add the trained model

app.py

@@ -3,7 +3,7 @@ import streamlit as st
 
 from transformers import AutoModelForCausalLM, AutoTokenizer
 
-checkpoint = "gpt2-large"
+checkpoint = "."
   
 tokenizer = AutoTokenizer.from_pretrained(checkpoint)
 @st.cache
@@ -13,10 +13,11 @@ def load_model(model_name):
 
 model = load_model(checkpoint)
 
-def infer(input_ids, max_tokens, temperature, top_k, top_p):
+def infer(input_ids, bad_words_ids, max_tokens, temperature, top_k, top_p):
 
     output_sequences = model.generate(
         input_ids=input_ids,
+        bad_words_ids = bad_words_ids,
         max_new_tokens=max_tokens,
         temperature=temperature,
         top_k=top_k,
@@ -39,18 +40,24 @@ st.write("This is a LLM that was fine-tuned on a dataset of investment memos to
 
 sent = st.text_area("Text", default_value, height = 400)
 
-max_tokens = st.sidebar.slider("Max Tokens", min_value = 32, max_value=512)
+max_tokens = st.sidebar.slider("Max Tokens", min_value = 16, max_value=64)
 temperature = st.sidebar.slider("Temperature", value = 0.8, min_value = 0.0, max_value=1.0, step=0.05)
 top_k = st.sidebar.slider("Top-k", min_value = 0, max_value=5, value = 4)
 top_p = st.sidebar.slider("Top-p", min_value = 0.0, max_value=1.0, step = 0.05, value = 0.9)
 
-encoded_prompt = tokenizer.encode(tokenizer.eos_token+sent, add_special_tokens=False, return_tensors="pt")
+# print(model.config.max_position_embeddings)
+
+encoded_prompt = tokenizer.encode(tokenizer.eos_token+sent, max_length=1024, return_tensors="pt", truncation=True)
+
+# get tokens of words that should not be generated
+bad_words_ids = tokenizer(["confidential", "angel.co", "angellist.com"], add_special_tokens=False).input_ids
+
 if encoded_prompt.size()[-1] == 0:
     input_ids = None
 else:
     input_ids = encoded_prompt
 
-output_sequences = infer(input_ids, max_tokens, temperature, top_k, top_p)
+output_sequences = infer(input_ids, bad_words_ids, max_tokens, temperature, top_k, top_p)
 
 for generated_sequence_idx, generated_sequence in enumerate(output_sequences):
     print(f"=== GENERATED SEQUENCE {generated_sequence_idx + 1} ===")
@@ -70,4 +77,4 @@ for generated_sequence_idx, generated_sequence in enumerate(output_sequences):
     generated_sequences.append(total_sequence)
     print(total_sequence)
 
-st.write(generated_sequences[-1])
+st.markdown(generated_sequences[-1])

config.json

@@ -0,0 +1,39 @@
+{
+  "_name_or_path": "gpt2-large",
+  "activation_function": "gelu_new",
+  "architectures": [
+    "GPT2LMHeadModel"
+  ],
+  "attn_pdrop": 0.1,
+  "bos_token_id": 50256,
+  "embd_pdrop": 0.1,
+  "eos_token_id": 50256,
+  "initializer_range": 0.02,
+  "layer_norm_epsilon": 1e-05,
+  "model_type": "gpt2",
+  "n_ctx": 1024,
+  "n_embd": 1280,
+  "n_head": 20,
+  "n_inner": null,
+  "n_layer": 36,
+  "n_positions": 1024,
+  "reorder_and_upcast_attn": false,
+  "resid_pdrop": 0.1,
+  "scale_attn_by_inverse_layer_idx": false,
+  "scale_attn_weights": true,
+  "summary_activation": null,
+  "summary_first_dropout": 0.1,
+  "summary_proj_to_labels": true,
+  "summary_type": "cls_index",
+  "summary_use_proj": true,
+  "task_specific_params": {
+    "text-generation": {
+      "do_sample": true,
+      "max_length": 50
+    }
+  },
+  "torch_dtype": "float32",
+  "transformers_version": "4.27.0.dev0",
+  "use_cache": true,
+  "vocab_size": 50257
+}

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90483692644e017cc03a2b5470912ab80369b4a79deb4f031e3fced773988bbb
+size 3134031497