Training in progress, step 40000

README.md

@@ -78,13 +78,13 @@ GPT2LMHeadModel(
 
 # Resource Usage
 
-- Max Train VRAM Use: 15.7121 GB
-- Available VRAM: 23.6497 GB
+- Max Train VRAM Use: 15.7128 GB
+- Available VRAM: 23.6429 GB
 - GPUs: 
   - 1x NVIDIA GeForce RTX 4090
-- CPUs: 32
-- CPU Memory: 125.6976 GB
-- CPU Memory Bandwidth: 800 GB/s
+- CPUs: 128
+- CPU Memory: 503.5412 GB
+- CPU Memory Bandwidth: 3200 GB/s
 
 # Distillation (Teacher -> Student) Architecture Difference:
 
@@ -115,7 +115,7 @@ GPT2LMHeadModel(
 <br/>
 
 # Train Dataset
-Trained on 521,283,083 tokens from the [wikimedia/wikipedia](https://huggingface.co/datasets/wikimedia/wikipedia) dataset.
+Trained on 521,413,804 tokens from the [wikimedia/wikipedia](https://huggingface.co/datasets/wikimedia/wikipedia) dataset.
 
 - Num Samples: `990,000`
 - Subset: `20231101.en`
@@ -125,7 +125,7 @@ Trained on 521,283,083 tokens from the [wikimedia/wikipedia](https://huggingface
 # Training Objective
 
 ```
-DistillationObjective(logits_loss_component=LossComponent(label=logits, weight=1, loss_fn=kl), attn_loss_component=LossComponent(label=attn, weight=5, loss_fn=raw_mse, layer_mapper=layer-2, norm=layernorm_teacher_only_affine, projector=mlp_256_l2))
+DistillationObjective(logits_loss_component=LossComponent(label=logits, weight=1, loss_fn=kl), attn_loss_component=LossComponent(label=attn, weight=5, loss_fn=raw_mse, layer_mapper=layer-2, norm=layernorm_teacher_only_affine, projector=orthogonal))
 ```
 
 # Hyperparameters
@@ -141,8 +141,8 @@ The following hyperparameters were used during training:
 - optimizer: `Adam with betas=(0.9,0.999) and epsilon=1e-08`
 - lr_scheduler_type: `polynomial`
 - num_epochs: `1.0`
-- distillation_objective: `DistillationObjective(logits_loss_component=LossComponent(label=logits, weight=1, loss_fn=kl), attn_loss_component=LossComponent(label=attn, weight=5, loss_fn=raw_mse, layer_mapper=layer-2, norm=layernorm_teacher_only_affine, projector=mlp_256_l2))`
-- lr_scheduler: `<torch.optim.lr_scheduler.LambdaLR object at 0x7fb428e54250>`
+- distillation_objective: `DistillationObjective(logits_loss_component=LossComponent(label=logits, weight=1, loss_fn=kl), attn_loss_component=LossComponent(label=attn, weight=5, loss_fn=raw_mse, layer_mapper=layer-2, norm=layernorm_teacher_only_affine, projector=orthogonal))`
+- lr_scheduler: `<torch.optim.lr_scheduler.LambdaLR object at 0x7f5b3de02770>`
 - student_model_name_or_path: `None`
 - student_config_name_or_path: `distilbert/distilgpt2`
 - student_model_config: `None`
@@ -171,6 +171,6 @@ The following hyperparameters were used during training:
 
 # Framework Versions
 - Distily 0.5.0
-- Transformers 4.44.1
+- Transformers 4.44.2
 - Pytorch 2.4.0+cu121
-- Datasets 2.21.0
+- Datasets 2.18.0

config.json

@@ -40,7 +40,7 @@
     }
   },
   "torch_dtype": "bfloat16",
-  "transformers_version": "4.44.1",
+  "transformers_version": "4.44.2",
   "use_cache": true,
   "vocab_size": 50257
 }

generation_config.json

@@ -2,5 +2,5 @@
   "_from_model_config": true,
   "bos_token_id": 50256,
   "eos_token_id": 50256,
-  "transformers_version": "4.44.1"
+  "transformers_version": "4.44.2"
 }

logs/attn_norm=layernorm_teacher_only_affine, attn_projector=mlp_64_l2, attn_weight=5, learning_rate=0.0002, per_device_train_batch_size=16, warmup_ratio=0/events.out.tfevents.1725524926.a7e428977e35

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2507ee7055f3110e56d901fb8ccfa64a1cc7cee988afe516795f1f66b843173d
-size 1104318
+oid sha256:4618308bb131825bbcb34cbbabd60e97a9242af8823723efa76dd0dea577b93a
+size 1262872

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1c07b2b96c0f954cfca7b4623ca1585a1e76253e7da58e159be4569e2957144c
+oid sha256:23f433b1ccb03fa077bcf2597eea5479a6597bfc82fb3b82159f5a5f906d1f03
 size 163832792