CompactAI commited on
Commit
45a8066
Β·
verified Β·
1 Parent(s): c7e87e5

Update pruned model - 8 files

Browse files
Files changed (4) hide show
  1. README.md +16 -10
  2. comparison_graph.png +0 -0
  3. model.safetensors +1 -1
  4. tokenizer.json +1 -1
README.md CHANGED
@@ -11,24 +11,30 @@ pipeline_tag: text-generation
11
 
12
  # LFM2.5-1.2B-Base-python-aggressive
13
 
14
- > 🎯 **PYTHON-optimized** | πŸ“¦ **Aggressive** pruning | ⚑ **1% weights pruned**
15
 
16
  This model is a **aggressively pruned** version of [LiquidAI/LFM2.5-1.2B-Base](https://huggingface.co/LiquidAI/LFM2.5-1.2B-Base).
17
 
 
 
 
 
 
 
18
  ## Performance Comparison
19
 
20
  | Category | Original | Pruned | Change |
21
  |----------|----------|--------|--------|
22
  | **Python** | 0.0% | 0.0% ⭐ | β†’ |
23
- | Html | 16.7% | 16.7% | β†’ |
24
- | Trivia | 91.7% | 91.7% | β†’ |
25
- | Math | 75.0% | 75.0% | β†’ |
26
- | Reasoning | 41.7% | 41.7% | β†’ |
27
- | Medical | 66.7% | 66.7% | β†’ |
28
- | Linux | 16.7% | 16.7% | β†’ |
29
- | Writing | 33.3% | 33.3% | β†’ |
30
 
31
- **Average**: 42.7% β†’ 42.7% (+0.0%)
32
 
33
 
34
 
@@ -54,7 +60,7 @@ print(tokenizer.decode(outputs[0], skip_special_tokens=True))
54
  | Base Model | [LiquidAI/LFM2.5-1.2B-Base](https://huggingface.co/LiquidAI/LFM2.5-1.2B-Base) |
55
  | Specialization | Python |
56
  | Prune Mode | Aggressive |
57
- | Weight Reduction | 1% weights pruned |
58
 
59
  ## License
60
 
 
11
 
12
  # LFM2.5-1.2B-Base-python-aggressive
13
 
14
+ > **PYTHON-optimized** | **Aggressive** pruning | **35% weights pruned**
15
 
16
  This model is a **aggressively pruned** version of [LiquidAI/LFM2.5-1.2B-Base](https://huggingface.co/LiquidAI/LFM2.5-1.2B-Base).
17
 
18
+
19
+
20
+ > **Note:** Minimal quality drop detected. The Wanda pruning algorithm effectively identifies and removes less important weights while preserving model capability.
21
+
22
+
23
+
24
  ## Performance Comparison
25
 
26
  | Category | Original | Pruned | Change |
27
  |----------|----------|--------|--------|
28
  | **Python** | 0.0% | 0.0% ⭐ | β†’ |
29
+ | Html | 0.0% | 0.0% | β†’ |
30
+ | Trivia | 15.0% | 15.0% | β†’ |
31
+ | Math | 15.0% | 15.0% | β†’ |
32
+ | Reasoning | 0.0% | 0.0% | β†’ |
33
+ | Medical | 20.0% | 20.0% | β†’ |
34
+ | Linux | 30.0% | 25.0% | ↓ 5.0% |
35
+ | Writing | 5.0% | 5.0% | β†’ |
36
 
37
+ **Average**: 10.6% -> 10.0% (-0.6%)
38
 
39
 
40
 
 
60
  | Base Model | [LiquidAI/LFM2.5-1.2B-Base](https://huggingface.co/LiquidAI/LFM2.5-1.2B-Base) |
61
  | Specialization | Python |
62
  | Prune Mode | Aggressive |
63
+ | Weight Reduction | 35% weights pruned |
64
 
65
  ## License
66
 
comparison_graph.png CHANGED
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:f037cc94307b2672c8e1b29170435431e19053ff0740015c5e301ac0d94c203f
3
  size 2340697784
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:878bc57aaab79e1f145b2c1fcbf8ebea9547d45d344ee3a5d6e5fafda64d6150
3
  size 2340697784
tokenizer.json CHANGED
@@ -2,7 +2,7 @@
2
  "version": "1.0",
3
  "truncation": {
4
  "direction": "Right",
5
- "max_length": 127850,
6
  "strategy": "LongestFirst",
7
  "stride": 0
8
  },
 
2
  "version": "1.0",
3
  "truncation": {
4
  "direction": "Right",
5
+ "max_length": 126976,
6
  "strategy": "LongestFirst",
7
  "stride": 0
8
  },