names adjustment (#10)

Adjust hyperparameters and comments

Co-authored-by: RichardoLuo <[email protected]>

README.md

@@ -12,8 +12,8 @@ This repository is the official implementation of [Enhance-A-Video: Better Gener
 Install the dependencies:
 
 ```bash
-conda create -n feta python=3.10
-conda activate feta
+conda create -n enhanceAvideo python=3.10
+conda activate enhanceAvideo
 pip install -r requirements.txt
 ```
 

cogvideox.py

@@ -4,21 +4,22 @@
 
 from enhance_a_video import enable_enhance, inject_feta_for_cogvideox, set_enhance_weight
 
-pipe = CogVideoXPipeline.from_pretrained("THUDM/CogVideoX-5b", torch_dtype=torch.bfloat16)
+pipe = CogVideoXPipeline.from_pretrained("THUDM/CogVideoX-2b", torch_dtype=torch.float16)
 
 pipe.to("cuda")
 # pipe.enable_sequential_cpu_offload()
 pipe.vae.enable_slicing()
 # pipe.vae.enable_tiling()
 
-# ============ FETA ============
+# ============ Enhance-A-Video ============
 # comment the following if you want to use the original model
 inject_feta_for_cogvideox(pipe.transformer)
+# enhance_weight can be adjusted for better visual quality
 set_enhance_weight(1)
 enable_enhance()
-# ============ FETA ============
+# ============ Enhance-A-Video ============
 
-prompt = "A Japanese tram glides through the snowy streets of a city, its sleek design cutting through the falling snowflakes with grace."
+prompt = "A cute happy Corgi playing in park"
 
 video_generate = pipe(
     prompt=prompt,

enhance_a_video/models/cogvideox.py

@@ -120,10 +120,10 @@ def __call__(
             if not attn.is_cross_attention:
                 key[:, :, text_seq_length:] = apply_rotary_emb(key[:, :, text_seq_length:], image_rotary_emb)
 
-        # ========== FETA ==========
+        # ========== Enhance-A-Video ==========
         if is_enhance_enabled():
             feta_scores = self._get_feta_scores(attn, query, key, head_dim, text_seq_length)
-        # ========== FETA ==========
+        # ========== Enhance-A-Video ==========
 
         hidden_states = F.scaled_dot_product_attention(
             query, key, value, attn_mask=attention_mask, dropout_p=0.0, is_causal=False
@@ -140,9 +140,9 @@ def __call__(
             [text_seq_length, hidden_states.size(1) - text_seq_length], dim=1
         )
 
-        # ========== FETA ==========
+        # ========== Enhance-A-Video ==========
         if is_enhance_enabled():
             hidden_states = hidden_states * feta_scores
-        # ========== FETA ==========
+        # ========== Enhance-A-Video ==========
 
         return hidden_states, encoder_hidden_states

enhance_a_video/models/hunyuanvideo.py

@@ -114,10 +114,10 @@ def __call__(
                 query = apply_rotary_emb(query, image_rotary_emb)
                 key = apply_rotary_emb(key, image_rotary_emb)
 
-        # ========== FETA ==========
+        # ========== Enhance-A-Video ==========
         if is_enhance_enabled():
             feta_scores = self._get_feta_scores(attn, query, key, encoder_hidden_states)
-        # ========== FETA ==========
+        # ========== Enhance-A-Video ==========
 
         # 4. Encoder condition QKV projection and normalization
         if attn.add_q_proj is not None and encoder_hidden_states is not None:
@@ -159,9 +159,9 @@ def __call__(
             if getattr(attn, "to_add_out", None) is not None:
                 encoder_hidden_states = attn.to_add_out(encoder_hidden_states)
 
-        # ========== FETA ==========
+        # ========== Enhance-A-Video ==========
         if is_enhance_enabled():
             hidden_states = hidden_states * feta_scores
-        # ========== FETA ==========
+        # ========== Enhance-A-Video ==========
 
         return hidden_states, encoder_hidden_states

hunyuanvideo.py

@@ -16,12 +16,13 @@
 pipe.vae.enable_tiling()
 # pipe.vae.enable_tiling()
 
-# ============ FETA ============
+# ============ Enhance-A-Video ============
 # comment the following if you want to use the original model
 inject_feta_for_hunyuanvideo(pipe.transformer)
+# enhance_weight can be adjusted for better visual quality
 set_enhance_weight(4)
 enable_enhance()
-# ============ FETA ============
+# ============ Enhance-A-Video ============
 
 prompt = "A focused baseball player stands in the dugout, gripping his bat with determination, wearing a classic white jersey with blue pinstripes and a matching cap. The sunlight casts dramatic shadows across his face, highlighting his intense gaze as he prepares for the game. His hands, wrapped in black batting gloves, firmly hold the bat, showcasing his readiness and anticipation. The background reveals the bustling stadium, with blurred fans and vibrant green field, creating an atmosphere of excitement and competition. As he adjusts his stance, the player's concentration and passion for the sport are palpable, embodying the spirit of baseball."