Push model using huggingface_hub.

README.md

@@ -26,7 +26,7 @@ You can then generate text as follows:
 ```python
 from transformers import pipeline
 
-generator = pipeline("text-generation", model="arianaazarbal//tmp/tmpu6dj7w01/arianaazarbal/ppo-finetuned-model")
+generator = pipeline("text-generation", model="arianaazarbal//tmp/tmp3vx9jc19/arianaazarbal/ppo-finetuned-model")
 outputs = generator("Hello, my llama is cute")
 ```
 
@@ -36,8 +36,8 @@ If you want to use the model for training or to obtain the outputs from the valu
 from transformers import AutoTokenizer
 from trl import AutoModelForCausalLMWithValueHead
 
-tokenizer = AutoTokenizer.from_pretrained("arianaazarbal//tmp/tmpu6dj7w01/arianaazarbal/ppo-finetuned-model")
-model = AutoModelForCausalLMWithValueHead.from_pretrained("arianaazarbal//tmp/tmpu6dj7w01/arianaazarbal/ppo-finetuned-model")
+tokenizer = AutoTokenizer.from_pretrained("arianaazarbal//tmp/tmp3vx9jc19/arianaazarbal/ppo-finetuned-model")
+model = AutoModelForCausalLMWithValueHead.from_pretrained("arianaazarbal//tmp/tmp3vx9jc19/arianaazarbal/ppo-finetuned-model")
 
 inputs = tokenizer("Hello, my llama is cute", return_tensors="pt")
 outputs = model(**inputs, labels=inputs["input_ids"])

adapter_config.json

@@ -24,10 +24,10 @@
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "o_proj",
-    "v_proj",
     "q_proj",
-    "k_proj"
+    "v_proj",
+    "k_proj",
+    "o_proj"
   ],
   "task_type": "CAUSAL_LM",
   "trainable_token_indices": null,

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d0deea2490a4e8cf6b88413eda4ef11940cc91705fb6d75968dbe880f41979bc
+oid sha256:c9716c49b3f07fea478b4c3ebb4502749ef9192c782e39e427095ce34442bbb9
 size 40400200

config.json

@@ -2,7 +2,7 @@
   "accelerator_kwargs": {},
   "adap_kl_ctrl": true,
   "backward_batch_size": 2,
-  "batch_size": 2,
+  "batch_size": 4,
   "cliprange": 0.2,
   "cliprange_value": 0.2,
   "compare_steps": 1,
@@ -12,7 +12,7 @@
   "forward_batch_size": null,
   "gamma": 1.0,
   "global_backward_batch_size": 2,
-  "global_batch_size": 2,
+  "global_batch_size": 4,
   "gradient_accumulation_steps": 1,
   "gradient_checkpointing": false,
   "horizon": 10000.0,
@@ -28,7 +28,7 @@
   "model_name": "gpt2",
   "optimize_cuda_cache": null,
   "optimize_device_cache": false,
-  "ppo_epochs": 2,
+  "ppo_epochs": 6,
   "project_kwargs": {},
   "push_to_hub_if_best_kwargs": {},
   "query_dataset": "stanfordnlp/imdb",

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1db72a3ab3141e17bd54bfb9fb31d38b29129500c4b1b57f42fdb659267dff1e
+oid sha256:6ad3ce68c38ef303e0ed48af5f36ef1a39cbd73c5124a4f21f59344f5bb6b150
 size 16277