FSI Edge commited on
Commit
f03f789
·
1 Parent(s): 66d5feb

Add YAML metadata and laptop training instructions

Browse files
Files changed (1) hide show
  1. README.md +21 -1
README.md CHANGED
@@ -1,3 +1,16 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  # FSI_Edge: From-Scratch Novel Architecture Coding Model
2
 
3
  A tiny but capable code generation model trained from scratch on ARM CPU, with a novel DNA-inspired architecture.
@@ -24,12 +37,19 @@ A tiny but capable code generation model trained from scratch on ARM CPU, with a
24
  ## Quick Start
25
 
26
  ```bash
 
 
 
27
  pip install -r requirements.txt
28
 
29
  # Train on CPU
30
  python training/run_cpu.py --model-size 4K --steps 1000
31
 
32
- # Or on Colab T4 GPU
 
 
 
 
33
  # Upload scripts/fsi_edge_colab.ipynb to Google Colab
34
  ```
35
 
 
1
+ ---
2
+ license: mit
3
+ language:
4
+ - en
5
+ pipeline_tag: text-generation
6
+ tags:
7
+ - code-generation
8
+ - from-scratch
9
+ - novel-architecture
10
+ - helix-memory
11
+ - cpu-training
12
+ ---
13
+
14
  # FSI_Edge: From-Scratch Novel Architecture Coding Model
15
 
16
  A tiny but capable code generation model trained from scratch on ARM CPU, with a novel DNA-inspired architecture.
 
37
  ## Quick Start
38
 
39
  ```bash
40
+ # Clone from HuggingFace
41
+ git clone https://huggingface.co/FerrellSyntheticIntelligence/FSI-Edge
42
+ cd FSI-Edge
43
  pip install -r requirements.txt
44
 
45
  # Train on CPU
46
  python training/run_cpu.py --model-size 4K --steps 1000
47
 
48
+ # Resume training from checkpoint (step 4132)
49
+ python training/run_cpu.py --model-size 4K --steps 10000 \
50
+ --resume checkpoints/cpu_ckpt_004132.pt --lr 2e-4
51
+
52
+ # Or on Colab T4 GPU (100x faster)
53
  # Upload scripts/fsi_edge_colab.ipynb to Google Colab
54
  ```
55