Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -356,40 +356,49 @@ def create_cluster_dataframes(processed_df):
|
|
356 |
|
357 |
|
358 |
|
359 |
-
|
360 |
from transformers import GPTNeoForCausalLM, GPT2Tokenizer
|
361 |
def generate_project_proposal(prompt): # Generate the proposal
|
362 |
# model_Name = "EleutherAI/gpt-neo-2.7B"
|
|
|
|
|
363 |
model_Name = "EleutherAI/gpt-neo-1.3B"
|
|
|
364 |
|
365 |
consoleMessage_and_Print(f"Trying to access {model_Name} model. The Prompt is: \n{prompt}")
|
366 |
|
367 |
model = GPTNeoForCausalLM.from_pretrained(model_Name)
|
368 |
tokenizer = GPT2Tokenizer.from_pretrained(model_Name)
|
369 |
-
model_max_token_limit =
|
370 |
|
371 |
try:
|
372 |
# input_ids = tokenizer.encode(prompt, return_tensors="pt")
|
373 |
# Truncate the prompt to fit within the model's input limits
|
374 |
# Adjust as per your model's limit
|
375 |
-
input_ids = tokenizer.encode(prompt, return_tensors="pt", truncation=True, max_length = model_max_token_limit/2)
|
376 |
|
377 |
|
378 |
print("Input IDs shape:", input_ids.shape)
|
|
|
|
|
|
|
|
|
379 |
# Generate the output
|
380 |
output = model.generate(
|
381 |
input_ids,
|
|
|
382 |
max_new_tokens = model_max_token_limit,
|
383 |
num_return_sequences=1,
|
384 |
no_repeat_ngram_size=2,
|
385 |
-
temperature=
|
386 |
-
|
|
|
387 |
)
|
388 |
print("Output shape:", output.shape)
|
389 |
|
390 |
|
391 |
# Decode the output to text
|
392 |
-
full_returned_segment = tokenizer.decode(output[0], skip_special_tokens=True)
|
393 |
|
394 |
# Slice off the input part if the input length is known
|
395 |
input_length = input_ids.shape[1]
|
@@ -410,10 +419,9 @@ def generate_project_proposal(prompt): # Generate the proposal
|
|
410 |
|
411 |
|
412 |
|
413 |
-
import copy
|
414 |
-
|
415 |
|
416 |
|
|
|
417 |
def create_project_proposals(budget_cluster_df, problem_cluster_df, location_clusters, problem_clusters):
|
418 |
consoleMessage_and_Print("\n Starting function: create_project_proposals")
|
419 |
proposals = {}
|
|
|
356 |
|
357 |
|
358 |
|
359 |
+
from random import uniform
|
360 |
from transformers import GPTNeoForCausalLM, GPT2Tokenizer
|
361 |
def generate_project_proposal(prompt): # Generate the proposal
|
362 |
# model_Name = "EleutherAI/gpt-neo-2.7B"
|
363 |
+
# tempareCHUR = uniform(0.3,0.6)
|
364 |
+
|
365 |
model_Name = "EleutherAI/gpt-neo-1.3B"
|
366 |
+
tempareCHUR = uniform(0.5,0.8)
|
367 |
|
368 |
consoleMessage_and_Print(f"Trying to access {model_Name} model. The Prompt is: \n{prompt}")
|
369 |
|
370 |
model = GPTNeoForCausalLM.from_pretrained(model_Name)
|
371 |
tokenizer = GPT2Tokenizer.from_pretrained(model_Name)
|
372 |
+
model_max_token_limit = 2047
|
373 |
|
374 |
try:
|
375 |
# input_ids = tokenizer.encode(prompt, return_tensors="pt")
|
376 |
# Truncate the prompt to fit within the model's input limits
|
377 |
# Adjust as per your model's limit
|
378 |
+
input_ids = tokenizer.encode(prompt, return_tensors="pt", truncation=True, max_length = int(model_max_token_limit/2) )
|
379 |
|
380 |
|
381 |
print("Input IDs shape:", input_ids.shape)
|
382 |
+
|
383 |
+
pad_tokenId = tokenizer.pad_token_id if tokenizer.pad_token_id is not None else tokenizer.eos_token_id # Padding with EOS token may always be great
|
384 |
+
attentionMask = input_ids.ne(pad_tokenId).long()
|
385 |
+
|
386 |
# Generate the output
|
387 |
output = model.generate(
|
388 |
input_ids,
|
389 |
+
min_length = int(model_max_token_limit/3), # minimum length of the generated output
|
390 |
max_new_tokens = model_max_token_limit,
|
391 |
num_return_sequences=1,
|
392 |
no_repeat_ngram_size=2,
|
393 |
+
temperature=tempareCHUR,
|
394 |
+
attention_mask=attentionMask, # This was previously not being used
|
395 |
+
pad_token_id=pad_tokenId
|
396 |
)
|
397 |
print("Output shape:", output.shape)
|
398 |
|
399 |
|
400 |
# Decode the output to text
|
401 |
+
# full_returned_segment = tokenizer.decode(output[0], skip_special_tokens=True)
|
402 |
|
403 |
# Slice off the input part if the input length is known
|
404 |
input_length = input_ids.shape[1]
|
|
|
419 |
|
420 |
|
421 |
|
|
|
|
|
422 |
|
423 |
|
424 |
+
import copy
|
425 |
def create_project_proposals(budget_cluster_df, problem_cluster_df, location_clusters, problem_clusters):
|
426 |
consoleMessage_and_Print("\n Starting function: create_project_proposals")
|
427 |
proposals = {}
|