Push model using huggingface_hub.

README.md

@@ -25,7 +25,7 @@ You can then generate text as follows:
 ```python
 from transformers import pipeline
 
-generator = pipeline("text-generation", model="nteku1//tmp/tmp3pjve72m/nteku1/Jon_GPT2L_PPO_epi_point1")
+generator = pipeline("text-generation", model="nteku1//tmp/tmpu05xwfmf/nteku1/Jon_GPT2L_PPO_epi_point1")
 outputs = generator("Hello, my llama is cute")
 ```
 
@@ -35,8 +35,8 @@ 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("nteku1//tmp/tmp3pjve72m/nteku1/Jon_GPT2L_PPO_epi_point1")
-model = AutoModelForCausalLMWithValueHead.from_pretrained("nteku1//tmp/tmp3pjve72m/nteku1/Jon_GPT2L_PPO_epi_point1")
+tokenizer = AutoTokenizer.from_pretrained("nteku1//tmp/tmpu05xwfmf/nteku1/Jon_GPT2L_PPO_epi_point1")
+model = AutoModelForCausalLMWithValueHead.from_pretrained("nteku1//tmp/tmpu05xwfmf/nteku1/Jon_GPT2L_PPO_epi_point1")
 
 inputs = tokenizer("Hello, my llama is cute", return_tensors="pt")
 outputs = model(**inputs, labels=inputs["input_ids"])

config.json

@@ -1,8 +1,8 @@
 {
   "accelerator_kwargs": {},
   "adap_kl_ctrl": true,
-  "backward_batch_size": 16,
-  "batch_size": 16,
+  "backward_batch_size": 1,
+  "batch_size": 256,
   "cliprange": 0.2,
   "cliprange_value": 0.2,
   "compare_steps": 1,
@@ -10,21 +10,21 @@
   "exp_name": "colab_kernel_launcher",
   "forward_batch_size": null,
   "gamma": 1,
-  "global_backward_batch_size": 16,
-  "global_batch_size": 16,
-  "gradient_accumulation_steps": 4,
+  "global_backward_batch_size": 1,
+  "global_batch_size": 256,
+  "gradient_accumulation_steps": 1,
   "horizon": 10000,
   "init_kl_coef": 0.2,
   "is_encoder_decoder": false,
   "is_peft_model": true,
   "kl_penalty": "kl",
   "lam": 0.95,
-  "learning_rate": 1.41e-05,
+  "learning_rate": 1.41e-08,
   "log_with": null,
   "max_grad_norm": null,
-  "mini_batch_size": 4,
+  "mini_batch_size": 1,
   "model_name": "Setpember/sft_gpt2_large",
-  "optimize_cuda_cache": true,
+  "optimize_cuda_cache": null,
   "optimize_device_cache": false,
   "ppo_epochs": 4,
   "project_kwargs": {},

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5bdcebd3247fc826505a69a4a257c8632e750e6d27d97042c1c23ef58ab45f30
+oid sha256:de06f6e25ee44c6eaa85f340acbe8ef8a7c9d94a01a79e04ebbf2b297dc25c51
 size 6652