Update Qwen7B.yaml
Browse files- Qwen7B.yaml +4 -4
Qwen7B.yaml
CHANGED
|
@@ -65,10 +65,10 @@ lora_dropout:
|
|
| 65 |
lora_target_linear: true
|
| 66 |
lora_fan_in_fan_out:
|
| 67 |
|
| 68 |
-
wandb_project: henbane 7b-
|
| 69 |
wandb_entity:
|
| 70 |
wandb_watch:
|
| 71 |
-
wandb_name: henbane 7b-
|
| 72 |
wandb_log_model:
|
| 73 |
|
| 74 |
|
|
@@ -80,8 +80,8 @@ liger_swiglu: true
|
|
| 80 |
liger_fused_linear_cross_entropy: true
|
| 81 |
|
| 82 |
|
| 83 |
-
gradient_accumulation_steps:
|
| 84 |
-
micro_batch_size:
|
| 85 |
num_epochs: 2
|
| 86 |
optimizer: adamw_torch
|
| 87 |
lr_scheduler: cosine
|
|
|
|
| 65 |
lora_target_linear: true
|
| 66 |
lora_fan_in_fan_out:
|
| 67 |
|
| 68 |
+
wandb_project: henbane 7b-attempt2
|
| 69 |
wandb_entity:
|
| 70 |
wandb_watch:
|
| 71 |
+
wandb_name: henbane 7b-attempt2
|
| 72 |
wandb_log_model:
|
| 73 |
|
| 74 |
|
|
|
|
| 80 |
liger_fused_linear_cross_entropy: true
|
| 81 |
|
| 82 |
|
| 83 |
+
gradient_accumulation_steps: 32
|
| 84 |
+
micro_batch_size: 4
|
| 85 |
num_epochs: 2
|
| 86 |
optimizer: adamw_torch
|
| 87 |
lr_scheduler: cosine
|