New model from https://wandb.ai/wandb/huggingtweets/runs/5uf3iuhw

README.md

@@ -47,15 +47,15 @@ The model was trained on tweets from Vaush.
 | Short tweets | 357 |
 | Tweets kept | 1956 |
 
-[Explore the data](https://wandb.ai/wandb/huggingtweets/runs/qq671ti1/artifacts), which is tracked with [W&B artifacts](https://docs.wandb.com/artifacts) at every step of the pipeline.
+[Explore the data](https://wandb.ai/wandb/huggingtweets/runs/kae7zvq9/artifacts), which is tracked with [W&B artifacts](https://docs.wandb.com/artifacts) at every step of the pipeline.
 
 ## Training procedure
 
 The model is based on a pre-trained [GPT-2](https://huggingface.co/gpt2) which is fine-tuned on @vaushv's tweets.
 
-Hyperparameters and metrics are recorded in the [W&B training run](https://wandb.ai/wandb/huggingtweets/runs/szxjdg8v) for full transparency and reproducibility.
+Hyperparameters and metrics are recorded in the [W&B training run](https://wandb.ai/wandb/huggingtweets/runs/5uf3iuhw) for full transparency and reproducibility.
 
-At the end of training, [the final model](https://wandb.ai/wandb/huggingtweets/runs/szxjdg8v/artifacts) is logged and versioned.
+At the end of training, [the final model](https://wandb.ai/wandb/huggingtweets/runs/5uf3iuhw/artifacts) is logged and versioned.
 
 ## How to use
 

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "gpt2",
+  "_name_or_path": "gpt2-medium",
   "activation_function": "gelu_new",
   "architectures": [
     "GPT2LMHeadModel"
@@ -12,11 +12,13 @@
   "layer_norm_epsilon": 1e-05,
   "model_type": "gpt2",
   "n_ctx": 1024,
-  "n_embd": 768,
-  "n_head": 12,
+  "n_embd": 1024,
+  "n_head": 16,
   "n_inner": null,
-  "n_layer": 12,
+  "n_layer": 24,
   "n_positions": 1024,
+  "n_special": 0,
+  "predict_special_tokens": true,
   "reorder_and_upcast_attn": false,
   "resid_pdrop": 0.1,
   "scale_attn_by_inverse_layer_idx": false,

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f1fa2c4ee8ba3f9f8fd053222669040d7edd1e3cd5c6ef6c7ca74c803e69add0
-size 510398013
+oid sha256:328acded8822a23d2aec8b74f0d34eaa506205b93f9ebc5e6a86eccee3651a3b
+size 1444569373

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b94df647ad1d1d5da072a9758c39f33230735ffcbce8e7db009b1591ff3d3db8
+oid sha256:298816583456fc2bc79f966f1282ab6883c82b04457493775dc433fd8a83cdbc
 size 3579