Hugging Face's logo

Casper0508
/
MSc_llama3_finetuned_model_secondData

PEFT
TensorBoard
Safetensors
Generated from Trainer
Model card Files Metrics Community
Casper0508 commited on
Commit
ebed6c7
•
1 Parent(s): 78294d5

End of training

Browse files
Files changed (6) hide show
  1. README.md +28 -28
  2. adapter_config.json +1 -1
  3. adapter_model.safetensors +1 -1
  4. emissions.csv +1 -1
  5. runs/Jul17_21-03-56_msc-modeltrain-pod/events.out.tfevents.1721250240.msc-modeltrain-pod.3569.0 +3 -0
  6. training_args.bin +1 -1
README.md CHANGED
@@ -16,7 +16,7 @@ should probably proofread and complete it, then remove this comment. -->
16
 
17
  This model is a fine-tuned version of [meta-llama/Meta-Llama-3-8B](https://huggingface.co/meta-llama/Meta-Llama-3-8B) on the None dataset.
18
  It achieves the following results on the evaluation set:
19
- - Loss: 0.7565
20
 
21
  ## Model description
22
 
@@ -50,10 +50,10 @@ The following `bitsandbytes` quantization config was used during training:
50
 
51
  The following hyperparameters were used during training:
52
  - learning_rate: 5e-05
53
- - train_batch_size: 32
54
  - eval_batch_size: 8
55
  - seed: 42
56
- - gradient_accumulation_steps: 2
57
  - total_train_batch_size: 64
58
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
59
  - lr_scheduler_type: cosine
@@ -64,31 +64,31 @@ The following hyperparameters were used during training:
64
 
65
  | Training Loss | Epoch | Step | Validation Loss |
66
  |:-------------:|:-----:|:----:|:---------------:|
67
- | 3.7475 | 1.33 | 10 | 3.3981 |
68
- | 2.804 | 2.67 | 20 | 2.0339 |
69
- | 1.5726 | 4.0 | 30 | 1.2647 |
70
- | 1.1893 | 5.33 | 40 | 1.1114 |
71
- | 1.0463 | 6.67 | 50 | 1.0339 |
72
- | 0.9701 | 8.0 | 60 | 0.9736 |
73
- | 0.9065 | 9.33 | 70 | 0.9295 |
74
- | 0.8548 | 10.67 | 80 | 0.9125 |
75
- | 0.7974 | 12.0 | 90 | 0.8830 |
76
- | 0.7462 | 13.33 | 100 | 0.8595 |
77
- | 0.7056 | 14.67 | 110 | 0.8447 |
78
- | 0.6521 | 16.0 | 120 | 0.8119 |
79
- | 0.5957 | 17.33 | 130 | 0.7680 |
80
- | 0.5399 | 18.67 | 140 | 0.7439 |
81
- | 0.486 | 20.0 | 150 | 0.7158 |
82
- | 0.4572 | 21.33 | 160 | 0.7180 |
83
- | 0.4422 | 22.67 | 170 | 0.7232 |
84
- | 0.4255 | 24.0 | 180 | 0.7245 |
85
- | 0.421 | 25.33 | 190 | 0.7257 |
86
- | 0.4129 | 26.67 | 200 | 0.7458 |
87
- | 0.4065 | 28.0 | 210 | 0.7558 |
88
- | 0.4053 | 29.33 | 220 | 0.7577 |
89
- | 0.3984 | 30.67 | 230 | 0.7583 |
90
- | 0.3987 | 32.0 | 240 | 0.7567 |
91
- | 0.4005 | 33.33 | 250 | 0.7565 |
92
 
93
 
94
  ### Framework versions
 
16
 
17
  This model is a fine-tuned version of [meta-llama/Meta-Llama-3-8B](https://huggingface.co/meta-llama/Meta-Llama-3-8B) on the None dataset.
18
  It achieves the following results on the evaluation set:
19
+ - Loss: 0.7284
20
 
21
  ## Model description
22
 
 
50
 
51
  The following hyperparameters were used during training:
52
  - learning_rate: 5e-05
53
+ - train_batch_size: 8
54
  - eval_batch_size: 8
55
  - seed: 42
56
+ - gradient_accumulation_steps: 8
57
  - total_train_batch_size: 64
58
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
59
  - lr_scheduler_type: cosine
 
64
 
65
  | Training Loss | Epoch | Step | Validation Loss |
66
  |:-------------:|:-----:|:----:|:---------------:|
67
+ | 3.7573 | 1.36 | 10 | 3.3722 |
68
+ | 2.8144 | 2.71 | 20 | 2.0325 |
69
+ | 1.5867 | 4.07 | 30 | 1.2586 |
70
+ | 1.1875 | 5.42 | 40 | 1.0898 |
71
+ | 1.0487 | 6.78 | 50 | 1.0168 |
72
+ | 0.9716 | 8.14 | 60 | 0.9688 |
73
+ | 0.9134 | 9.49 | 70 | 0.9304 |
74
+ | 0.8505 | 10.85 | 80 | 0.8940 |
75
+ | 0.7963 | 12.2 | 90 | 0.8640 |
76
+ | 0.7497 | 13.56 | 100 | 0.8379 |
77
+ | 0.7067 | 14.92 | 110 | 0.8080 |
78
+ | 0.6484 | 16.27 | 120 | 0.7751 |
79
+ | 0.5945 | 17.63 | 130 | 0.7385 |
80
+ | 0.5349 | 18.98 | 140 | 0.6990 |
81
+ | 0.4854 | 20.34 | 150 | 0.6956 |
82
+ | 0.4729 | 21.69 | 160 | 0.6990 |
83
+ | 0.4496 | 23.05 | 170 | 0.6885 |
84
+ | 0.4405 | 24.41 | 180 | 0.7089 |
85
+ | 0.435 | 25.76 | 190 | 0.7215 |
86
+ | 0.4243 | 27.12 | 200 | 0.7162 |
87
+ | 0.418 | 28.47 | 210 | 0.7205 |
88
+ | 0.4156 | 29.83 | 220 | 0.7296 |
89
+ | 0.4153 | 31.19 | 230 | 0.7275 |
90
+ | 0.4146 | 32.54 | 240 | 0.7292 |
91
+ | 0.4094 | 33.9 | 250 | 0.7284 |
92
 
93
 
94
  ### Framework versions
adapter_config.json CHANGED
@@ -8,7 +8,7 @@
8
  "layers_pattern": null,
9
  "layers_to_transform": null,
10
  "lora_alpha": 64,
11
- "lora_dropout": 0.1,
12
  "modules_to_save": null,
13
  "peft_type": "LORA",
14
  "r": 32,
 
8
  "layers_pattern": null,
9
  "layers_to_transform": null,
10
  "lora_alpha": 64,
11
+ "lora_dropout": 0.2,
12
  "modules_to_save": null,
13
  "peft_type": "LORA",
14
  "r": 32,
adapter_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:2866fc3eb9f31ea49c5c5d4a806c99f33cb8132c0df139c0e58730f4d260dd0e
3
  size 75523312
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dcd4cce18bd06fc14145a73cf25cd3cab8a326057bc4fba56c7d5e68c2670d70
3
  size 75523312
emissions.csv CHANGED
@@ -1,2 +1,2 @@
1
  timestamp,experiment_id,project_name,duration,emissions,energy_consumed,country_name,country_iso_code,region,on_cloud,cloud_provider,cloud_region
2
- 2024-07-17T20:28:57,58cd8dfb-7bcb-45db-b0fc-26c9ef823a21,codecarbon,409.7803256511688,0.033807288347631255,0.050300301789424406,United Kingdom,GBR,scotland,N,,
 
1
  timestamp,experiment_id,project_name,duration,emissions,energy_consumed,country_name,country_iso_code,region,on_cloud,cloud_provider,cloud_region
2
+ 2024-07-17T21:26:16,875bb1fb-55b3-4654-806d-76f2ddaca959,codecarbon,1335.8051824569702,0.07996877307198384,0.1189818413677245,United Kingdom,GBR,scotland,N,,
runs/Jul17_21-03-56_msc-modeltrain-pod/events.out.tfevents.1721250240.msc-modeltrain-pod.3569.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:84044fa5e8f3c3e61b99b3d2f370c81abeb4740e93268e34f7384e7720a962c2
3
+ size 17467
training_args.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:4cb836eeb3e02d59f39d4ba4165989382c380cfca1edbf41c54479268e5dd6cf
3
  size 4984
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1f116e75b2c681e3d8ef6631a40874a06edfdccef281260daae484fdda3bebfd
3
  size 4984