Update app.py

app.py

@@ -39,16 +39,16 @@ def generate(
  ctx,
  image_features,
  token_count=200,
- temperature=0.2,
+ temperature=1.0,
  top_p=0.3,
  presencePenalty = 0.1,
  countPenalty = 0.1,
 ):
  args = PIPELINE_ARGS(temperature = max(0.2, float(temperature)), top_p = float(top_p),
- alpha_frequency = countPenalty,
- alpha_presence = presencePenalty,
- token_ban = [], # ban the generation of some tokens
- token_stop = [0, 261]) # stop generation whenever you see any token here
+ alpha_frequency = countPenalty,
+ alpha_presence = presencePenalty,
+ token_ban = [], # ban the generation of some tokens
+ token_stop = [0, 261]) # stop generation whenever you see any token here
  ctx = ctx.strip()
  all_tokens = []
  out_last = 0
@@ -56,9 +56,11 @@ def generate(
  occurrence = {}
  for i in range(int(token_count)):
  if i == 0:
+ prefix_ids = pipeline.encode("User: ")
+ prefix_embs = model.w['emb.weight'][prefix_ids]
  input_ids = pipeline.encode(ctx)
  text_embs = model.w['emb.weight'][input_ids]
- input_embs = torch.cat((image_features, text_embs), dim=0)[-ctx_limit:]
+ input_embs = torch.cat((prefix_embs, image_features, text_embs), dim=0)[-ctx_limit:]
  out, state = model.forward(embs=input_embs, state=None)
  else:
  input_ids = [token]