diff --git a/medcat-trainer/webapp/api/api/metrics.py b/medcat-trainer/webapp/api/api/metrics.py
index 1cdfaf93d..b6e17e44b 100644
--- a/medcat-trainer/webapp/api/api/metrics.py
+++ b/medcat-trainer/webapp/api/api/metrics.py
@@ -245,7 +245,7 @@ def rename_meta_anns(self, meta_anns2rename=dict(), meta_ann_values2rename=dict(
         self.annotations = self._annotations()
         return
 
-    def _eval_model(self, model: nn.Module, data: List, config: ConfigMetaCAT, tokenizer: TokenizerWrapperBase) -> Dict:
+    def _eval_model(self, model: nn.Module, data: List, config: ConfigMetaCAT) -> Dict:
         device = torch.device(config.general.device)  # Create a torch device
         batch_size_eval = config.general.batch_size_eval
         pad_id = config.model.padding_idx
@@ -267,7 +267,7 @@ def _eval_model(self, model: nn.Module, data: List, config: ConfigMetaCAT, token
 
         with torch.no_grad():
             for i in range(num_batches):
-                x, cpos, attention_mask, y = create_batch_piped_data(data,
+                x, cpos, _, y = create_batch_piped_data(data,
                                                      i*batch_size_eval,
                                                      (i+1)*batch_size_eval,
                                                      device=device,
@@ -283,23 +283,22 @@ def _eval_model(self, model: nn.Module, data: List, config: ConfigMetaCAT, token
         return predictions
 
     def _eval(self, metacat_model, mct_export):
-        # TODO: Should be moved into
         g_config = metacat_model.config.general
-        t_config = metacat_model.config.train
-        t_config['test_size'] = 0
-        t_config['shuffle_data'] = False
-        t_config['prerequisites'] = {}
-        t_config['cui_filter'] = {}
 
         # Prepare the data
-        assert metacat_model.tokenizer is not None
-        data = prepare_from_json(mct_export, g_config['cntx_left'], g_config['cntx_right'], metacat_model.tokenizer,
-                                 cui_filter=t_config['cui_filter'],
-                                 replace_center=g_config['replace_center'], prerequisites=t_config['prerequisites'],
-                                 lowercase=g_config['lowercase'])
+        data = prepare_from_json(
+            mct_export,
+            g_config.cntx_left,
+            g_config.cntx_right,
+            metacat_model.mc.tokenizer,
+            cui_filter={},
+            replace_center=g_config.replace_center,
+            prerequisites={},
+            lowercase=g_config.lowercase
+        )
 
         # Check is the name there
-        category_name = g_config['category_name']
+        category_name = g_config.category_name
         if category_name not in data:
             warnings.warn(f"The meta_model {category_name} does not exist in this MedCATtrainer export.", UserWarning)
             return {category_name: f"{category_name} does not exist"}
@@ -307,12 +306,11 @@ def _eval(self, metacat_model, mct_export):
         data = data[category_name]
 
         # We already have everything, just get the data
-        category_value2id = g_config['category_value2id']
+        category_value2id = g_config.category_value2id
         data, _, _ = encode_category_values(data, existing_category_value2id=category_value2id)
         logger.info(_)
         # Run evaluation
-        assert metacat_model.tokenizer is not None
-        result = self._eval_model(metacat_model.model, data, config=metacat_model.config, tokenizer=metacat_model.tokenizer)
+        result = self._eval_model(metacat_model.mc.model, data, config=metacat_model.mc.config)
 
         return {'predictions': result, 'meta_values': _}
 
diff --git a/medcat-trainer/webapp/api/api/models.py b/medcat-trainer/webapp/api/api/models.py
index d8df36223..e89459752 100644
--- a/medcat-trainer/webapp/api/api/models.py
+++ b/medcat-trainer/webapp/api/api/models.py
@@ -45,52 +45,72 @@ class ModelPack(models.Model):
     last_modified_by = models.ForeignKey(settings.AUTH_USER_MODEL, on_delete=models.CASCADE, default=None, null=True)
 
     @transaction.atomic
-    def save(self, *args, **kwargs):
+    def save(self, *args, skip_load=False, **kwargs):
         is_new = self._state.adding
         if is_new:
             super().save(*args, **kwargs)
 
+        if skip_load:
+            super().save(*args, **kwargs)
+            return
+
         # Process the model pack
         logger.info('Loading model pack: %s', self.model_pack)
-        model_pack_name = str(self.model_pack).replace(".zip", "")
+        model_pack_path = self.model_pack.path
+
         try:
-            CAT.attempt_unpack(self.model_pack.path)
+            CAT.attempt_unpack(model_pack_path)
         except BadZipFile as exc:
             # potential for CRC-32 errors in Trainer process - ignore and still use
             logger.warning(f'Possibly corrupt cdb.dat decompressing {self.model_pack}\nFull Exception: {exc}')
-        unpacked_model_pack_path = self.model_pack.path.replace('.zip', '')
-        # attempt to load cdb
+
+        unpacked_model_pack_path = model_pack_path.replace('.zip', '')
+        unpacked_file_name = self.model_pack.file.name.replace('.zip', '')
+
+        # attempt to load cdb - use absolute paths for file existence checks
         try:
-            CAT.load_cdb(unpacked_model_pack_path)
+            # Check for v2 (directory) first
+            cdb_path_abs = os.path.join(unpacked_model_pack_path, 'cdb')
+            if os.path.exists(cdb_path_abs) and os.path.isdir(cdb_path_abs):
+                # v2: CDB is a directory
+                cdb_path_rel = os.path.join(unpacked_file_name, 'cdb')
+            else:
+                # v1: CDB is a file
+                cdb_path_abs = os.path.join(unpacked_model_pack_path, 'cdb.dat')
+                cdb_path_rel = os.path.join(unpacked_file_name, 'cdb.dat')
+                if not os.path.exists(cdb_path_abs):
+                    raise FileNotFoundError(f'CDB not found in model pack: {unpacked_model_pack_path}')
+
+            # Validate CDB can be loaded
+            CDB.load(cdb_path_abs)
+
+            # Create ConceptDB model with relative path (to MEDIA_ROOT)
             concept_db = ConceptDB()
-            unpacked_file_name = self.model_pack.file.name.replace('.zip', '')
-            # cdb path for v2
-            cdb_path = os.path.join(unpacked_file_name, 'cdb')
-            if not os.path.exists(cdb_path):
-                # cdb path for v1
-                cdb_path = os.path.join(unpacked_file_name, 'cdb.dat')
-            concept_db.cdb_file.name = cdb_path
+            concept_db.cdb_file.name = cdb_path_rel
             concept_db.name = f'{self.name}_CDB'
             concept_db.save(skip_load=True)
             self.concept_db = concept_db
         except Exception as exc:
             raise FileNotFoundError(f'Error loading the CDB from this model pack: {self.model_pack.path}') from exc
 
-        # Load Vocab, v2
-        vocab_path = os.path.join(unpacked_model_pack_path, "vocab")
-        if not os.path.exists(vocab_path):
+        # Load Vocab, v2 (directory) or v1 (file)
+        vocab_path_abs = os.path.join(unpacked_model_pack_path, "vocab")
+        vocab_path_rel = os.path.join(unpacked_file_name, "vocab")
+        if not os.path.exists(vocab_path_abs):
             # v1
-            vocab_path = os.path.join(unpacked_model_pack_path, "vocab.dat")
-        if os.path.exists(vocab_path):
-            Vocab.load(vocab_path)
+            vocab_path_abs = os.path.join(unpacked_model_pack_path, "vocab.dat")
+            vocab_path_rel = os.path.join(unpacked_file_name, "vocab.dat")
+        if os.path.exists(vocab_path_abs):
+            Vocab.load(vocab_path_abs)
             vocab = Vocabulary()
-            vocab.vocab_file.name = vocab_path.replace(f'{MEDIA_ROOT}/', '')
+            # Use relative path for saving to model
+            vocab.vocab_file.name = vocab_path_rel
             vocab.save(skip_load=True)
             self.vocab = vocab
         else:
             # DeID model packs do not have a vocab.dat file
             logger.warn('Error loading the Vocab from this model pack - '
-                        f'if this is a DeID model pack, this is expected: {vocab_path}')
+                        f'if this is a DeID model pack, this is expected: {vocab_path_abs}')
 
         # load MetaCATs
         try:
@@ -462,7 +482,7 @@ class Meta:
                                             help_text='Use a remote MedCAT service API for document processing instead of local models'\
                                                       '(note: interim model training is not supported for remote model service projects)')
     model_service_url = models.CharField(max_length=500, blank=True, null=True,
-                                         help_text='URL of the remote MedCAT service API (e.g., http://medcat-service:8003)')
+                                         help_text='URL of the remote MedCAT service API (e.g., http://medcat-service:8000)')
 
     def save(self, *args, **kwargs):
         # If using remote model service, skip local model validation
diff --git a/medcat-trainer/webapp/api/api/tests/test_data_utils.py b/medcat-trainer/webapp/api/api/tests/test_data_utils.py
index 87d230ee4..a7c5832bb 100644
--- a/medcat-trainer/webapp/api/api/tests/test_data_utils.py
+++ b/medcat-trainer/webapp/api/api/tests/test_data_utils.py
@@ -172,26 +172,14 @@ def test_upload_projects_export_with_modelpack(self, mock_exists, mock_vocab_loa
             # Mock all file operations
             mock_exists.return_value = False
             mock_load_addons.return_value = []
-            # Create a model pack - the save will be mocked to avoid actual file operations
+            # Create a model pack - use skip_load to avoid file validation
             from django.core.files.uploadedfile import SimpleUploadedFile
             modelpack = ModelPack(
                 name='test_modelpack',
                 model_pack=SimpleUploadedFile('test_modelpack.zip', b'fake zip')
             )
-            # Save with mocked file operations - it will fail on file loading but that's ok
-            try:
-                modelpack.save()
-            except (FileNotFoundError, Exception):
-                # If save fails, create it directly in the database
-                from django.utils import timezone
-                ModelPack.objects.filter(name='test_modelpack').delete()
-                modelpack = ModelPack.objects.create(
-                    name='test_modelpack',
-                    model_pack='test_modelpack.zip'
-                )
-                # Manually set the file field to avoid save() being called again
-                ModelPack.objects.filter(id=modelpack.id).update(model_pack='test_modelpack.zip')
-                modelpack.refresh_from_db()
+            # Save with skip_load=True to skip file validation
+            modelpack.save(skip_load=True)
 
             # Call the function
             upload_projects_export(
diff --git a/medcat-trainer/webapp/api/api/views.py b/medcat-trainer/webapp/api/api/views.py
index f9d9e0d36..713938e52 100644
--- a/medcat-trainer/webapp/api/api/views.py
+++ b/medcat-trainer/webapp/api/api/views.py
@@ -3,6 +3,7 @@
 import traceback
 from smtplib import SMTPException
 from tempfile import NamedTemporaryFile
+from typing import Any
 
 from background_task.models import Task, CompletedTask
 from django.contrib.auth.views import PasswordResetView
@@ -781,7 +782,8 @@ def serialize_task(task, state):
                 'projects': task.verbose_name.split('-')[1].split('_'),
                 'created_user': task.creator.username,
                 'create_time': task.run_at.strftime(dt_fmt),
-                'status': state
+                'error_msg': '\n'.join(task.last_error.split('\n')[-2:]),
+                'status': state,
             }
         running_reports = [serialize_task(t, 'running') for t in running_metrics_tasks_qs]
         for r, t in zip(running_reports, running_metrics_tasks_qs):
@@ -790,6 +792,8 @@ def serialize_task(task, state):
 
         comp_reports = [serialize_task(t, 'complete') for t in completed_metrics_tasks]
         for comp_task, comp_rep in zip(completed_metrics_tasks, comp_reports):
+            if comp_task.has_error():
+                comp_rep['status'] = 'Failed'
             pm_obj = ProjectMetrics.objects.filter(report_name_generated=comp_task.verbose_name).first()
             if pm_obj is not None and pm_obj.report_name is not None:
                 comp_rep['report_name'] = pm_obj.report_name
diff --git a/medcat-trainer/webapp/frontend/src/views/MetricsHome.vue b/medcat-trainer/webapp/frontend/src/views/MetricsHome.vue
index c4a83fb15..bc584da8e 100644
--- a/medcat-trainer/webapp/frontend/src/views/MetricsHome.vue
+++ b/medcat-trainer/webapp/frontend/src/views/MetricsHome.vue
@@ -5,7 +5,18 @@
                   :hover="true"
                   @click:row="loadMetrics"
                   hide-default-footer
-                  :items-per-page="-1">
+                  :items-per-page="-1"
+                  :row-props="getRowProps">
+      <template #item.report_id="{ item }">
+        <v-tooltip v-if="item.status === 'Failed' && item.error_msg" location="top">
+          <template #activator="{ props }">
+            <span v-bind="props" class="failed-row-tooltip-trigger">{{ item.report_id }}</span>
+          </template>
+          <span>{{ item.error_msg }}</span>
+        </v-tooltip>
+        <span v-else>{{ item.report_id }}</span>
+      </template>
+
       <template #item.projects="{ item }" >
         <div @click.stop>
           <v-runtime-template :template="projectsFormatter(item.projects)"></v-runtime-template>
@@ -22,6 +33,17 @@
         <span v-if="item.status === 'complete'">Complete
           <font-awesome-icon icon="check" class="status-icon success"></font-awesome-icon>
         </span>
+        <v-tooltip v-if="item.status === 'Failed' && item.error_msg" location="top">
+          <template #activator="{ props }">
+            <span v-bind="props" class="failed-row-tooltip-trigger">Failed
+              <font-awesome-icon icon="times-circle" class="status-icon error"></font-awesome-icon>
+            </span>
+          </template>
+          <span>{{ item.error_msg }}</span>
+        </v-tooltip>
+        <span v-else-if="item.status === 'Failed'">Failed
+          <font-awesome-icon icon="times-circle" class="status-icon error"></font-awesome-icon>
+        </span>
       </template>
       <template #item.cleanup="{ item }">
         <div @click.stop>
@@ -50,6 +72,10 @@
           Confirm complete metrics report deletion. To regenerate the report reselect the projects on the
           home screen and select metrics.
         </div>
+        <div v-if="confDeleteReportModal.status === 'Failed'">
+          Confirm failed metrics report deletion. To regenerate the report reselect the projects on the
+          home screen and select metrics.
+        </div>
       </template>
       <template #footer>
         <button class="btn btn-danger" @click="confRemoval">Confirm</button>
@@ -116,6 +142,10 @@ export default {
       })
     },
     loadMetrics (_, { item }) {
+      // Don't navigate if the report has failed
+      if (item.status === 'Failed') {
+        return
+      }
       this.$router.push({
         name: 'metrics',
         params: {
@@ -123,6 +153,13 @@ export default {
         }
       })
     },
+    getRowProps (data) {
+      // Apply disabled-row class to failed reports
+      if (data.item.status === 'Failed') {
+        return { class: 'disabled-row' }
+      }
+      return { class: '' }
+    },
     confRemoval () {
       const item = this.confDeleteReportModal
       this.$http.delete(`/api/metrics-job/${this.confDeleteReportModal.report_id}/`).then(_ => {
@@ -166,8 +203,35 @@ export default {
   padding-left: 3px;
 }
 
+.status-icon.error {
+  color: $danger;
+}
+
+:deep(.v-table > .v-table__wrapper > table > tbody > tr.disabled-row) {
+  pointer-events: none;
+  opacity: 0.6;
+  cursor: not-allowed !important;
+}
+
+:deep(.v-table > .v-table__wrapper > table > tbody > tr.disabled-row:hover) {
+  background-color: inherit !important;
+}
+
+:deep(.v-table > .v-table__wrapper > table > tbody > tr.disabled-row > td) {
+  pointer-events: none;
+}
+
+:deep(.v-table > .v-table__wrapper > table > tbody > tr.disabled-row > td > div) {
+  pointer-events: auto;
+}
+
+:deep(.v-table > .v-table__wrapper > table > tbody > tr.disabled-row .failed-row-tooltip-trigger) {
+  pointer-events: auto;
+  cursor: help;
+}
+
 .project-links {
-  color: #005EB8;
+  color: $primary-alt;
 
   &:hover {
     color: #fff;
diff --git a/v1/medcat/medcat/utils/meta_cat/data_utils.py b/v1/medcat/medcat/utils/meta_cat/data_utils.py
index 9272d1a2d..e3a0e054d 100644
--- a/v1/medcat/medcat/utils/meta_cat/data_utils.py
+++ b/v1/medcat/medcat/utils/meta_cat/data_utils.py
@@ -193,9 +193,11 @@ def find_alternate_classname(category_value2id: Dict, category_values: Set, alte
                 failed_to_find = True
             if failed_to_find:
                 raise Exception("The classes set in the config are not the same as the one found in the data. "
-                                "The classes present in the config vs the ones found in the data - "
-                                f"{set(category_value2id.keys())}, {category_values}. Additionally, ensure the "
-                                "populate the 'alternative_class_names' attribute to accommodate for variations.")
+                                f"The classes present in the config: {set(category_value2id.keys())} vs "
+                                f"the ones found in the data: {category_values}. Additionally, ensure to "
+                                "populate the 'alternative_class_names' attribute to accommodate for variations. "
+                                " This also could be due to the data not including enough examples of a given class"
+                                " after train / test splitting.")
     category_value2id = copy.deepcopy(updated_category_value2id)
     logger.info("Updated categoryvalue2id mapping - %s", category_value2id)
     return category_value2id