000
|
| 245 |
+
top_k = 40, tfs_z = 1.000, top_p = 0.950, min_p = 0.050, typical_p = 1.000, temp = 0.800
|
| 246 |
+
mirostat = 0, mirostat_lr = 0.100, mirostat_ent = 5.000
|
| 247 |
+
sampling order:
|
| 248 |
+
CFG -> Penalties -> top_k -> tfs_z -> typical_p -> top_p -> min_p -> temperature
|
| 249 |
+
generate: n_ctx = 512, n_batch = 2048, n_predict = -1, n_keep = 1
|
| 250 |
+
|
| 251 |
+
|
| 252 |
+
hello world the gruff man said he had a dream. He thought about what he had to do. He took a deep breath and ran around the woods. The man was so excited! He couldn't wait to do the whole day.
|
| 253 |
+
The man looked around and found a small box. He wanted to see what was inside. He picked up the box and started to climb. But the box was too hard. He looked around for the cake.
|
| 254 |
+
The man was sad and he didn't know what to do. He asked his friend to help him. The man said no. He said he couldn't get the oven. He was sad and began to cry.
|
| 255 |
+
The man felt bad because he knew it was okay. He wished he had been a good friend. He was very sad and he couldn't find the cake. [end of text]
|
| 256 |
+
|
| 257 |
+
|
| 258 |
+
llama_print_timings: load time = 6.74 ms
|
| 259 |
+
llama_print_timings: sample time = 6.76 ms / 169 runs ( 0.04 ms per token, 25011.10 tokens per second)
|
| 260 |
+
llama_print_timings: prompt eval time = 3.64 ms / 8 tokens ( 0.46 ms per token, 2196.60 tokens per second)
|
| 261 |
+
llama_print_timings: eval time = 340.85 ms / 168 runs ( 2.03 ms per token, 492.88 tokens per second)
|
| 262 |
+
llama_print_timings: total time = 392.63 ms / 176 tokens
|
| 263 |
+
Log end
|
| 264 |
+
note: if you have an AMD or NVIDIA GPU then you need to pass -ngl 9999 to enable GPU offloading
|
| 265 |
+
main: llamafile version 0.8.9
|
| 266 |
+
main: seed = 1721531082
|
| 267 |
+
llama_model_loader: loaded meta data with 37 key-value pairs and 75 tensors from TinyLLama-4.6M-v0.0-F16.gguf (version GGUF V3 (latest))
|
| 268 |
+
llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
|
| 269 |
+
llama_model_loader: - kv 0: general.architecture str = llama
|
| 270 |
+
llama_model_loader: - kv 1: general.type str = model
|
| 271 |
+
llama_model_loader: - kv 2: general.name str = TinyLLama
|
| 272 |
+
llama_model_loader: - kv 3: general.author str = Maykeye
|
| 273 |
+
llama_model_loader: - kv 4: general.version str = v0.0
|
| 274 |
+
llama_model_loader: - kv 5: general.description str = This gguf is ported from a first vers...
|
| 275 |
+
llama_model_loader: - kv 6: general.quantized_by str = Mofosyne
|
| 276 |
+
llama_model_loader: - kv 7: general.size_label str = 4.6M
|
| 277 |
+
llama_model_loader: - kv 8: general.license str = apache-2.0
|
| 278 |
+
llama_model_loader: - kv 9: general.license.name str = Apache License Version 2.0, January 2004
|
| 279 |
+
llama_model_loader: - kv 10: general.license.link str = https://huggingface.co/datasets/choos...
|
| 280 |
+
llama_model_loader: - kv 11: general.url str = https://huggingface.co/mofosyne/TinyL...
|
| 281 |
+
llama_model_loader: - kv 12: general.repo_url str = https://huggingface.co/mofosyne/TinyL...
|
| 282 |
+
llama_model_loader: - kv 13: general.source.url str = https://huggingface.co/Maykeye/TinyLL...
|
| 283 |
+
llama_model_loader: - kv 14: general.source.repo_url str = https://huggingface.co/Maykeye/TinyLL...
|
| 284 |
+
llama_model_loader: - kv 15: general.tags arr[str,5] = ["text generation", "transformer", "l...
|
| 285 |
+
llama_model_loader: - kv 16: general.languages arr[str,1] = ["en"]
|
| 286 |
+
llama_model_loader: - kv 17: general.datasets arr[str,2] = ["https://huggingface.co/datasets/ron...
|
| 287 |
+
llama_model_loader: - kv 18: llama.block_count u32 = 8
|
| 288 |
+
llama_model_loader: - kv 19: llama.context_length u32 = 2048
|
| 289 |
+
llama_model_loader: - kv 20: llama.embedding_length u32 = 64
|
| 290 |
+
llama_model_loader: - kv 21: llama.feed_forward_length u32 = 256
|
| 291 |
+
llama_model_loader: - kv 22: llama.attention.head_count u32 = 16
|
| 292 |
+
llama_model_loader: - kv 23: llama.attention.layer_norm_rms_epsilon f32 = 0.000001
|
| 293 |
+
llama_model_loader: - kv 24: general.file_type u32 = 1
|
| 294 |
+
llama_model_loader: - kv 25: llama.vocab_size u32 = 32000
|
| 295 |
+
llama_model_loader: - kv 26: llama.rope.dimension_count u32 = 4
|
| 296 |
+
llama_model_loader: - kv 27: tokenizer.ggml.model str = llama
|
| 297 |
+
llama_model_loader: - kv 28: tokenizer.ggml.pre str = default
|
| 298 |
+
llama_model_loader: - kv 29: tokenizer.ggml.tokens arr[str,32000] = ["<unk>", "<s>", "</s>", "<0x00>", "<...
|
| 299 |
+
llama_model_loader: - kv 30: tokenizer.ggml.scores arr[f32,32000] = [0.000000, 0.000000, 0.000000, 0.0000...
|
| 300 |
+
llama_model_loader: - kv 31: tokenizer.ggml.token_type arr[i32,32000] = [2, 3, 3, 6, 6, 6, 6, 6, 6, 6, 6, 6, ...
|
| 301 |
+
llama_model_loader: - kv 32: tokenizer.ggml.bos_token_id u32 = 1
|
| 302 |
+
llama_model_loader: - kv 33: tokenizer.ggml.eos_token_id u32 = 2
|
| 303 |
+
llama_model_loader: - kv 34: tokenizer.ggml.unknown_token_id u32 = 0
|
| 304 |
+
llama_model_loader: - kv 35: tokenizer.ggml.padding_token_id u32 = 0
|
| 305 |
+
llama_model_loader: - kv 36: general.quantization_version u32 = 2
|
| 306 |
+
llama_model_loader: - type f32: 17 tensors
|
| 307 |
+
llama_model_loader: - type f16: 58 tensors
|
| 308 |
+
llm_load_vocab: special tokens definition check successful ( 259/32000 ).
|
| 309 |
+
llm_load_print_meta: format = GGUF V3 (latest)
|
| 310 |
+
llm_load_print_meta: arch = llama
|
| 311 |
+
llm_load_print_meta: vocab type = SPM
|
| 312 |
+
llm_load_print_meta: n_vocab = 32000
|
| 313 |
+
llm_load_print_meta: n_merges = 0
|
| 314 |
+
llm_load_print_meta: n_ctx_train = 2048
|
| 315 |
+
llm_load_print_meta: n_embd = 64
|
| 316 |
+
llm_load_print_meta: n_head = 16
|
| 317 |
+
llm_load_print_meta: n_head_kv = 16
|
| 318 |
+
llm_load_print_meta: n_layer = 8
|
| 319 |
+
llm_load_print_meta: n_rot = 4
|
| 320 |
+
llm_load_print_meta: n_swa = 0
|
| 321 |
+
llm_load_print_meta: n_embd_head_k = 4
|
| 322 |
+
llm_load_print_meta: n_embd_head_v = 4
|
| 323 |
+
llm_load_print_meta: n_gqa = 1
|
| 324 |
+
llm_load_print_meta: n_embd_k_gqa = 64
|
| 325 |
+
llm_load_print_meta: n_embd_v_gqa = 64
|
| 326 |
+
llm_load_print_meta: f_norm_eps = 0.0e+00
|
| 327 |
+
llm_load_print_meta: f_norm_rms_eps = 1.0e-06
|
| 328 |
+
llm_load_print_meta: f_clamp_kqv = 0.0e+00
|
| 329 |
+
llm_load_print_meta: f_max_alibi_bias = 0.0e+00
|
| 330 |
+
llm_load_print_meta: f_logit_scale = 0.0e+00
|
| 331 |
+
llm_load_print_meta: n_ff = 256
|
| 332 |
+
llm_load_print_meta: n_expert = 0
|
| 333 |
+
llm_load_print_meta: n_expert_used = 0
|
| 334 |
+
llm_load_print_meta: causal attn = 1
|
| 335 |
+
llm_load_print_meta: pooling type = 0
|
| 336 |
+
llm_load_print_meta: rope type = 0
|
| 337 |
+
llm_load_print_meta: rope scaling = linear
|
| 338 |
+
llm_load_print_meta: freq_base_train = 10000.0
|
| 339 |
+
llm_load_print_meta: freq_scale_train = 1
|
| 340 |
+
llm_load_print_meta: n_yarn_orig_ctx = 2048
|
| 341 |
+
llm_load_print_meta: rope_finetuned = unknown
|
| 342 |
+
llm_load_print_meta: ssm_d_conv = 0
|
| 343 |
+
llm_load_print_meta: ssm_d_inner = 0
|
| 344 |
+
llm_load_print_meta: ssm_d_state = 0
|
| 345 |
+
llm_load_print_meta: ssm_dt_rank = 0
|
| 346 |
+
llm_load_print_meta: model type = ?B
|
| 347 |
+
llm_load_print_meta: model ftype = F16
|
| 348 |
+
llm_load_print_meta: model params = 4.62 M
|
| 349 |
+
llm_load_print_meta: model size = 8.82 MiB (16.00 BPW)
|
| 350 |
+
llm_load_print_meta: general.name = TinyLLama
|
| 351 |
+
llm_load_print_meta: BOS token = 1 '<s>'
|
| 352 |
+
llm_load_print_meta: EOS token = 2 '</s>'
|
| 353 |
+
llm_load_print_meta: UNK token = 0 '<unk>'
|
| 354 |
+
llm_load_print_meta: PAD token = 0 '<unk>'
|
| 355 |
+
llm_load_print_meta: LF token = 13 '<0x0A>'
|
| 356 |
+
llm_load_tensors: ggml ctx size = 0.04 MiB
|
| 357 |
+
llm_load_tensors: CPU buffer size = 8.82 MiB
|
| 358 |
+
..............
|
| 359 |
+
llama_new_context_with_model: n_ctx = 512
|
| 360 |
+
llama_new_context_with_model: n_batch = 512
|
| 361 |
+
llama_new_context_with_model: n_ubatch = 512
|
| 362 |
+
llama_new_context_with_model: flash_attn = 0
|
| 363 |
+
llama_new_context_with_model: freq_base = 10000.0
|
| 364 |
+
llama_new_context_with_model: freq_scale = 1
|
| 365 |
+
llama_kv_cache_init: CPU KV buffer size = 1.00 MiB
|
| 366 |
+
llama_new_context_with_model: KV self size = 1.00 MiB, K (f16): 0.50 MiB, V (f16): 0.50 MiB
|
| 367 |
+
llama_new_context_with_model: CPU output buffer size = 0.12 MiB
|
| 368 |
+
llama_new_context_with_model: CPU compute buffer size = 62.75 MiB
|
| 369 |
+
llama_new_context_with_model: graph nodes = 262
|
| 370 |
+
llama_new_context_with_model: graph splits = 1
|
| 371 |
+
|
| 372 |
+
system_info: n_threads = 4 / 8 | AVX = 1 | AVX_VNNI = 0 | AVX2 = 1 | AVX512 = 0 | AVX512_VBMI = 0 | AVX512_VNNI = 0 | AVX512_BF16 = 0 | FMA = 1 | NEON = 0 | ARM_FMA = 0 | F16C = 1 | FP16_VA = 0 | WASM_SIMD = 0 | BLAS = 0 | SSE3 = 1 | SSSE3 = 1 | VSX = 0 | MATMUL_INT8 = 0 | LLAMAFILE = 1 |
|
| 373 |
+
sampling:
|
| 374 |
+
repeat_last_n = 64, repeat_penalty = 1.000, frequency_penalty = 0.000, presence_penalty = 0.000
|
| 375 |
+
top_k = 40, tfs_z = 1.000, top_p = 0.950, min_p = 0.050, typical_p = 1.000, temp = 0.800
|
| 376 |
+
mirostat = 0, mirostat_lr = 0.100, mirostat_ent = 5.000
|
| 377 |
+
sampling order:
|
| 378 |
+
CFG -> Penalties -> top_k -> tfs_z -> typical_p -> top_p -> min_p -> temperature
|
| 379 |
+
generate: n_ctx = 512, n_batch = 2048, n_predict = -1, n_keep = 1
|
| 380 |
+
|
| 381 |
+
|
| 382 |
+
hello world the gruff man said yes and said hello to her.
|
| 383 |
+
The lady said to her, “I want to keep you a new friend! You are very nice and smart. I will be very proud."
|
| 384 |
+
The man smiled and said, “Yes, I will be more careful!”
|
| 385 |
+
The man looked up, but he knew she had a secret. He said, “I'm the right thing, but I have a special plan!”
|
| 386 |
+
The man was very surprised, but he smiled.
|
| 387 |
+
"That's so smart," he said. "I have to be careful to be kind to others. It's special to help me find things, but it's too late."
|
| 388 |
+
The man smiled and said, "No, you can't do it to be nice to you."
|
| 389 |
+
The man smiled and said, "You're welcome, I'm sorry for helping you. You are very nosy!"
|
| 390 |
+
The man was so happy. He knew that it had helped his friends, and he could be a friend. The people in the town loved to learn together. [end of text]
|
| 391 |
+
|
| 392 |
+
|
| 393 |
+
llama_print_timings: load time = 8.66 ms
|
| 394 |
+
llama_print_timings: sample time = 8.79 ms / 220 runs ( 0.04 ms per token, 25017.06 tokens per second)
|
| 395 |
+
llama_print_timings: prompt eval time = 1.59 ms / 8 tokens ( 0.20 ms per token, 5015.67 tokens per second)
|
| 396 |
+
llama_print_timings: eval time = 352.66 ms / 219 runs ( 1.61 ms per token, 620.99 tokens per second)
|
| 397 |
+
llama_print_timings: total time = 415.12 ms / 227 tokens
|
| 398 |
+
Log end
|
| 399 |
+
note: if you have an AMD or NVIDIA GPU then you need to pass -ngl 9999 to enable GPU offloading
|
| 400 |
+
main: llamafile version 0.8.9
|
| 401 |
+
main: seed = 1721531150
|
| 402 |
+
llama_model_loader: loaded meta data with 37 key-value pairs and 75 tensors from TinyLLama-4.6M-v0.0-F16.gguf (version GGUF V3 (latest))
|
| 403 |
+
llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
|
| 404 |
+
llama_model_loader: - kv 0: general.architecture str = llama
|
| 405 |
+
llama_model_loader: - kv 1: general.type str = model
|
| 406 |
+
llama_model_loader: - kv 2: general.name str = TinyLLama
|
| 407 |
+
llama_model_loader: - kv 3: general.author str = Maykeye
|
| 408 |
+
llama_model_loader: - kv 4: general.version str = v0.0
|
| 409 |
+
llama_model_loader: - kv 5: general.description str = This gguf is ported from a first vers...
|
| 410 |
+
llama_model_loader: - kv 6: general.quantized_by str = Mofosyne
|
| 411 |
+
llama_model_loader: - kv 7: general.size_label str = 4.6M
|
| 412 |
+
llama_model_loader: - kv 8: general.license str = apache-2.0
|
| 413 |
+
llama_model_loader: - kv 9: general.license.name str = Apache License Version 2.0, January 2004
|
| 414 |
+
llama_model_loader: - kv 10: general.license.link str = https://huggingface.co/datasets/choos...
|
| 415 |
+
llama_model_loader: - kv 11: general.url str = https://huggingface.co/mofosyne/TinyL...
|
| 416 |
+
llama_model_loader: - kv 12: general.repo_url str = https://huggingface.co/mofosyne/TinyL...
|
| 417 |
+
llama_model_loader: - kv 13: general.source.url str = https://huggingface.co/Maykeye/TinyLL...
|
| 418 |
+
llama_model_loader: - kv 14: general.source.repo_url str = https://huggingface.co/Maykeye/TinyLL...
|
| 419 |
+
llama_model_loader: - kv 15: general.tags arr[str,5] = ["text generation", "transformer", "l...
|
| 420 |
+
llama_model_loader: - kv 16: general.languages arr[str,1] = ["en"]
|
| 421 |
+
llama_model_loader: - kv 17: general.datasets arr[str,2] = ["https://huggingface.co/datasets/ron...
|
| 422 |
+
llama_model_loader: - kv 18: llama.block_count u32 = 8
|
| 423 |
+
llama_model_loader: - kv 19: llama.context_length u32 = 2048
|
| 424 |
+
llama_model_loader: - kv 20: llama.embedding_length u32 = 64
|
| 425 |
+
llama_model_loader: - kv 21: llama.feed_forward_length u32 = 256
|
| 426 |
+
llama_model_loader: - kv 22: llama.attention.head_count u32 = 16
|
| 427 |
+
llama_model_loader: - kv 23: llama.attention.layer_norm_rms_epsilon f32 = 0.000001
|
| 428 |
+
llama_model_loader: - kv 24: general.file_type u32 = 1
|
| 429 |
+
llama_model_loader: - kv 25: llama.vocab_size u32 = 32000
|
| 430 |
+
llama_model_loader: - kv 26: llama.rope.dimension_count u32 = 4
|
| 431 |
+
llama_model_loader: - kv 27: tokenizer.ggml.model str = llama
|
| 432 |
+
llama_model_loader: - kv 28: tokenizer.ggml.pre str = default
|
| 433 |
+
llama_model_loader: - kv 29: tokenizer.ggml.tokens arr[str,32000] = ["<unk>", "<s>", "</s>", "<0x00>", "<...
|
| 434 |
+
llama_model_loader: - kv 30: tokenizer.ggml.scores arr[f32,32000] = [0.000000, 0.000000, 0.000000, 0.0000...
|
| 435 |
+
llama_model_loader: - kv 31: tokenizer.ggml.token_type arr[i32,32000] = [2, 3, 3, 6, 6, 6, 6, 6, 6, 6, 6, 6, ...
|
| 436 |
+
llama_model_loader: - kv 32: tokenizer.ggml.bos_token_id u32 = 1
|
| 437 |
+
llama_model_loader: - kv 33: tokenizer.ggml.eos_token_id u32 = 2
|
| 438 |
+
llama_model_loader: - kv 34: tokenizer.ggml.unknown_token_id u32 = 0
|
| 439 |
+
llama_model_loader: - kv 35: tokenizer.ggml.padding_token_id u32 = 0
|
| 440 |
+
llama_model_loader: - kv 36: general.quantization_version u32 = 2
|
| 441 |
+
llama_model_loader: - type f32: 17 tensors
|
| 442 |
+
llama_model_loader: - type f16: 58 tensors
|
| 443 |
+
llm_load_vocab: special tokens definition check successful ( 259/32000 ).
|
| 444 |
+
llm_load_print_meta: format = GGUF V3 (latest)
|
| 445 |
+
llm_load_print_meta: arch = llama
|
| 446 |
+
llm_load_print_meta: vocab type = SPM
|
| 447 |
+
llm_load_print_meta: n_vocab = 32000
|
| 448 |
+
llm_load_print_meta: n_merges = 0
|
| 449 |
+
llm_load_print_meta: n_ctx_train = 2048
|
| 450 |
+
llm_load_print_meta: n_embd = 64
|
| 451 |
+
llm_load_print_meta: n_head = 16
|
| 452 |
+
llm_load_print_meta: n_head_kv = 16
|
| 453 |
+
llm_load_print_meta: n_layer = 8
|
| 454 |
+
llm_load_print_meta: n_rot = 4
|
| 455 |
+
llm_load_print_meta: n_swa = 0
|
| 456 |
+
llm_load_print_meta: n_embd_head_k = 4
|
| 457 |
+
llm_load_print_meta: n_embd_head_v = 4
|
| 458 |
+
llm_load_print_meta: n_gqa = 1
|
| 459 |
+
llm_load_print_meta: n_embd_k_gqa = 64
|
| 460 |
+
llm_load_print_meta: n_embd_v_gqa = 64
|
| 461 |
+
llm_load_print_meta: f_norm_eps = 0.0e+00
|
| 462 |
+
llm_load_print_meta: f_norm_rms_eps = 1.0e-06
|
| 463 |
+
llm_load_print_meta: f_clamp_kqv = 0.0e+00
|
| 464 |
+
llm_load_print_meta: f_max_alibi_bias = 0.0e+00
|
| 465 |
+
llm_load_print_meta: f_logit_scale = 0.0e+00
|
| 466 |
+
llm_load_print_meta: n_ff = 256
|
| 467 |
+
llm_load_print_meta: n_expert = 0
|
| 468 |
+
llm_load_print_meta: n_expert_used = 0
|
| 469 |
+
llm_load_print_meta: causal attn = 1
|
| 470 |
+
llm_load_print_meta: pooling type = 0
|
| 471 |
+
llm_load_print_meta: rope type = 0
|
| 472 |
+
llm_load_print_meta: rope scaling = linear
|
| 473 |
+
llm_load_print_meta: freq_base_train = 10000.0
|
| 474 |
+
llm_load_print_meta: freq_scale_train = 1
|
| 475 |
+
llm_load_print_meta: n_yarn_orig_ctx = 2048
|
| 476 |
+
llm_load_print_meta: rope_finetuned = unknown
|
| 477 |
+
llm_load_print_meta: ssm_d_conv = 0
|
| 478 |
+
llm_load_print_meta: ssm_d_inner = 0
|
| 479 |
+
llm_load_print_meta: ssm_d_state = 0
|
| 480 |
+
llm_load_print_meta: ssm_dt_rank = 0
|
| 481 |
+
llm_load_print_meta: model type = ?B
|
| 482 |
+
llm_load_print_meta: model ftype = F16
|
| 483 |
+
llm_load_print_meta: model params = 4.62 M
|
| 484 |
+
llm_load_print_meta: model size = 8.82 MiB (16.00 BPW)
|
| 485 |
+
llm_load_print_meta: general.name = TinyLLama
|
| 486 |
+
llm_load_print_meta: BOS token = 1 '<s>'
|
| 487 |
+
llm_load_print_meta: EOS token = 2 '</s>'
|
| 488 |
+
llm_load_print_meta: UNK token = 0 '<unk>'
|
| 489 |
+
llm_load_print_meta: PAD token = 0 '<unk>'
|
| 490 |
+
llm_load_print_meta: LF token = 13 '<0x0A>'
|
| 491 |
+
llm_load_tensors: ggml ctx size = 0.04 MiB
|
| 492 |
+
llm_load_tensors: CPU buffer size = 8.82 MiB
|
| 493 |
+
..............
|
| 494 |
+
llama_new_context_with_model: n_ctx = 512
|
| 495 |
+
llama_new_context_with_model: n_batch = 512
|
| 496 |
+
llama_new_context_with_model: n_ubatch = 512
|
| 497 |
+
llama_new_context_with_model: flash_attn = 0
|
| 498 |
+
llama_new_context_with_model: freq_base = 10000.0
|
| 499 |
+
llama_new_context_with_model: freq_scale = 1
|
| 500 |
+
llama_kv_cache_init: CPU KV buffer size = 1.00 MiB
|
| 501 |
+
llama_new_context_with_model: KV self size = 1.00 MiB, K (f16): 0.50 MiB, V (f16): 0.50 MiB
|
| 502 |
+
llama_new_context_with_model: CPU output buffer size = 0.12 MiB
|
| 503 |
+
llama_new_context_with_model: CPU compute buffer size = 62.75 MiB
|
| 504 |
+
llama_new_context_with_model: graph nodes = 262
|
| 505 |
+
llama_new_context_with_model: graph splits = 1
|
| 506 |
+
|
| 507 |
+
system_info: n_threads = 4 / 8 | AVX = 1 | AVX_VNNI = 0 | AVX2 = 1 | AVX512 = 0 | AVX512_VBMI = 0 | AVX512_VNNI = 0 | AVX512_BF16 = 0 | FMA = 1 | NEON = 0 | ARM_FMA = 0 | F16C = 1 | FP16_VA = 0 | WASM_SIMD = 0 | BLAS = 0 | SSE3 = 1 | SSSE3 = 1 | VSX = 0 | MATMUL_INT8 = 0 | LLAMAFILE = 1 |
|
| 508 |
+
sampling:
|
| 509 |
+
repeat_last_n = 64, repeat_penalty = 1.000, frequency_penalty = 0.000, presence_penalty = 0.000
|
| 510 |
+
top_k = 40, tfs_z = 1.000, top_p = 0.950, min_p = 0.050, typical_p = 1.000, temp = 0.800
|
| 511 |
+
mirostat = 0, mirostat_lr = 0.100, mirostat_ent = 5.000
|
| 512 |
+
sampling order:
|
| 513 |
+
CFG -> Penalties -> top_k -> tfs_z -> typical_p -> top_p -> min_p -> temperature
|
| 514 |
+
generate: n_ctx = 512, n_batch = 2048, n_predict = -1, n_keep = 1
|
| 515 |
+
|
| 516 |
+
|
| 517 |
+
hello world the gruff man said, “Hi, do you want to do something?". He looked up and said, “No!"
|
| 518 |
+
The man looked around and saw the big, tall hill. He wanted to see where he could see it. He jumped in and was so excited!
|
| 519 |
+
The lady said, “It's so messy! This is important to be careful if you can't make a big smile. Thank you for being very nice here."
|
| 520 |
+
The man thought for a moment and then said, “I'm sorry, I would want to eat it now!"
|
| 521 |
+
The man nodded. He knew he could help the girl. He said, “No, we should not try it."
|
| 522 |
+
The boy was sad and said, “No, I don’t need some food. I'll be my friend."
|
| 523 |
+
The man thought for a moment and then said, “You can't have to leave the garden when you're going to the park! We must always share it with your friends".
|
| 524 |
+
The boy smiled and said, “Of course".
|
| 525 |
+
The man smiled and said, “I'm glad you found this party! I'm very excited to play with you.”
|
| 526 |
+
The boy smiled. He was very proud of himself. He said, “Do you want to have some fun? I like to watch this game. I don’t want to ask me if you can be good."
|
| 527 |
+
The boy smiled and said, "Yes, I will, let's go, let's go!"
|
| 528 |
+
So, the boy and the boy hopped and started to walk and see the stars. They laughed and laughed, but he was very proud. [end of text]
|
| 529 |
+
|
| 530 |
+
|
| 531 |
+
llama_print_timings: load time = 17.26 ms
|
| 532 |
+
llama_print_timings: sample time = 13.20 ms / 347 runs ( 0.04 ms per token, 26287.88 tokens per second)
|
| 533 |
+
llama_print_timings: prompt eval time = 2.46 ms / 8 tokens ( 0.31 ms per token, 3257.33 tokens per second)
|
| 534 |
+
llama_print_timings: eval time = 550.67 ms / 346 runs ( 1.59 ms per token, 628.33 tokens per second)
|
| 535 |
+
llama_print_timings: total time = 649.30 ms / 354 tokens
|
| 536 |
+
Log end
|
| 537 |
+
note: if you have an AMD or NVIDIA GPU then you need to pass -ngl 9999 to enable GPU offloading
|
| 538 |
+
main: llamafile version 0.8.9
|
| 539 |
+
main: seed = 1721531259
|
| 540 |
+
llama_model_loader: loaded meta data with 37 key-value pairs and 75 tensors from TinyLLama-4.6M-v0.0-F16.gguf (version GGUF V3 (latest))
|
| 541 |
+
llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
|
| 542 |
+
llama_model_loader: - kv 0: general.architecture str = llama
|
| 543 |
+
llama_model_loader: - kv 1: general.type str = model
|
| 544 |
+
llama_model_loader: - kv 2: general.name str = TinyLLama
|
| 545 |
+
llama_model_loader: - kv 3: general.author str = Maykeye
|
| 546 |
+
llama_model_loader: - kv 4: general.version str = v0.0
|
| 547 |
+
llama_model_loader: - kv 5: general.description str = This gguf is ported from a first vers...
|
| 548 |
+
llama_model_loader: - kv 6: general.quantized_by str = Mofosyne
|
| 549 |
+
llama_model_loader: - kv 7: general.size_label str = 4.6M
|
| 550 |
+
llama_model_loader: - kv 8: general.license str = apache-2.0
|
| 551 |
+
llama_model_loader: - kv 9: general.license.name str = Apache License Version 2.0, January 2004
|
| 552 |
+
llama_model_loader: - kv 10: general.license.link str = https://huggingface.co/datasets/choos...
|
| 553 |
+
llama_model_loader: - kv 11: general.url str = https://huggingface.co/mofosyne/TinyL...
|
| 554 |
+
llama_model_loader: - kv 12: general.repo_url str = https://huggingface.co/mofosyne/TinyL...
|
| 555 |
+
llama_model_loader: - kv 13: general.source.url str = https://huggingface.co/Maykeye/TinyLL...
|
| 556 |
+
llama_model_loader: - kv 14: general.source.repo_url str = https://huggingface.co/Maykeye/TinyLL...
|
| 557 |
+
llama_model_loader: - kv 15: general.tags arr[str,5] = ["text generation", "transformer", "l...
|
| 558 |
+
llama_model_loader: - kv 16: general.languages arr[str,1] = ["en"]
|
| 559 |
+
llama_model_loader: - kv 17: general.datasets arr[str,2] = ["https://huggingface.co/datasets/ron...
|
| 560 |
+
llama_model_loader: - kv 18: llama.block_count u32 = 8
|
| 561 |
+
llama_model_loader: - kv 19: llama.context_length u32 = 2048
|
| 562 |
+
llama_model_loader: - kv 20: llama.embedding_length u32 = 64
|
| 563 |
+
llama_model_loader: - kv 21: llama.feed_forward_length u32 = 256
|
| 564 |
+
llama_model_loader: - kv 22: llama.attention.head_count u32 = 16
|
| 565 |
+
llama_model_loader: - kv 23: llama.attention.layer_norm_rms_epsilon f32 = 0.000001
|
| 566 |
+
llama_model_loader: - kv 24: general.file_type u32 = 1
|
| 567 |
+
llama_model_loader: - kv 25: llama.vocab_size u32 = 32000
|
| 568 |
+
llama_model_loader: - kv 26: llama.rope.dimension_count u32 = 4
|
| 569 |
+
llama_model_loader: - kv 27: tokenizer.ggml.model str = llama
|
| 570 |
+
llama_model_loader: - kv 28: tokenizer.ggml.pre str = default
|
| 571 |
+
llama_model_loader: - kv 29: tokenizer.ggml.tokens arr[str,32000] = ["<unk>", "<s>", "</s>", "<0x00>", "<...
|
| 572 |
+
llama_model_loader: - kv 30: tokenizer.ggml.scores arr[f32,32000] = [0.000000, 0.000000, 0.000000, 0.0000...
|
| 573 |
+
llama_model_loader: - kv 31: tokenizer.ggml.token_type arr[i32,32000] = [2, 3, 3, 6, 6, 6, 6, 6, 6, 6, 6, 6, ...
|
| 574 |
+
llama_model_loader: - kv 32: tokenizer.ggml.bos_token_id u32 = 1
|
| 575 |
+
llama_model_loader: - kv 33: tokenizer.ggml.eos_token_id u32 = 2
|
| 576 |
+
llama_model_loader: - kv 34: tokenizer.ggml.unknown_token_id u32 = 0
|
| 577 |
+
llama_model_loader: - kv 35: tokenizer.ggml.padding_token_id u32 = 0
|
| 578 |
+
llama_model_loader: - kv 36: general.quantization_version u32 = 2
|
| 579 |
+
llama_model_loader: - type f32: 17 tensors
|
| 580 |
+
llama_model_loader: - type f16: 58 tensors
|
| 581 |
+
llm_load_vocab: special tokens definition check successful ( 259/32000 ).
|
| 582 |
+
llm_load_print_meta: format = GGUF V3 (latest)
|
| 583 |
+
llm_load_print_meta: arch = llama
|
| 584 |
+
llm_load_print_meta: vocab type = SPM
|
| 585 |
+
llm_load_print_meta: n_vocab = 32000
|
| 586 |
+
llm_load_print_meta: n_merges = 0
|
| 587 |
+
llm_load_print_meta: n_ctx_train = 2048
|
| 588 |
+
llm_load_print_meta: n_embd = 64
|
| 589 |
+
llm_load_print_meta: n_head = 16
|
| 590 |
+
llm_load_print_meta: n_head_kv = 16
|
| 591 |
+
llm_load_print_meta: n_layer = 8
|
| 592 |
+
llm_load_print_meta: n_rot = 4
|
| 593 |
+
llm_load_print_meta: n_swa = 0
|
| 594 |
+
llm_load_print_meta: n_embd_head_k = 4
|
| 595 |
+
llm_load_print_meta: n_embd_head_v = 4
|
| 596 |
+
llm_load_print_meta: n_gqa = 1
|
| 597 |
+
llm_load_print_meta: n_embd_k_gqa = 64
|
| 598 |
+
llm_load_print_meta: n_embd_v_gqa = 64
|
| 599 |
+
llm_load_print_meta: f_norm_eps = 0.0e+00
|
| 600 |
+
llm_load_print_meta: f_norm_rms_eps = 1.0e-06
|
| 601 |
+
llm_load_print_meta: f_clamp_kqv = 0.0e+00
|
| 602 |
+
llm_load_print_meta: f_max_alibi_bias = 0.0e+00
|
| 603 |
+
llm_load_print_meta: f_logit_scale = 0.0e+00
|
| 604 |
+
llm_load_print_meta: n_ff = 256
|
| 605 |
+
llm_load_print_meta: n_expert = 0
|
| 606 |
+
llm_load_print_meta: n_expert_used = 0
|
| 607 |
+
llm_load_print_meta: causal attn = 1
|
| 608 |
+
llm_load_print_meta: pooling type = 0
|
| 609 |
+
llm_load_print_meta: rope type = 0
|
| 610 |
+
llm_load_print_meta: rope scaling = linear
|
| 611 |
+
llm_load_print_meta: freq_base_train = 10000.0
|
| 612 |
+
llm_load_print_meta: freq_scale_train = 1
|
| 613 |
+
llm_load_print_meta: n_yarn_orig_ctx = 2048
|
| 614 |
+
llm_load_print_meta: rope_finetuned = unknown
|
| 615 |
+
llm_load_print_meta: ssm_d_conv = 0
|
| 616 |
+
llm_load_print_meta: ssm_d_inner = 0
|
| 617 |
+
llm_load_print_meta: ssm_d_state = 0
|
| 618 |
+
llm_load_print_meta: ssm_dt_rank = 0
|
| 619 |
+
llm_load_print_meta: model type = ?B
|
| 620 |
+
llm_load_print_meta: model ftype = F16
|
| 621 |
+
llm_load_print_meta: model params = 4.62 M
|
| 622 |
+
llm_load_print_meta: model size = 8.82 MiB (16.00 BPW)
|
| 623 |
+
llm_load_print_meta: general.name = TinyLLama
|
| 624 |
+
llm_load_print_meta: BOS token = 1 '<s>'
|
| 625 |
+
llm_load_print_meta: EOS token = 2 '</s>'
|
| 626 |
+
llm_load_print_meta: UNK token = 0 '<unk>'
|
| 627 |
+
llm_load_print_meta: PAD token = 0 '<unk>'
|
| 628 |
+
llm_load_print_meta: LF token = 13 '<0x0A>'
|
| 629 |
+
llm_load_tensors: ggml ctx size = 0.04 MiB
|
| 630 |
+
llm_load_tensors: CPU buffer size = 8.82 MiB
|
| 631 |
+
..............
|
| 632 |
+
llama_new_context_with_model: n_ctx = 512
|
| 633 |
+
llama_new_context_with_model: n_batch = 512
|
| 634 |
+
llama_new_context_with_model: n_ubatch = 512
|
| 635 |
+
llama_new_context_with_model: flash_attn = 0
|
| 636 |
+
llama_new_context_with_model: freq_base = 10000.0
|
| 637 |
+
llama_new_context_with_model: freq_scale = 1
|
| 638 |
+
llama_kv_cache_init: CPU KV buffer size = 1.00 MiB
|
| 639 |
+
llama_new_context_with_model: KV self size = 1.00 MiB, K (f16): 0.50 MiB, V (f16): 0.50 MiB
|
| 640 |
+
llama_new_context_with_model: CPU output buffer size = 0.12 MiB
|
| 641 |
+
llama_new_context_with_model: CPU compute buffer size = 62.75 MiB
|
| 642 |
+
llama_new_context_with_model: graph nodes = 262
|
| 643 |
+
llama_new_context_with_model: graph splits = 1
|
| 644 |
+
|
| 645 |
+
system_info: n_threads = 4 / 8 | AVX = 1 | AVX_VNNI = 0 | AVX2 = 1 | AVX512 = 0 | AVX512_VBMI = 0 | AVX512_VNNI = 0 | AVX512_BF16 = 0 | FMA = 1 | NEON = 0 | ARM_FMA = 0 | F16C = 1 | FP16_VA = 0 | WASM_SIMD = 0 | BLAS = 0 | SSE3 = 1 | SSSE3 = 1 | VSX = 0 | MATMUL_INT8 = 0 | LLAMAFILE = 1 |
|
| 646 |
+
sampling:
|
| 647 |
+
repeat_last_n = 64, repeat_penalty = 1.000, frequency_penalty = 0.000, presence_penalty = 0.000
|
| 648 |
+
top_k = 40, tfs_z = 1.000, top_p = 0.950, min_p = 0.050, typical_p = 1.000, temp = 0.800
|
| 649 |
+
mirostat = 0, mirostat_lr = 0.100, mirostat_ent = 5.000
|
| 650 |
+
sampling order:
|
| 651 |
+
CFG -> Penalties -> top_k -> tfs_z -> typical_p -> top_p -> min_p -> temperature
|
| 652 |
+
generate: n_ctx = 512, n_batch = 2048, n_predict = -1, n_keep = 1
|
| 653 |
+
|
| 654 |
+
|
| 655 |
+
hello world the gruff man said he would help her. As he ran, he noticed something shiny in the sky. He looked around and saw a small, old lady, who was so excited! She said, “Can I try this?"
|
| 656 |
+
The old lady smiled and said, “Yes, but I have to keep the egg. It is so nice!”
|
| 657 |
+
The old man smiled. He said, “Yes, that is a good idea! I will stay in your house and give you a hug!"
|
| 658 |
+
The old man smiled, but then he said, “We can be very careful when you take it away". He said: “I want to be brave," The old man was so proud of his work. He said, “I need to be happy with this. Let's play together!"
|
| 659 |
+
The old man said, “No, I don’t want to go. We need to be careful."
|
| 660 |
+
The old man said, “Don't worry, I will be happy."
|
| 661 |
+
The old man smiled and said, "It's okay. We can try to take some more times. But be careful. Maybe you can't stop your friends."
|
| 662 |
+
The old man smiled and said, “Yes, you can. I'm here to help you. It's time for this problem." The old man nodded and said, “I will find it! We can are very careful with it".
|
| 663 |
+
The old man agreed. He gave the ugly man a big hug and said, “I know you would like it, but you don't need it."
|
| 664 |
+
The old man smiled and said, “You do so. I like the old man. I can be nice. He is my friend and I will help you get your way back to the party."
|
| 665 |
+
The old man smiled and said, “You don’s okay. You're so brave to find it, and I'm glad you have a new friend. That's a very nice idea and I'll take your things. [end of text]
|
| 666 |
+
|
| 667 |
+
|
| 668 |
+
llama_print_timings: load time = 7.35 ms
|
| 669 |
+
llama_print_timings: sample time = 13.73 ms / 409 runs ( 0.03 ms per token, 29780.11 tokens per second)
|
| 670 |
+
llama_print_timings: prompt eval time = 1.57 ms / 8 tokens ( 0.20 ms per token, 5102.04 tokens per second)
|
| 671 |
+
llama_print_timings: eval time = 972.37 ms / 408 runs ( 2.38 ms per token, 419.59 tokens per second)
|
| 672 |
+
llama_print_timings: total time = 1089.29 ms / 416 tokens
|
| 673 |
+
Log end
|
| 674 |
+
note: if you have an AMD or NVIDIA GPU then you need to pass -ngl 9999 to enable GPU offloading
|
| 675 |
+
main: llamafile version 0.8.9
|
| 676 |
+
main: seed = 1721532670
|
| 677 |
+
llama_model_loader: loaded meta data with 37 key-value pairs and 75 tensors from TinyLLama-4.6M-v0.0-F16.gguf (version GGUF V3 (latest))
|
| 678 |
+
llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
|
| 679 |
+
llama_model_loader: - kv 0: general.architecture str = llama
|
| 680 |
+
llama_model_loader: - kv 1: general.type str = model
|
| 681 |
+
llama_model_loader: - kv 2: general.name str = TinyLLama
|
| 682 |
+
llama_model_loader: - kv 3: general.author str = Maykeye
|
| 683 |
+
llama_model_loader: - kv 4: general.version str = v0.0
|
| 684 |
+
llama_model_loader: - kv 5: general.description str = This gguf is ported from a first vers...
|
| 685 |
+
llama_model_loader: - kv 6: general.quantized_by str = Mofosyne
|
| 686 |
+
llama_model_loader: - kv 7: general.size_label str = 4.6M
|
| 687 |
+
llama_model_loader: - kv 8: general.license str = apache-2.0
|
| 688 |
+
llama_model_loader: - kv 9: general.license.name str = Apache License Version 2.0, January 2004
|
| 689 |
+
llama_model_loader: - kv 10: general.license.link str = https://huggingface.co/datasets/choos...
|
| 690 |
+
llama_model_loader: - kv 11: general.url str = https://huggingface.co/mofosyne/TinyL...
|
| 691 |
+
llama_model_loader: - kv 12: general.repo_url str = https://huggingface.co/mofosyne/TinyL...
|
| 692 |
+
llama_model_loader: - kv 13: general.source.url str = https://huggingface.co/Maykeye/TinyLL...
|
| 693 |
+
llama_model_loader: - kv 14: general.source.repo_url str = https://huggingface.co/Maykeye/TinyLL...
|
| 694 |
+
llama_model_loader: - kv 15: general.tags arr[str,5] = ["text generation", "transformer", "l...
|
| 695 |
+
llama_model_loader: - kv 16: general.languages arr[str,1] = ["en"]
|
| 696 |
+
llama_model_loader: - kv 17: general.datasets arr[str,2] = ["https://huggingface.co/datasets/ron...
|
| 697 |
+
llama_model_loader: - kv 18: llama.block_count u32 = 8
|
| 698 |
+
llama_model_loader: - kv 19: llama.context_length u32 = 2048
|
| 699 |
+
llama_model_loader: - kv 20: llama.embedding_length u32 = 64
|
| 700 |
+
llama_model_loader: - kv 21: llama.feed_forward_length u32 = 256
|
| 701 |
+
llama_model_loader: - kv 22: llama.attention.head_count u32 = 16
|
| 702 |
+
llama_model_loader: - kv 23: llama.attention.layer_norm_rms_epsilon f32 = 0.000001
|
| 703 |
+
llama_model_loader: - kv 24: general.file_type u32 = 1
|
| 704 |
+
llama_model_loader: - kv 25: llama.vocab_size u32 = 32000
|
| 705 |
+
llama_model_loader: - kv 26: llama.rope.dimension_count u32 = 4
|
| 706 |
+
llama_model_loader: - kv 27: tokenizer.ggml.model str = llama
|
| 707 |
+
llama_model_loader: - kv 28: tokenizer.ggml.pre str = default
|
| 708 |
+
llama_model_loader: - kv 29: tokenizer.ggml.tokens arr[str,32000] = ["<unk>", "<s>", "</s>", "<0x00>", "<...
|
| 709 |
+
llama_model_loader: - kv 30: tokenizer.ggml.scores arr[f32,32000] = [0.000000, 0.000000, 0.000000, 0.0000...
|
| 710 |
+
llama_model_loader: - kv 31: tokenizer.ggml.token_type arr[i32,32000] = [2, 3, 3, 6, 6, 6, 6, 6, 6, 6, 6, 6, ...
|
| 711 |
+
llama_model_loader: - kv 32: tokenizer.ggml.bos_token_id u32 = 1
|
| 712 |
+
llama_model_loader: - kv 33: tokenizer.ggml.eos_token_id u32 = 2
|
| 713 |
+
llama_model_loader: - kv 34: tokenizer.ggml.unknown_token_id u32 = 0
|
| 714 |
+
llama_model_loader: - kv 35: tokenizer.ggml.padding_token_id u32 = 0
|
| 715 |
+
llama_model_loader: - kv 36: general.quantization_version u32 = 2
|
| 716 |
+
llama_model_loader: - type f32: 17 tensors
|
| 717 |
+
llama_model_loader: - type f16: 58 tensors
|
| 718 |
+
llm_load_vocab: special tokens definition check successful ( 259/32000 ).
|
| 719 |
+
llm_load_print_meta: format = GGUF V3 (latest)
|
| 720 |
+
llm_load_print_meta: arch = llama
|
| 721 |
+
llm_load_print_meta: vocab type = SPM
|
| 722 |
+
llm_load_print_meta: n_vocab = 32000
|
| 723 |
+
llm_load_print_meta: n_merges = 0
|
| 724 |
+
llm_load_print_meta: n_ctx_train = 2048
|
| 725 |
+
llm_load_print_meta: n_embd = 64
|
| 726 |
+
llm_load_print_meta: n_head = 16
|
| 727 |
+
llm_load_print_meta: n_head_kv = 16
|
| 728 |
+
llm_load_print_meta: n_layer = 8
|
| 729 |
+
llm_load_print_meta: n_rot = 4
|
| 730 |
+
llm_load_print_meta: n_swa = 0
|
| 731 |
+
llm_load_print_meta: n_embd_head_k = 4
|
| 732 |
+
llm_load_print_meta: n_embd_head_v = 4
|
| 733 |
+
llm_load_print_meta: n_gqa = 1
|
| 734 |
+
llm_load_print_meta: n_embd_k_gqa = 64
|
| 735 |
+
llm_load_print_meta: n_embd_v_gqa = 64
|
| 736 |
+
llm_load_print_meta: f_norm_eps = 0.0e+00
|
| 737 |
+
llm_load_print_meta: f_norm_rms_eps = 1.0e-06
|
| 738 |
+
llm_load_print_meta: f_clamp_kqv = 0.0e+00
|
| 739 |
+
llm_load_print_meta: f_max_alibi_bias = 0.0e+00
|
| 740 |
+
llm_load_print_meta: f_logit_scale = 0.0e+00
|
| 741 |
+
llm_load_print_meta: n_ff = 256
|
| 742 |
+
llm_load_print_meta: n_expert = 0
|
| 743 |
+
llm_load_print_meta: n_expert_used = 0
|
| 744 |
+
llm_load_print_meta: causal attn = 1
|
| 745 |
+
llm_load_print_meta: pooling type = 0
|
| 746 |
+
llm_load_print_meta: rope type = 0
|
| 747 |
+
llm_load_print_meta: rope scaling = linear
|
| 748 |
+
llm_load_print_meta: freq_base_train = 10000.0
|
| 749 |
+
llm_load_print_meta: freq_scale_train = 1
|
| 750 |
+
llm_load_print_meta: n_yarn_orig_ctx = 2048
|
| 751 |
+
llm_load_print_meta: rope_finetuned = unknown
|
| 752 |
+
llm_load_print_meta: ssm_d_conv = 0
|
| 753 |
+
llm_load_print_meta: ssm_d_inner = 0
|
| 754 |
+
llm_load_print_meta: ssm_d_state = 0
|
| 755 |
+
llm_load_print_meta: ssm_dt_rank = 0
|
| 756 |
+
llm_load_print_meta: model type = ?B
|
| 757 |
+
llm_load_print_meta: model ftype = F16
|
| 758 |
+
llm_load_print_meta: model params = 4.62 M
|
| 759 |
+
llm_load_print_meta: model size = 8.82 MiB (16.00 BPW)
|
| 760 |
+
llm_load_print_meta: general.name = TinyLLama
|
| 761 |
+
llm_load_print_meta: BOS token = 1 '<s>'
|
| 762 |
+
llm_load_print_meta: EOS token = 2 '</s>'
|
| 763 |
+
llm_load_print_meta: UNK token = 0 '<unk>'
|
| 764 |
+
llm_load_print_meta: PAD token = 0 '<unk>'
|
| 765 |
+
llm_load_print_meta: LF token = 13 '<0x0A>'
|
| 766 |
+
llm_load_tensors: ggml ctx size = 0.04 MiB
|
| 767 |
+
llm_load_tensors: CPU buffer size = 8.82 MiB
|
| 768 |
+
..............
|
| 769 |
+
llama_new_context_with_model: n_ctx = 512
|
| 770 |
+
llama_new_context_with_model: n_batch = 512
|
| 771 |
+
llama_new_context_with_model: n_ubatch = 512
|
| 772 |
+
llama_new_context_with_model: flash_attn = 0
|
| 773 |
+
llama_new_context_with_model: freq_base = 10000.0
|
| 774 |
+
llama_new_context_with_model: freq_scale = 1
|
| 775 |
+
llama_kv_cache_init: CPU KV buffer size = 1.00 MiB
|
| 776 |
+
llama_new_context_with_model: KV self size = 1.00 MiB, K (f16): 0.50 MiB, V (f16): 0.50 MiB
|
| 777 |
+
llama_new_context_with_model: CPU output buffer size = 0.12 MiB
|
| 778 |
+
llama_new_context_with_model: CPU compute buffer size = 62.75 MiB
|
| 779 |
+
llama_new_context_with_model: graph nodes = 262
|
| 780 |
+
llama_new_context_with_model: graph splits = 1
|
| 781 |
+
|
| 782 |
+
system_info: n_threads = 4 / 8 | AVX = 1 | AVX_VNNI = 0 | AVX2 = 1 | AVX512 = 0 | AVX512_VBMI = 0 | AVX512_VNNI = 0 | AVX512_BF16 = 0 | FMA = 1 | NEON = 0 | ARM_FMA = 0 | F16C = 1 | FP16_VA = 0 | WASM_SIMD = 0 | BLAS = 0 | SSE3 = 1 | SSSE3 = 1 | VSX = 0 | MATMUL_INT8 = 0 | LLAMAFILE = 1 |
|
| 783 |
+
sampling:
|
| 784 |
+
repeat_last_n = 64, repeat_penalty = 1.000, frequency_penalty = 0.000, presence_penalty = 0.000
|
| 785 |
+
top_k = 40, tfs_z = 1.000, top_p = 0.950, min_p = 0.050, typical_p = 1.000, temp = 0.800
|
| 786 |
+
mirostat = 0, mirostat_lr = 0.100, mirostat_ent = 5.000
|
| 787 |
+
sampling order:
|
| 788 |
+
CFG -> Penalties -> top_k -> tfs_z -> typical_p -> top_p -> min_p -> temperature
|
| 789 |
+
generate: n_ctx = 512, n_batch = 2048, n_predict = -1, n_keep = 1
|
| 790 |
+
|
| 791 |
+
|
| 792 |
+
hello world the gruff man said yes. The man said he could go. The man said the man was very brave and he wanted to go. The man was very curious and he decided to go on a trip.
|
| 793 |
+
After the story, the man was ready to go. He waved goodbye and said, "Don't worry, I will go on a walk." The man stepped down and said, "Don't be scared, I'xy. The man was a brave boy!"
|
| 794 |
+
The man smiled and said, "I'm sorry, I am so hungry." The man smiled and said, "That's nice. I'm proud of you."
|
| 795 |
+
The man said, "You can go with me. The man can fly fast for you, and I'll come with you." The man said, "Yes, let's do it!"
|
| 796 |
+
The man and the man ran around the village, and the man was happy. He was very kind. He was happy to have a friend. The man said, "You are very nice!"
|
| 797 |
+
The man hugged the man and said, "We are welcome, little boy. We were so proud of you. We are best friends."
|
| 798 |
+
The man smiled and said, "You are right, little boy, it's not a good thing to do. I have the way to the party to have fun." [end of text]
|
| 799 |
+
|
| 800 |
+
|
| 801 |
+
llama_print_timings: load time = 6.42 ms
|
| 802 |
+
llama_print_timings: sample time = 8.73 ms / 279 runs ( 0.03 ms per token, 31969.75 tokens per second)
|
| 803 |
+
llama_print_timings: prompt eval time = 1.53 ms / 8 tokens ( 0.19 ms per token, 5239.03 tokens per second)
|
| 804 |
+
llama_print_timings: eval time = 366.91 ms / 278 runs ( 1.32 ms per token, 757.69 tokens per second)
|
| 805 |
+
llama_print_timings: total time = 440.41 ms / 286 tokens
|
| 806 |
+
Log end
|