Upload folder using huggingface_hub
Browse files- README.md +20 -20
- adapter_config.json +5 -5
- adapter_model.safetensors +1 -1
- training_args.bin +1 -1
README.md
CHANGED
@@ -5,12 +5,8 @@ tags:
|
|
5 |
- generated_from_trainer
|
6 |
base_model: meta-llama/Meta-Llama-3-8B-Instruct
|
7 |
model-index:
|
8 |
-
- name: llama-3-8b-claudstruct-
|
9 |
results: []
|
10 |
-
datasets:
|
11 |
-
- Norquinal/claude_multi_instruct_30k
|
12 |
-
language:
|
13 |
-
- en
|
14 |
---
|
15 |
|
16 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
@@ -35,7 +31,7 @@ datasets:
|
|
35 |
type: alpaca
|
36 |
dataset_prepared_path: last_run_prepared
|
37 |
val_set_size: 0.05
|
38 |
-
output_dir: ./outputs/llama-3-8b-claudstruct-
|
39 |
|
40 |
adapter: qlora
|
41 |
lora_model_dir:
|
@@ -58,8 +54,8 @@ wandb_name:
|
|
58 |
wandb_log_model:
|
59 |
|
60 |
gradient_accumulation_steps: 1
|
61 |
-
micro_batch_size:
|
62 |
-
num_epochs:
|
63 |
optimizer: adamw_torch
|
64 |
lr_scheduler: cosine
|
65 |
learning_rate: 0.00001
|
@@ -107,11 +103,11 @@ special_tokens:
|
|
107 |
|
108 |
</details><br>
|
109 |
|
110 |
-
# llama-3-8b-claudstruct-
|
111 |
|
112 |
-
This model is a fine-tuned version of [meta-llama/Meta-Llama-3-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct) on the
|
113 |
It achieves the following results on the evaluation set:
|
114 |
-
- Loss: 1.
|
115 |
|
116 |
## Model description
|
117 |
|
@@ -131,26 +127,30 @@ More information needed
|
|
131 |
|
132 |
The following hyperparameters were used during training:
|
133 |
- learning_rate: 1e-05
|
134 |
-
- train_batch_size:
|
135 |
-
- eval_batch_size:
|
136 |
- seed: 42
|
137 |
- distributed_type: multi-GPU
|
138 |
- num_devices: 2
|
139 |
-
- total_train_batch_size:
|
140 |
-
- total_eval_batch_size:
|
141 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
142 |
- lr_scheduler_type: cosine
|
143 |
- lr_scheduler_warmup_steps: 10
|
144 |
-
- num_epochs:
|
145 |
|
146 |
### Training results
|
147 |
|
148 |
| Training Loss | Epoch | Step | Validation Loss |
|
149 |
|:-------------:|:------:|:----:|:---------------:|
|
150 |
-
| 2.
|
151 |
-
| 1.
|
152 |
-
| 1.
|
153 |
-
| 1.
|
|
|
|
|
|
|
|
|
154 |
|
155 |
|
156 |
### Framework versions
|
|
|
5 |
- generated_from_trainer
|
6 |
base_model: meta-llama/Meta-Llama-3-8B-Instruct
|
7 |
model-index:
|
8 |
+
- name: outputs/llama-3-8b-claudstruct-v3/
|
9 |
results: []
|
|
|
|
|
|
|
|
|
10 |
---
|
11 |
|
12 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
|
|
31 |
type: alpaca
|
32 |
dataset_prepared_path: last_run_prepared
|
33 |
val_set_size: 0.05
|
34 |
+
output_dir: ./outputs/llama-3-8b-claudstruct-v3/
|
35 |
|
36 |
adapter: qlora
|
37 |
lora_model_dir:
|
|
|
54 |
wandb_log_model:
|
55 |
|
56 |
gradient_accumulation_steps: 1
|
57 |
+
micro_batch_size: 8
|
58 |
+
num_epochs: 2
|
59 |
optimizer: adamw_torch
|
60 |
lr_scheduler: cosine
|
61 |
learning_rate: 0.00001
|
|
|
103 |
|
104 |
</details><br>
|
105 |
|
106 |
+
# outputs/llama-3-8b-claudstruct-v3/
|
107 |
|
108 |
+
This model is a fine-tuned version of [meta-llama/Meta-Llama-3-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct) on the None dataset.
|
109 |
It achieves the following results on the evaluation set:
|
110 |
+
- Loss: 1.6226
|
111 |
|
112 |
## Model description
|
113 |
|
|
|
127 |
|
128 |
The following hyperparameters were used during training:
|
129 |
- learning_rate: 1e-05
|
130 |
+
- train_batch_size: 8
|
131 |
+
- eval_batch_size: 8
|
132 |
- seed: 42
|
133 |
- distributed_type: multi-GPU
|
134 |
- num_devices: 2
|
135 |
+
- total_train_batch_size: 16
|
136 |
+
- total_eval_batch_size: 16
|
137 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
138 |
- lr_scheduler_type: cosine
|
139 |
- lr_scheduler_warmup_steps: 10
|
140 |
+
- num_epochs: 2
|
141 |
|
142 |
### Training results
|
143 |
|
144 |
| Training Loss | Epoch | Step | Validation Loss |
|
145 |
|:-------------:|:------:|:----:|:---------------:|
|
146 |
+
| 2.2209 | 0.0007 | 1 | 2.0399 |
|
147 |
+
| 1.7842 | 0.2502 | 341 | 1.6960 |
|
148 |
+
| 1.6914 | 0.5004 | 682 | 1.6590 |
|
149 |
+
| 1.6757 | 0.7506 | 1023 | 1.6414 |
|
150 |
+
| 1.5182 | 1.0007 | 1364 | 1.6319 |
|
151 |
+
| 1.8421 | 1.2509 | 1705 | 1.6264 |
|
152 |
+
| 1.7271 | 1.5011 | 2046 | 1.6237 |
|
153 |
+
| 1.4817 | 1.7513 | 2387 | 1.6226 |
|
154 |
|
155 |
|
156 |
### Framework versions
|
adapter_config.json
CHANGED
@@ -20,13 +20,13 @@
|
|
20 |
"rank_pattern": {},
|
21 |
"revision": null,
|
22 |
"target_modules": [
|
23 |
-
"o_proj",
|
24 |
-
"gate_proj",
|
25 |
-
"q_proj",
|
26 |
-
"k_proj",
|
27 |
"v_proj",
|
|
|
|
|
28 |
"down_proj",
|
29 |
-
"
|
|
|
|
|
30 |
],
|
31 |
"task_type": "CAUSAL_LM",
|
32 |
"use_dora": false,
|
|
|
20 |
"rank_pattern": {},
|
21 |
"revision": null,
|
22 |
"target_modules": [
|
|
|
|
|
|
|
|
|
23 |
"v_proj",
|
24 |
+
"up_proj",
|
25 |
+
"gate_proj",
|
26 |
"down_proj",
|
27 |
+
"o_proj",
|
28 |
+
"k_proj",
|
29 |
+
"q_proj"
|
30 |
],
|
31 |
"task_type": "CAUSAL_LM",
|
32 |
"use_dora": false,
|
adapter_model.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 83945296
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:87a0ecab10e267e31e2d676cb9219607001f5b9c2e4496c5c3ded514352d594d
|
3 |
size 83945296
|
training_args.bin
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 6456
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:899f1662c0af7238f2da34f9cfa528d22053871728445d51705cf917a4b82a05
|
3 |
size 6456
|