Downloads · 30 days
453
22% of all-time downloads
RichardErkhov/yujiepan_-_llama-3-tiny-random-gguf
yujiepan_-_llama-3-tiny-random-gguf is a machine learning model from RichardErkhov. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
Downloads · 30 days
453
22% of all-time downloads
All-time downloads
2K
Public
Repo size
7.8 MB
Likes
0
Public
Click a slice to open those files.
.gguf172 MB · 100%
From the Hugging Face model README
Quantization made by Richard Erkhov.
llama-3-tiny-random - GGUF
| Name | Quant method | Size |
|---|---|---|
| llama-3-tiny-random.Q2_K.gguf | Q2_K | 0.01GB |
| llama-3-tiny-random.IQ3_XS.gguf | IQ3_XS | 0.01GB |
| llama-3-tiny-random.IQ3_S.gguf | IQ3_S | 0.01GB |
| llama-3-tiny-random.Q3_K_S.gguf | Q3_K_S | 0.01GB |
| llama-3-tiny-random.IQ3_M.gguf | IQ3_M | 0.01GB |
| llama-3-tiny-random.Q3_K.gguf | Q3_K | 0.01GB |
| llama-3-tiny-random.Q3_K_M.gguf | Q3_K_M | 0.01GB |
| llama-3-tiny-random.Q3_K_L.gguf | Q3_K_L | 0.01GB |
| llama-3-tiny-random.IQ4_XS.gguf | IQ4_XS | 0.01GB |
| llama-3-tiny-random.Q4_0.gguf | Q4_0 | 0.01GB |
| llama-3-tiny-random.IQ4_NL.gguf | IQ4_NL | 0.01GB |
| llama-3-tiny-random.Q4_K_S.gguf | Q4_K_S | 0.01GB |
| llama-3-tiny-random.Q4_K.gguf | Q4_K | 0.01GB |
| llama-3-tiny-random.Q4_K_M.gguf | Q4_K_M | 0.01GB |
| llama-3-tiny-random.Q4_1.gguf | Q4_1 | 0.01GB |
| llama-3-tiny-random.Q5_0.gguf | Q5_0 | 0.01GB |
| llama-3-tiny-random.Q5_K_S.gguf | Q5_K_S | 0.01GB |
| llama-3-tiny-random.Q5_K.gguf | Q5_K | 0.01GB |
| llama-3-tiny-random.Q5_K_M.gguf | Q5_K_M | 0.01GB |
| llama-3-tiny-random.Q5_1.gguf | Q5_1 | 0.01GB |
| llama-3-tiny-random.Q6_K.gguf | Q6_K | 0.01GB |
| llama-3-tiny-random.Q8_0.gguf | Q8_0 | 0.01GB |
library_name: transformers pipeline_tag: text-generation inference: true widget:
This model is randomly initialized, using the config from meta-llama/Meta-Llama-3-8B-Instruct but with smaller size. Note the model is in bfloat16.
"yujiepan/llama-3-tiny-random" and "yujiepan/meta-llama-3-tiny-random" shares exactly the same files except the repo name.
Codes:
import transformers
import torch
import os
from huggingface_hub import create_repo, upload_folder
import accelerate
source_model_id = 'meta-llama/Meta-Llama-3-8B-Instruct'
save_path = '/tmp/yujiepan/meta-llama-3-tiny-random'
repo_id = 'yujiepan/meta-llama-3-tiny-random'
os.system(f'rm -rf {save_path}')
config = transformers.AutoConfig.from_pretrained(
source_model_id,
trust_remote_code=True,
)
config._name_or_path = source_model_id
config.hidden_size = 4
config.intermediate_size = 14
config.num_attention_heads = 2
config.num_key_value_heads = 1
config.num_hidden_layers = 2
config.torch_dtype = "bfloat16"
model = transformers.AutoModelForCausalLM.from_config(
config,
trust_remote_code=True,
)
with accelerate.init_empty_weights():
model.generation_config = transformers.AutoModelForCausalLM.from_pretrained(source_model_id).generation_config
model = model.to(torch.bfloat16)
model.save_pretrained(save_path)
tokenizer = transformers.AutoTokenizer.from_pretrained(
source_model_id,
trust_remote_code=True,
)
tokenizer.save_pretrained(save_path)
model.float().generate(torch.tensor([[1, 2, 3]]).long(), max_length=16)
os.system(f'ls -alh {save_path}')
# os.system(f'rm -rf {save_path}/model.safetensors')
create_repo(repo_id, exist_ok=True)
upload_folder(repo_id='yujiepan/meta-llama-3-tiny-random', folder_path=save_path)
upload_folder(repo_id='yujiepan/llama-3-tiny-random', folder_path=save_path)