entrenar 0.7.13

Training & Optimization library with autograd, LoRA, quantization, and model merging
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85

//! Gradual pruning schedule methods.

#[cfg(test)]
mod proptests;
#[cfg(test)]
mod tests;

use super::PruningSchedule;

impl PruningSchedule {
    /// Compute the target sparsity at a given training step for Gradual schedule.
    pub(super) fn gradual_sparsity_at_step(
        start_step: usize,
        end_step: usize,
        initial_sparsity: f32,
        final_sparsity: f32,
        step: usize,
    ) -> f32 {
        if step < start_step {
            initial_sparsity
        } else if step >= end_step {
            final_sparsity
        } else {
            let progress = (step - start_step) as f32 / (end_step - start_step) as f32;
            initial_sparsity + progress * (final_sparsity - initial_sparsity)
        }
    }

    /// Check if pruning should be applied at this step for Gradual schedule.
    pub(super) fn gradual_should_prune_at_step(
        start_step: usize,
        end_step: usize,
        frequency: usize,
        step: usize,
    ) -> bool {
        if step < start_step || step > end_step {
            return false;
        }
        if frequency == 0 {
            return step == start_step;
        }
        (step - start_step).is_multiple_of(frequency)
    }

    /// Get the total number of pruning operations for Gradual schedule.
    pub(super) fn gradual_num_pruning_steps(
        start_step: usize,
        end_step: usize,
        frequency: usize,
    ) -> usize {
        if frequency == 0 {
            1
        } else {
            (end_step - start_step) / frequency + 1
        }
    }

    /// Validate Gradual schedule.
    pub(super) fn gradual_validate(
        start_step: usize,
        end_step: usize,
        initial_sparsity: f32,
        final_sparsity: f32,
    ) -> Result<(), String> {
        if end_step <= start_step {
            return Err(format!(
                "end_step ({end_step}) must be greater than start_step ({start_step})"
            ));
        }
        if !(0.0..=1.0).contains(&initial_sparsity) {
            return Err(format!(
                "initial_sparsity ({initial_sparsity}) must be between 0.0 and 1.0"
            ));
        }
        if !(0.0..=1.0).contains(&final_sparsity) {
            return Err(format!("final_sparsity ({final_sparsity}) must be between 0.0 and 1.0"));
        }
        Ok(())
    }

    /// Check if Gradual pruning has completed.
    pub(super) fn gradual_is_complete(end_step: usize, step: usize) -> bool {
        step > end_step
    }
}