afrias5 commited on
Commit
b951e69
1 Parent(s): a64e663

End of training

Browse files
Files changed (1) hide show
  1. README.md +6 -6
README.md CHANGED
@@ -34,8 +34,8 @@ datasets:
34
  dataset_prepared_path: FinUpTagsNoTestNoExNewCodeLlama
35
  val_set_size: 0
36
  output_dir: models/codellama34bTestL4
37
- # lora_model_dir: models/codellamaTestLora4
38
- # auto_resume_from_checkpoints: true
39
  sequence_len: 4096
40
  sample_packing: true
41
  pad_to_sequence_len: true
@@ -54,12 +54,12 @@ wandb_project: 'codellamaFeed'
54
  wandb_entity:
55
  wandb_watch:
56
  wandb_run_id:
57
- wandb_name: '34bLora4'
58
  wandb_log_model:
59
 
60
  gradient_accumulation_steps: 4
61
  micro_batch_size: 1
62
- num_epochs: 4
63
  optimizer: adamw_torch
64
  lr_scheduler: cosine
65
  learning_rate: 0.0002
@@ -98,7 +98,7 @@ special_tokens:
98
 
99
  </details><br>
100
 
101
- [<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="200" height="32"/>](https://wandb.ai/afrias5/codellamaFeed/runs/qazhmiqa)
102
  # CodeLlamaL4
103
 
104
  This model is a fine-tuned version of [codellama/CodeLlama-34b-Python-hf](https://huggingface.co/codellama/CodeLlama-34b-Python-hf) on the None dataset.
@@ -132,7 +132,7 @@ The following hyperparameters were used during training:
132
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
133
  - lr_scheduler_type: cosine
134
  - lr_scheduler_warmup_steps: 10
135
- - num_epochs: 4
136
 
137
  ### Training results
138
 
 
34
  dataset_prepared_path: FinUpTagsNoTestNoExNewCodeLlama
35
  val_set_size: 0
36
  output_dir: models/codellama34bTestL4
37
+ lora_model_dir: models/codellama34bTestL4/checkpoint-40
38
+ auto_resume_from_checkpoints: true
39
  sequence_len: 4096
40
  sample_packing: true
41
  pad_to_sequence_len: true
 
54
  wandb_entity:
55
  wandb_watch:
56
  wandb_run_id:
57
+ wandb_name: '34bLora4'
58
  wandb_log_model:
59
 
60
  gradient_accumulation_steps: 4
61
  micro_batch_size: 1
62
+ num_epochs: 8
63
  optimizer: adamw_torch
64
  lr_scheduler: cosine
65
  learning_rate: 0.0002
 
98
 
99
  </details><br>
100
 
101
+ [<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="200" height="32"/>](https://wandb.ai/afrias5/codellamaFeed/runs/81byeenq)
102
  # CodeLlamaL4
103
 
104
  This model is a fine-tuned version of [codellama/CodeLlama-34b-Python-hf](https://huggingface.co/codellama/CodeLlama-34b-Python-hf) on the None dataset.
 
132
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
133
  - lr_scheduler_type: cosine
134
  - lr_scheduler_warmup_steps: 10
135
+ - num_epochs: 8
136
 
137
  ### Training results
138