File size: 3,495 Bytes
1b7df7f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
---
library_name: transformers
license: apache-2.0
base_model: Qwen/Qwen2.5-Coder-14B
tags:
- generated_from_trainer
datasets:
- winglian/gpumode-py2triton-reasoning
model-index:
- name: outputs/out-fft
  results: []
---

<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->

[<img src="https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/axolotl-ai-cloud/axolotl)
<details><summary>See axolotl config</summary>

axolotl version: `0.8.0`
```yaml
base_model: Qwen/Qwen2.5-Coder-14B
strict: false       

plugins:             
  - axolotl.integrations.cut_cross_entropy.CutCrossEntropyPlugin
  - axolotl.integrations.liger.LigerPlugin
                                                               
cut_cross_entropy: true
  # liger_rope: true
liger_rms_norm: true
liger_layer_norm: true

# gemma3 doesn't seem to play nice with ddp
ddp_find_unused_parameters: true
                                                               
chat_template: qwen_25
datasets:           
  - path: winglian/gpumode-py2triton-reasoning
    type: chat_template
                                                               
dataset_prepared_path: last_run_prepared
val_set_size: 0.0
output_dir: ./outputs/out-fft
save_safetensors: true
save_only_model: true

sequence_len: 32768
sample_packing: true 
pad_to_sequence_len: true
                                                               
sequence_parallel_degree: 1

# unfrozen_parameters:
#   - language_model.model

wandb_project: qwen25-kernel-llm
wandb_entity: axolotl-ai
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 1
micro_batch_size: 3
num_epochs: 3
optimizer: adamw_8bit
lr_scheduler: rex
learning_rate: 3.0e-6
lr_groups:
  - name: embeddings
    lr: 3.0e-5
    modules:
      - lm_head
      - embed_tokens

train_on_inputs: false
group_by_length: false
bf16: true
tf32: true

gradient_checkpointing: true
gradient_checkpointing_kwargs:
  use_reentrant: false
logging_steps: 1
flash_attention: true

warmup_ratio: 0.1
evals_per_epoch: 1
saves_per_epoch: 1
  # deepspeed: deepspeed_configs/zero1.json
weight_decay: 0.0
deepspeed: deepspeed_configs/zero1.json
tokens:
 - <think>
 - </think>
special_tokens:
  eos_token: <|im_end|>
fix_untrained_tokens:
  - 151665
  - 151666


```

</details><br>

# outputs/out-fft

This model is a fine-tuned version of [Qwen/Qwen2.5-Coder-14B](https://huggingface.co/Qwen/Qwen2.5-Coder-14B) on the winglian/gpumode-py2triton-reasoning dataset.

## Model description

More information needed

## Intended uses & limitations

More information needed

## Training and evaluation data

More information needed

## Training procedure

### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 3e-06
- train_batch_size: 3
- eval_batch_size: 3
- seed: 42
- distributed_type: multi-GPU
- num_devices: 8
- total_train_batch_size: 24
- total_eval_batch_size: 24
- optimizer: Use OptimizerNames.ADAMW_8BIT with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 27
- num_epochs: 3.0

### Training results



### Framework versions

- Transformers 4.51.3
- Pytorch 2.6.0+cu124
- Datasets 3.5.0
- Tokenizers 0.21.1