Update README.md
Browse files
README.md
CHANGED
@@ -17,6 +17,34 @@ tags:
|
|
17 |
- **License:** apache-2.0
|
18 |
- **Finetuned from model :** unsloth/gemma-2-9b-bnb-4bit
|
19 |
|
20 |
-
This gemma2
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
21 |
|
22 |
[<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)
|
|
|
17 |
- **License:** apache-2.0
|
18 |
- **Finetuned from model :** unsloth/gemma-2-9b-bnb-4bit
|
19 |
|
20 |
+
This is gemma2 trained on semeval restaurant data 2014 by Harsh Jain.
|
21 |
+
|
22 |
+
`trainer = SFTTrainer(
|
23 |
+
model = model,
|
24 |
+
tokenizer = tokenizer,
|
25 |
+
train_dataset = dataset,
|
26 |
+
dataset_text_field = "text_new",
|
27 |
+
max_seq_length = max_seq_length,
|
28 |
+
dataset_num_proc = 2,
|
29 |
+
packing = False, # Can make training 5x faster for short sequences.
|
30 |
+
args = TrainingArguments(
|
31 |
+
per_device_train_batch_size = 2,
|
32 |
+
gradient_accumulation_steps = 4,
|
33 |
+
warmup_steps = 6, #Previous 5
|
34 |
+
#num_train_epochs = 1, # Set this for 1 full training run.
|
35 |
+
max_steps = 60,
|
36 |
+
#learning_rate = 2e-4,
|
37 |
+
learning_rate = 1e-4,
|
38 |
+
fp16 = not is_bfloat16_supported(),
|
39 |
+
bf16 = is_bfloat16_supported(),
|
40 |
+
logging_steps = 1,
|
41 |
+
optim = "adamw_8bit",
|
42 |
+
weight_decay = 0.01,
|
43 |
+
lr_scheduler_type = "linear",
|
44 |
+
seed = 3407,
|
45 |
+
output_dir = "./tensorLog",
|
46 |
+
report_to="wandb"
|
47 |
+
),
|
48 |
+
)`
|
49 |
|
50 |
[<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)
|