Hugging Face's logo

Pelochus
/
ezrkllm-collection

Text Generation
RKLLM
rockchip
rk3588
text-generation-inference
Model card Files
xet
Community
1
Pelochus commited on
Commit
9479cdb
·
verified ·
1 Parent(s): 703a738

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +12 -11
README.md CHANGED
@@ -18,17 +18,18 @@ Check the main repo on GitHub for how to install and use: https://github.com/Pel
18
  Before running any LLM, take into account that the required RAM is between 1.5-3 times the model size (this is an estimation, haven't done extensive testing yet).
19
 
20
  Right now, only converted the following models:
21
- | LLM | Parameters | Link |
22
- | --------------------- | ----------- | ---------------------------------------------------------- |
23
- | Qwen Chat | 1.8B | https://huggingface.co/Pelochus/qwen-1_8B-rk3588 |
24
- | Gemma | 2B | https://huggingface.co/Pelochus/gemma-2b-rk3588 |
25
- | Microsoft Phi-2 | 2.7B | https://huggingface.co/Pelochus/phi-2-rk3588 |
26
- | Microsoft Phi-3 Mini | 3.8B | https://huggingface.co/Pelochus/phi-3-mini-rk3588 |
27
- | Llama 2 7B | 7B | https://huggingface.co/Pelochus/llama2-chat-7b-hf-rk3588 |
28
- | Llama 2 13B | 13B | https://huggingface.co/Pelochus/llama2-chat-13b-hf-rk3588 |
29
- | TinyLlama v1 | 1.1B | https://huggingface.co/Pelochus/tinyllama-v1-rk3588 |
30
- | Qwen 1.5 Chat | 4B | https://huggingface.co/Pelochus/qwen1.5-chat-4B-rk3588 |
31
- | Qwen 2 | 1.5B | https://huggingface.co/Pelochus/qwen2-1_5B-rk3588 |
 
32
 
33
  Llama 2 was converted using Azure servers.
34
  For reference, converting Phi-2 peaked at about 15 GBs of RAM + 25 GBs of swap (counting OS, but that was using about 2 GBs max).
 
18
  Before running any LLM, take into account that the required RAM is between 1.5-3 times the model size (this is an estimation, haven't done extensive testing yet).
19
 
20
  Right now, only converted the following models:
21
+ | LLM | Parameters | Link |
22
+ | --------------------- | ----------- | ------------------------------------------------------------- |
23
+ | Qwen 2 | 1.5B | https://huggingface.co/Pelochus/deepseek-R1-distill-qwen-1.5B |
24
+ | Qwen Chat | 1.8B | https://huggingface.co/Pelochus/qwen-1_8B-rk3588 |
25
+ | Gemma | 2B | https://huggingface.co/Pelochus/gemma-2b-rk3588 |
26
+ | Microsoft Phi-2 | 2.7B | https://huggingface.co/Pelochus/phi-2-rk3588 |
27
+ | Microsoft Phi-3 Mini | 3.8B | https://huggingface.co/Pelochus/phi-3-mini-rk3588 |
28
+ | Llama 2 7B | 7B | https://huggingface.co/Pelochus/llama2-chat-7b-hf-rk3588 |
29
+ | Llama 2 13B | 13B | https://huggingface.co/Pelochus/llama2-chat-13b-hf-rk3588 |
30
+ | TinyLlama v1 | 1.1B | https://huggingface.co/Pelochus/tinyllama-v1-rk3588 |
31
+ | Qwen 1.5 Chat | 4B | https://huggingface.co/Pelochus/qwen1.5-chat-4B-rk3588 |
32
+ | Qwen 2 | 1.5B | https://huggingface.co/Pelochus/qwen2-1_5B-rk3588 |
33
 
34
  Llama 2 was converted using Azure servers.
35
  For reference, converting Phi-2 peaked at about 15 GBs of RAM + 25 GBs of swap (counting OS, but that was using about 2 GBs max).