Merge branch 'main' of github.com:herobd/FUDGE into main

README.md

@@ -15,7 +15,7 @@ This code is licensed under GNU GPL v3. If you would like it distributed to you
 
 ## Pre-trained Model weights
 
-See "Releases"
+See `pretrained.tar.gz` in "Releases"
 
 
 ## Reproducability instructions
@@ -34,11 +34,11 @@ FUNSD: `python train.py -c configs/cf_FUNSDLines_detect_augR_staggerLighter.json
 NAF: `python train.py -c configs/cf_NAF_detect_augR_staggerLighter.json`
 
 ### Training the full networks
-FUNSD: `python train.py -c cf_FUNSDLines_pair_graph663rv_new.json`
+FUNSD: `python train.py -c configs/cf_FUNSDLines_pair_graph663rv_new.json`
 
-Word-FUDGE: `python train.py -c cf_FUNSDWords_pair_graph663rv_new.json` [ACTUALLY, THERE WAS AN ERROR IN MY ORIGINAL EVAL. NEW RESULTS FORTHCOMING (for Word-FUDGE)]
+Word-FUDGE: `python train.py -c configs/cf_FUNSDLinesAndWords_pair_graph663rv_new.json` 
 
-NAF: `python train.py -c cf_NAF_pair_graph663rv_new.json`
+NAF: `python train.py -c configs/cf_NAF_pair_graph663rv_new.json`
 
 The ablation uses the following configs:
 * cf_FUNSDLines_pair_binary333rv_new.json
@@ -50,7 +50,7 @@ The ablation uses the following configs:
 #### Wait, how long does this take to train?
 If trained to the full 700,000 iterations, it takes a couple weeks, depending on your GPU. I used a batch size of 1 due to hardware limitations. I also hard-coded the batch size of 1, so you have to as well (GCNs handle batches specially and I didn't want to code that up).
 
-However, from an experiment I ran, I think you can get the same results with only 250,000 iterations by accumulating the gradient to pretend a batch size of 5. This is done by adding `"accum_grad_steps": 5` to `trainer` in the config json. Yes, that means it only updates the weights 50,000 times.
+However, from an experiment I ran, I think you can get the same results with only 250,000 iterations by accumulating the gradient to pretend a batch size of 5. This is done by adding `"accum_grad_steps": 5` to `trainer` in the config json. Yes, that means it only updates the weights 50,000 times. It never hurts to train a bit more, it doesn't overfit in my experience.
 
 ### Evaluating
 

configs/cf_FUNSDLinesAndWords_pair_graph663rv_new.json

@@ -83,7 +83,6 @@
         "use_word_bbs_gt": 0.5,
         "max_use_pred": 0.33,
         "valid_with_gt": true,
-
 
         "use_learning_schedule": "multi_rise then ramp_to_swa",
 "ramp_down_steps": 10000,
@@ -98,7 +97,7 @@
         "start_frozen": true,
         "relationship_proposal": "feature_nn",
         "percent_rel_to_keep": 0.5,
-        "max_rel_to_keep": 800,
+        "max_rel_to_keep": 900,
         "roi_batch_size": 400,
 
 	"use_rel_shape_feats": "corner",
@@ -124,7 +123,6 @@
         "featurizer_bb_start_w": 10,
         "bb_featurizer_conv": ["sep64","sep64","M","sep128"],
 
-
         "reintroduce_visual_features": "fixed map",
         "graph_config": [
 	    {