Update README.md
Browse files
README.md
CHANGED
@@ -16,7 +16,66 @@ should probably proofread and complete it, then remove this comment. -->
|
|
16 |
This model is a fine-tuned version of [openlm-research/open_llama_3b_v2](https://huggingface.co/openlm-research/open_llama_3b_v2) on the None dataset.
|
17 |
|
18 |
## Model description
|
19 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
20 |
More information needed
|
21 |
|
22 |
## Intended uses & limitations
|
|
|
16 |
This model is a fine-tuned version of [openlm-research/open_llama_3b_v2](https://huggingface.co/openlm-research/open_llama_3b_v2) on the None dataset.
|
17 |
|
18 |
## Model description
|
19 |
+
training_arguments = TrainingArguments(
|
20 |
+
per_device_train_batch_size=8,
|
21 |
+
num_train_epochs=10,
|
22 |
+
learning_rate=3e-5,
|
23 |
+
gradient_accumulation_steps=2,
|
24 |
+
optim="adamw_hf",
|
25 |
+
fp16=True,
|
26 |
+
logging_steps=1,
|
27 |
+
# debug=True,
|
28 |
+
output_dir="/kaggle/Tatvajsh/Lllama_AHS_V_7.0/"
|
29 |
+
# warmup_steps=100,
|
30 |
+
)
|
31 |
+
|
32 |
+
trainer = SFTTrainer(
|
33 |
+
model=model,
|
34 |
+
tokenizer=tokenizer,
|
35 |
+
train_dataset=dataset,
|
36 |
+
dataset_text_field="text",
|
37 |
+
peft_config=lora_config,
|
38 |
+
max_seq_length=512,
|
39 |
+
args=training_arguments,
|
40 |
+
# packing=True,#change
|
41 |
+
)
|
42 |
+
|
43 |
+
trainer.train()
|
44 |
+
|
45 |
+
|
46 |
+
EPOCHS=[30-50]
|
47 |
+
|
48 |
+
|
49 |
+
from peft import LoraConfig, get_peft_model
|
50 |
+
|
51 |
+
lora_config = LoraConfig(
|
52 |
+
r=16,
|
53 |
+
lora_alpha=64,
|
54 |
+
target_modules=['base_layer','gate_proj', 'v_proj','up_proj','down_proj','q_proj','k_proj','o_proj'],
|
55 |
+
lora_dropout=0.05,
|
56 |
+
bias="none",
|
57 |
+
task_type="CAUSAL_LM"
|
58 |
+
)
|
59 |
+
|
60 |
+
|
61 |
+
def generate_prompt(row) -> str:
|
62 |
+
prompt=f"""
|
63 |
+
Below is an instruction that describes a task. Write a response that appropriately completes the request.
|
64 |
+
|
65 |
+
### Instruction:
|
66 |
+
|
67 |
+
{row['Instruction']}
|
68 |
+
|
69 |
+
### Response:
|
70 |
+
|
71 |
+
{row['Answer']}
|
72 |
+
|
73 |
+
### End
|
74 |
+
"""
|
75 |
+
return prompt
|
76 |
+
|
77 |
+
|
78 |
+
WHEN THE TRAINING LOSS IN NOT REDUCING THEN TRY SETTING FOR LESSER VALUE OF LEARNING RATE I.E. 2E-5 TO 3E-5,ETC.
|
79 |
More information needed
|
80 |
|
81 |
## Intended uses & limitations
|