Updated model metadata, Added temperature slider

app.py

@@ -22,7 +22,7 @@ def load_model(model_name):
     model = AutoModelForCausalLM.from_pretrained(model_name)
     return model, tokenizer
 
-def extend(input_text, max_size=20, top_k=50, top_p=0.95):
+def extend(input_text, max_size=20, top_k=50, top_p=0.95, temperature=0.7):
     if len(input_text) == 0:
         input_text = ""
 
@@ -40,9 +40,10 @@ def extend(input_text, max_size=20, top_k=50, top_p=0.95):
     input_ids=input_ids,
     max_length=max_size + len(encoded_prompt[0]),
     top_k=top_k, 
-    top_p=top_p, 
+    top_p=top_p,
+    temperature=temperature,
     do_sample=True,
-    repetition_penalty=5.0,
+    repetition_penalty=2.0,
     num_return_sequences=1)
 
     # Remove the batch dimension when returning multiple sequences
@@ -100,9 +101,10 @@ if __name__ == "__main__":
 
     st.sidebar.subheader("Configurable parameters")
 
-    max_len = st.sidebar.slider("Max-Length", 0, 512, 160,help="The maximum length of the sequence to be generated.")
+    max_len = st.sidebar.slider("Max-Length", 0, 256, 160,help="The maximum length of the sequence to be generated.")
     top_k = st.sidebar.slider("Top-K", 0, 100, 40, help="The number of highest probability vocabulary tokens to keep for top-k-filtering.")
     top_p = st.sidebar.slider("Top-P", 0.0, 1.0, 0.92, help="If set to float < 1, only the most probable tokens with probabilities that add up to top_p or higher are kept for generation.")
+    temperature = st.sidebar.slider("Temperature", 0.0, 1.0, 1.0, help="The value used to module the randomness of the output tokens.")
 
     if st.button("Generate Text"):
         with st.spinner(text="Generating results..."):
@@ -113,12 +115,13 @@ if __name__ == "__main__":
             result = extend(input_text=text_area,                         
                             max_size=int(max_len),                         
                             top_k=int(top_k),
-                            top_p=float(top_p))
+                            top_p=float(top_p),
+                            temperature=float(temperature))
 
             print("Done length: " + str(len(result)) + " bytes") 
             #<div class="rtl" dir="rtl" style="text-align:right;">
             st.markdown(f"<p dir=\"rtl\" style=\"text-align:right;\"> {result} </p>", unsafe_allow_html=True)
-            st.write("\n\nResult length: " + str(len(result)) + " bytes\n Random seed: " + str(random_seed) + "\ntop_k: " + str(top_k) + "\ntop_p: " + str(top_p) + "\nmax_len: " + str(max_len) + "\ndevice: " + str(device) + "\nn_gpu: " + str(n_gpu))
+            st.write("\n\nResult length: " + str(len(result)) + " bytes\n Random seed: " + str(random_seed) + "\ntop_k: " + str(top_k) + "\ntop_p: " + str(top_p) + "\ntemperature: " + str(temperature) + "\nmax_len: " + str(max_len) + "\ndevice: " + str(device) + "\nn_gpu: " + str(n_gpu))
             print(f"\"{result}\"")      
     
     st.markdown(    

model/added_tokens.json

@@ -1,6 +1,6 @@
-{
-  "<|endoftext|>": 50257,
-  "<|pad|>": 50260,
-  "<|startoftext|>": 50258,
-  "<|unknown|>": 50259
-}
+{
+  "<|endoftext|>": 50257,
+  "<|pad|>": 50260,
+  "<|startoftext|>": 50258,
+  "<|unknown|>": 50259
+}

model/config.json

@@ -1,54 +1,54 @@
-{
-  "_name_or_path": "Norod78/hebrew-gpt_neo-small",
-  "activation_function": "gelu_new",
-  "architectures": [
-    "GPTNeoForCausalLM"
-  ],
-  "attention_dropout": 0,
-  "attention_layers": [
-    "global",
-    "global",
-    "global",
-    "global",
-    "global",
-    "global",
-    "global",
-    "global",
-    "global",
-    "global",
-    "global",
-    "global"
-  ],
-  "attention_types": [
-    [
-      [
-        "global"
-      ],
-      12
-    ]
-  ],
-  "bos_token_id": 50256,
-  "embed_dropout": 0,
-  "eos_token_id": 50256,
-  "gradient_checkpointing": false,
-  "hidden_size": 768,
-  "initializer_range": 0.02,
-  "intermediate_size": null,
-  "layer_norm_epsilon": 1e-05,
-  "max_position_embeddings": 2048,
-  "model_type": "gpt_neo",
-  "num_heads": 12,
-  "num_layers": 12,
-  "pad_token_id": 50256,
-  "resid_dropout": 0,
-  "summary_activation": null,
-  "summary_first_dropout": 0.1,
-  "summary_proj_to_labels": true,
-  "summary_type": "cls_index",
-  "summary_use_proj": true,
-  "torch_dtype": "float32",
-  "transformers_version": "4.21.0",
-  "use_cache": true,
-  "vocab_size": 50261,
-  "window_size": 256
-}
+{
+  "_name_or_path": "./FantasyChildrenScifi-hebrew-gpt_neo-small/model",
+  "activation_function": "gelu_new",
+  "architectures": [
+    "GPTNeoForCausalLM"
+  ],
+  "attention_dropout": 0,
+  "attention_layers": [
+    "global",
+    "global",
+    "global",
+    "global",
+    "global",
+    "global",
+    "global",
+    "global",
+    "global",
+    "global",
+    "global",
+    "global"
+  ],
+  "attention_types": [
+    [
+      [
+        "global"
+      ],
+      12
+    ]
+  ],
+  "bos_token_id": 50256,
+  "embed_dropout": 0,
+  "eos_token_id": 50256,
+  "gradient_checkpointing": false,
+  "hidden_size": 768,
+  "initializer_range": 0.02,
+  "intermediate_size": null,
+  "layer_norm_epsilon": 1e-05,
+  "max_position_embeddings": 2048,
+  "model_type": "gpt_neo",
+  "num_heads": 12,
+  "num_layers": 12,
+  "pad_token_id": 50256,
+  "resid_dropout": 0,
+  "summary_activation": null,
+  "summary_first_dropout": 0.1,
+  "summary_proj_to_labels": true,
+  "summary_type": "cls_index",
+  "summary_use_proj": true,
+  "torch_dtype": "float32",
+  "transformers_version": "4.21.0",
+  "use_cache": true,
+  "vocab_size": 50261,
+  "window_size": 256
+}

model/special_tokens_map.json

@@ -1,6 +1,6 @@
-{
-  "bos_token": "<|startoftext|>",
-  "eos_token": "<|endoftext|>",
-  "pad_token": "<|pad|>",
-  "unk_token": "<unk>"
-}
+{
+  "bos_token": "<|startoftext|>",
+  "eos_token": "<|endoftext|>",
+  "pad_token": "<|pad|>",
+  "unk_token": "<unk>"
+}

model/tokenizer_config.json

@@ -1,22 +1,22 @@
-{
-  "add_bos_token": false,
-  "add_prefix_space": false,
-  "bos_token": "<|startoftext|>",
-  "do_lower_case": false,
-  "eos_token": "<|endoftext|>",
-  "errors": "replace",
-  "full_tokenizer_file": null,
-  "max_len": 1024,
-  "name_or_path": "Norod78/hebrew-gpt_neo-small",
-  "pad_token": "<|pad|>",
-  "special_tokens_map_file": "special_tokens_map.json",
-  "tokenizer_class": "GPT2Tokenizer",
-  "unk_token": {
-    "__type": "AddedToken",
-    "content": "<|endoftext|>",
-    "lstrip": false,
-    "normalized": true,
-    "rstrip": false,
-    "single_word": false
-  }
-}
+{
+  "add_bos_token": false,
+  "add_prefix_space": false,
+  "bos_token": "<|startoftext|>",
+  "do_lower_case": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "full_tokenizer_file": null,
+  "max_len": 1024,
+  "name_or_path": "./FantasyChildrenScifi-hebrew-gpt_neo-small/model",
+  "pad_token": "<|pad|>",
+  "special_tokens_map_file": "special_tokens_map.json",
+  "tokenizer_class": "GPT2Tokenizer",
+  "unk_token": {
+    "__type": "AddedToken",
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  }
+}